Merge branch 'main' into bugfix/support-llm-managers-that-use-model

* before kickoff breaks if inputs are none.

* improve none type

* Fix failing tests

* add tests for new code

* Fix failing test

* drop extra comments

* clean up based on eduardo feedback

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -65,7 +65,6 @@ body:
         - '3.10'
         - '3.11'
         - '3.12'
-        - '3.13'
     validations:
       required: true
   - type: input
@@ -113,4 +112,4 @@ body:
       label: Additional context
       description: Add any other context about the problem here.
     validations:
-      required: true
+      required: true

.github/workflows/linter.yml

@@ -13,4 +13,4 @@ jobs:
           pip install ruff
 
       - name: Run Ruff Linter
-        run: ruff check --exclude "templates","__init__.py"
+        run: ruff check

.github/workflows/stale.yml

@@ -1,5 +1,10 @@
 name: Mark stale issues and pull requests
 
+permissions:
+  contents: write
+  issues: write
+  pull-requests: write
+
 on:
   schedule:
     - cron: '10 12 * * *'
@@ -8,9 +13,6 @@ on:
 jobs:
   stale:
     runs-on: ubuntu-latest
-    permissions:
-      issues: write
-      pull-requests: write
     steps:
     - uses: actions/stale@v9
       with:

.github/workflows/tests.yml

@@ -23,7 +23,7 @@ jobs:
 
 
       - name: Set up Python
-        run: uv python install 3.11.9
+        run: uv python install 3.12.8
 
       - name: Install the project
         run: uv sync --dev --all-extras

.gitignore

@@ -21,3 +21,4 @@ crew_tasks_output.json
 .mypy_cache
 .ruff_cache
 .venv
+agentops.log

.pre-commit-config.yaml

@@ -1,9 +1,7 @@
 repos:
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.4.4
+    rev: v0.8.2
     hooks:
       - id: ruff
         args: ["--fix"]
-        exclude: "templates"
       - id: ruff-format
-        exclude: "templates"

.ruff.toml

@@ -0,0 +1,9 @@
+exclude = [
+    "templates",
+    "__init__.py",
+]
+
+[lint]
+select = [
+    "I", # isort rules
+]

README.md

@@ -4,7 +4,7 @@
 
 # **CrewAI**
 
-🤖 **CrewAI**: Cutting-edge framework for orchestrating role-playing, autonomous AI agents. By fostering collaborative intelligence, CrewAI empowers agents to work together seamlessly, tackling complex tasks.
+🤖 **CrewAI**: Production-grade framework for orchestrating sophisticated AI agent systems. From simple automations to complex real-world applications, CrewAI provides precise control and deep customization. By fostering collaborative intelligence through flexible, production-ready architecture, CrewAI empowers agents to work together seamlessly, tackling complex business challenges with predictable, consistent results.
 
 <h3>
 
@@ -22,13 +22,17 @@
 - [Why CrewAI?](#why-crewai)
 - [Getting Started](#getting-started)
 - [Key Features](#key-features)
+- [Understanding Flows and Crews](#understanding-flows-and-crews)
+- [CrewAI vs LangGraph](#how-crewai-compares)
 - [Examples](#examples)
   - [Quick Tutorial](#quick-tutorial)
   - [Write Job Descriptions](#write-job-descriptions)
   - [Trip Planner](#trip-planner)
   - [Stock Analysis](#stock-analysis)
+  - [Using Crews and Flows Together](#using-crews-and-flows-together)
 - [Connecting Your Crew to a Model](#connecting-your-crew-to-a-model)
 - [How CrewAI Compares](#how-crewai-compares)
+- [Frequently Asked Questions (FAQ)](#frequently-asked-questions-faq)
 - [Contribution](#contribution)
 - [Telemetry](#telemetry)
 - [License](#license)
@@ -36,22 +40,51 @@
 ## Why CrewAI?
 
 The power of AI collaboration has too much to offer.
-CrewAI is designed to enable AI agents to assume roles, share goals, and operate in a cohesive unit - much like a well-oiled crew. Whether you're building a smart assistant platform, an automated customer service ensemble, or a multi-agent research team, CrewAI provides the backbone for sophisticated multi-agent interactions.
+CrewAI is a standalone framework, built from the ground up without dependencies on Langchain or other agent frameworks. It's designed to enable AI agents to assume roles, share goals, and operate in a cohesive unit - much like a well-oiled crew. Whether you're building a smart assistant platform, an automated customer service ensemble, or a multi-agent research team, CrewAI provides the backbone for sophisticated multi-agent interactions.
 
 ## Getting Started
 
+### Learning Resources
+
+Learn CrewAI through our comprehensive courses:
+- [Multi AI Agent Systems with CrewAI](https://www.deeplearning.ai/short-courses/multi-ai-agent-systems-with-crewai/) - Master the fundamentals of multi-agent systems
+- [Practical Multi AI Agents and Advanced Use Cases](https://www.deeplearning.ai/short-courses/practical-multi-ai-agents-and-advanced-use-cases-with-crewai/) - Deep dive into advanced implementations
+
+### Understanding Flows and Crews
+
+CrewAI offers two powerful, complementary approaches that work seamlessly together to build sophisticated AI applications:
+
+1. **Crews**: Teams of AI agents with true autonomy and agency, working together to accomplish complex tasks through role-based collaboration. Crews enable:
+   - Natural, autonomous decision-making between agents
+   - Dynamic task delegation and collaboration
+   - Specialized roles with defined goals and expertise
+   - Flexible problem-solving approaches
+
+2. **Flows**: Production-ready, event-driven workflows that deliver precise control over complex automations. Flows provide:
+   - Fine-grained control over execution paths for real-world scenarios
+   - Secure, consistent state management between tasks
+   - Clean integration of AI agents with production Python code
+   - Conditional branching for complex business logic
+
+The true power of CrewAI emerges when combining Crews and Flows. This synergy allows you to:
+- Build complex, production-grade applications
+- Balance autonomy with precise control
+- Handle sophisticated real-world scenarios
+- Maintain clean, maintainable code structure
+
+### Getting Started with Installation
+
 To get started with CrewAI, follow these simple steps:
 
 ### 1. Installation
 
-Ensure you have Python >=3.10 <=3.13 installed on your system. CrewAI uses [UV](https://docs.astral.sh/uv/) for dependency management and package handling, offering a seamless setup and execution experience.
+Ensure you have Python >=3.10 <3.13 installed on your system. CrewAI uses [UV](https://docs.astral.sh/uv/) for dependency management and package handling, offering a seamless setup and execution experience.
 
 First, install CrewAI:
 
 ```shell
 pip install crewai
 ```
-
 If you want to install the 'crewai' package along with its optional features that include additional tools for agents, you can do so by using the following command:
 
 ```shell
@@ -59,6 +92,22 @@ pip install 'crewai[tools]'
 ```
 The command above installs the basic package and also adds extra components which require more dependencies to function.
 
+### Troubleshooting Dependencies
+
+If you encounter issues during installation or usage, here are some common solutions:
+
+#### Common Issues
+
+1. **ModuleNotFoundError: No module named 'tiktoken'**
+   - Install tiktoken explicitly: `pip install 'crewai[embeddings]'`
+   - If using embedchain or other tools: `pip install 'crewai[tools]'`
+
+2. **Failed building wheel for tiktoken**
+   - Ensure Rust compiler is installed (see installation steps above)
+   - For Windows: Verify Visual C++ Build Tools are installed
+   - Try upgrading pip: `pip install --upgrade pip`
+   - If issues persist, use a pre-built wheel: `pip install tiktoken --prefer-binary`
+
 ### 2. Setting Up Your Crew with the YAML Configuration
 
 To create a new CrewAI project, run the following CLI (Command Line Interface) command:
@@ -264,13 +313,16 @@ In addition to the sequential process, you can use the hierarchical process, whi
 
 ## Key Features
 
-- **Role-Based Agent Design**: Customize agents with specific roles, goals, and tools.
-- **Autonomous Inter-Agent Delegation**: Agents can autonomously delegate tasks and inquire amongst themselves, enhancing problem-solving efficiency.
-- **Flexible Task Management**: Define tasks with customizable tools and assign them to agents dynamically.
-- **Processes Driven**: Currently only supports `sequential` task execution and `hierarchical` processes, but more complex processes like consensual and autonomous are being worked on.
-- **Save output as file**: Save the output of individual tasks as a file, so you can use it later.
-- **Parse output as Pydantic or Json**: Parse the output of individual tasks as a Pydantic model or as a Json if you want to.
-- **Works with Open Source Models**: Run your crew using Open AI or open source models refer to the [Connect CrewAI to LLMs](https://docs.crewai.com/how-to/LLM-Connections/) page for details on configuring your agents' connections to models, even ones running locally!
+**Note**: CrewAI is a standalone framework built from the ground up, without dependencies on Langchain or other agent frameworks.
+
+- **Deep Customization**: Build sophisticated agents with full control over the system - from overriding inner prompts to accessing low-level APIs. Customize roles, goals, tools, and behaviors while maintaining clean abstractions.
+- **Autonomous Inter-Agent Delegation**: Agents can autonomously delegate tasks and inquire amongst themselves, enabling complex problem-solving in real-world scenarios.
+- **Flexible Task Management**: Define and customize tasks with granular control, from simple operations to complex multi-step processes.
+- **Production-Grade Architecture**: Support for both high-level abstractions and low-level customization, with robust error handling and state management.
+- **Predictable Results**: Ensure consistent, accurate outputs through programmatic guardrails, agent training capabilities, and flow-based execution control. See our [documentation on guardrails](https://docs.crewai.com/how-to/guardrails/) for implementation details.
+- **Model Flexibility**: Run your crew using OpenAI or open source models with production-ready integrations. See [Connect CrewAI to LLMs](https://docs.crewai.com/how-to/LLM-Connections/) for detailed configuration options.
+- **Event-Driven Flows**: Build complex, real-world workflows with precise control over execution paths, state management, and conditional logic.
+- **Process Orchestration**: Achieve any workflow pattern through flows - from simple sequential and hierarchical processes to complex, custom orchestration patterns with conditional branching and parallel execution.
 
 ![CrewAI Mind Map](./docs/crewAI-mindmap.png "CrewAI Mind Map")
 
@@ -305,6 +357,98 @@ You can test different real life examples of AI crews in the [CrewAI-examples re
 
 [![Stock Analysis](https://img.youtube.com/vi/e0Uj4yWdaAg/maxresdefault.jpg)](https://www.youtube.com/watch?v=e0Uj4yWdaAg "Stock Analysis")
 
+### Using Crews and Flows Together
+
+CrewAI's power truly shines when combining Crews with Flows to create sophisticated automation pipelines. Here's how you can orchestrate multiple Crews within a Flow:
+
+```python
+from crewai.flow.flow import Flow, listen, start, router
+from crewai import Crew, Agent, Task
+from pydantic import BaseModel
+
+# Define structured state for precise control
+class MarketState(BaseModel):
+    sentiment: str = "neutral"
+    confidence: float = 0.0
+    recommendations: list = []
+
+class AdvancedAnalysisFlow(Flow[MarketState]):
+    @start()
+    def fetch_market_data(self):
+        # Demonstrate low-level control with structured state
+        self.state.sentiment = "analyzing"
+        return {"sector": "tech", "timeframe": "1W"}  # These parameters match the task description template
+
+    @listen(fetch_market_data)
+    def analyze_with_crew(self, market_data):
+        # Show crew agency through specialized roles
+        analyst = Agent(
+            role="Senior Market Analyst",
+            goal="Conduct deep market analysis with expert insight",
+            backstory="You're a veteran analyst known for identifying subtle market patterns"
+        )
+        researcher = Agent(
+            role="Data Researcher",
+            goal="Gather and validate supporting market data",
+            backstory="You excel at finding and correlating multiple data sources"
+        )
+        
+        analysis_task = Task(
+            description="Analyze {sector} sector data for the past {timeframe}",
+            expected_output="Detailed market analysis with confidence score",
+            agent=analyst
+        )
+        research_task = Task(
+            description="Find supporting data to validate the analysis",
+            expected_output="Corroborating evidence and potential contradictions",
+            agent=researcher
+        )
+        
+        # Demonstrate crew autonomy
+        analysis_crew = Crew(
+            agents=[analyst, researcher],
+            tasks=[analysis_task, research_task],
+            process=Process.sequential,
+            verbose=True
+        )
+        return analysis_crew.kickoff(inputs=market_data)  # Pass market_data as named inputs
+
+    @router(analyze_with_crew)
+    def determine_next_steps(self):
+        # Show flow control with conditional routing
+        if self.state.confidence > 0.8:
+            return "high_confidence"
+        elif self.state.confidence > 0.5:
+            return "medium_confidence"
+        return "low_confidence"
+
+    @listen("high_confidence")
+    def execute_strategy(self):
+        # Demonstrate complex decision making
+        strategy_crew = Crew(
+            agents=[
+                Agent(role="Strategy Expert",
+                      goal="Develop optimal market strategy")
+            ],
+            tasks=[
+                Task(description="Create detailed strategy based on analysis",
+                     expected_output="Step-by-step action plan")
+            ]
+        )
+        return strategy_crew.kickoff()
+
+    @listen("medium_confidence", "low_confidence")
+    def request_additional_analysis(self):
+        self.state.recommendations.append("Gather more data")
+        return "Additional analysis required"
+```
+
+This example demonstrates how to:
+1. Use Python code for basic data operations
+2. Create and execute Crews as steps in your workflow
+3. Use Flow decorators to manage the sequence of operations
+4. Implement conditional branching based on Crew results
+
 ## Connecting Your Crew to a Model
 
 CrewAI supports using various LLMs through a variety of connection options. By default your agents will use the OpenAI API when querying the model. However, there are several other ways to allow your agents to connect to models. For example, you can configure your agents to use a local model via the Ollama tool.
@@ -313,9 +457,13 @@ Please refer to the [Connect CrewAI to LLMs](https://docs.crewai.com/how-to/LLM-
 
 ## How CrewAI Compares
 
-**CrewAI's Advantage**: CrewAI is built with production in mind. It offers the flexibility of Autogen's conversational agents and the structured process approach of ChatDev, but without the rigidity. CrewAI's processes are designed to be dynamic and adaptable, fitting seamlessly into both development and production workflows.
+**CrewAI's Advantage**: CrewAI combines autonomous agent intelligence with precise workflow control through its unique Crews and Flows architecture. The framework excels at both high-level orchestration and low-level customization, enabling complex, production-grade systems with granular control.
 
-- **Autogen**: While Autogen does good in creating conversational agents capable of working together, it lacks an inherent concept of process. In Autogen, orchestrating agents' interactions requires additional programming, which can become complex and cumbersome as the scale of tasks grows.
+- **LangGraph**: While LangGraph provides a foundation for building agent workflows, its approach requires significant boilerplate code and complex state management patterns. The framework's tight coupling with LangChain can limit flexibility when implementing custom agent behaviors or integrating with external systems.
+
+*P.S. CrewAI demonstrates significant performance advantages over LangGraph, executing 5.76x faster in certain cases like this QA task example ([see comparison](https://github.com/crewAIInc/crewAI-examples/tree/main/Notebooks/CrewAI%20Flows%20%26%20Langgraph/QA%20Agent)) while achieving higher evaluation scores with faster completion times in certain coding tasks, like in this example ([detailed analysis](https://github.com/crewAIInc/crewAI-examples/blob/main/Notebooks/CrewAI%20Flows%20%26%20Langgraph/Coding%20Assistant/coding_assistant_eval.ipynb)).*
+
+- **Autogen**: While Autogen excels at creating conversational agents capable of working together, it lacks an inherent concept of process. In Autogen, orchestrating agents' interactions requires additional programming, which can become complex and cumbersome as the scale of tasks grows.
 
 - **ChatDev**: ChatDev introduced the idea of processes into the realm of AI agents, but its implementation is quite rigid. Customizations in ChatDev are limited and not geared towards production environments, which can hinder scalability and flexibility in real-world applications.
 
@@ -440,5 +588,8 @@ A: CrewAI uses anonymous telemetry to collect usage data for improvement purpose
 ### Q: Where can I find examples of CrewAI in action?
 A: You can find various real-life examples in the [CrewAI-examples repository](https://github.com/crewAIInc/crewAI-examples), including trip planners, stock analysis tools, and more.
 
+### Q: What is the difference between Crews and Flows?
+A: Crews and Flows serve different but complementary purposes in CrewAI. Crews are teams of AI agents working together to accomplish specific tasks through role-based collaboration, delivering accurate and predictable results. Flows, on the other hand, are event-driven workflows that can orchestrate both Crews and regular Python code, allowing you to build complex automation pipelines with secure state management and conditional execution paths.
+
 ### Q: How can I contribute to CrewAI?
 A: Contributions are welcome! You can fork the repository, create a new branch for your feature, add your improvement, and send a pull request. Check the Contribution section in the README for more details.

docs/concepts/agents.mdx

@@ -101,6 +101,8 @@ from crewai_tools import SerperDevTool
 class LatestAiDevelopmentCrew():
   """LatestAiDevelopment crew"""
 
+  agents_config = "config/agents.yaml"
+
   @agent
   def researcher(self) -> Agent:
     return Agent(

docs/concepts/cli.mdx

@@ -161,6 +161,7 @@ The CLI will initially prompt for API keys for the following services:
 * Groq
 * Anthropic
 * Google Gemini
+* SambaNova
 
 When you select a provider, the CLI will prompt you to enter your API key.
 

docs/concepts/crews.mdx

@@ -32,7 +32,6 @@ A crew in crewAI represents a collaborative group of agents working together to
 | **Share Crew** _(optional)_           | `share_crew`           | Whether you want to share the complete crew information and execution with the crewAI team to make the library better, and allow us to train models.                                                                                                      |
 | **Output Log File** _(optional)_      | `output_log_file`      | Whether you want to have a file with the complete crew output and execution. You can set it using True and it will default to the folder you are currently in and it will be called logs.txt or passing a string with the full path and name of the file. |
 | **Manager Agent** _(optional)_        | `manager_agent`        | `manager` sets a custom agent that will be used as a manager.                                                                                                                                                                                             |
-| **Manager Callbacks** _(optional)_    | `manager_callbacks`    | `manager_callbacks` takes a list of callback handlers to be executed by the manager agent when a hierarchical process is used.                                                                                                                            |
 | **Prompt File** _(optional)_          | `prompt_file`          | Path to the prompt JSON file to be used for the crew.                                                                                                                                                                                                     |
 | **Planning** *(optional)*             | `planning`             | Adds planning ability to the Crew. When activated before each Crew iteration, all Crew data is sent to an AgentPlanner that will plan the tasks and this plan will be added to each task description.                                                     |
 | **Planning LLM** *(optional)*         | `planning_llm`         | The language model used by the AgentPlanner in a planning process.                                                                                                                                                                                        |

docs/concepts/flows.mdx

@@ -35,6 +35,8 @@ class ExampleFlow(Flow):
     @start()
     def generate_city(self):
         print("Starting flow")
+        # Each flow state automatically gets a unique ID
+        print(f"Flow State ID: {self.state['id']}")
 
         response = completion(
             model=self.model,
@@ -47,6 +49,8 @@ class ExampleFlow(Flow):
         )
 
         random_city = response["choices"][0]["message"]["content"]
+        # Store the city in our state
+        self.state["city"] = random_city
         print(f"Random City: {random_city}")
 
         return random_city
@@ -64,6 +68,8 @@ class ExampleFlow(Flow):
         )
 
         fun_fact = response["choices"][0]["message"]["content"]
+        # Store the fun fact in our state
+        self.state["fun_fact"] = fun_fact
         return fun_fact
 
 
@@ -76,7 +82,15 @@ print(f"Generated fun fact: {result}")
 
 In the above example, we have created a simple Flow that generates a random city using OpenAI and then generates a fun fact about that city. The Flow consists of two tasks: `generate_city` and `generate_fun_fact`. The `generate_city` task is the starting point of the Flow, and the `generate_fun_fact` task listens for the output of the `generate_city` task.
 
-When you run the Flow, it will generate a random city and then generate a fun fact about that city. The output will be printed to the console.
+Each Flow instance automatically receives a unique identifier (UUID) in its state, which helps track and manage flow executions. The state can also store additional data (like the generated city and fun fact) that persists throughout the flow's execution.
+
+When you run the Flow, it will:
+1. Generate a unique ID for the flow state
+2. Generate a random city and store it in the state
+3. Generate a fun fact about that city and store it in the state
+4. Print the results to the console
+
+The state's unique ID and stored data can be useful for tracking flow executions and maintaining context between tasks.
 
 **Note:** Ensure you have set up your `.env` file to store your `OPENAI_API_KEY`. This key is necessary for authenticating requests to the OpenAI API.
 
@@ -138,7 +152,7 @@ print("---- Final Output ----")
 print(final_output)
 ````
 
-``` text Output
+```text Output
 ---- Final Output ----
 Second method received: Output from first_method
 ````
@@ -207,14 +221,17 @@ allowing developers to choose the approach that best fits their application's ne
 
 In unstructured state management, all state is stored in the `state` attribute of the `Flow` class.
 This approach offers flexibility, enabling developers to add or modify state attributes on the fly without defining a strict schema.
+Even with unstructured states, CrewAI Flows automatically generates and maintains a unique identifier (UUID) for each state instance.
 
 ```python Code
 from crewai.flow.flow import Flow, listen, start
 
-class UntructuredExampleFlow(Flow):
+class UnstructuredExampleFlow(Flow):
 
     @start()
     def first_method(self):
+        # The state automatically includes an 'id' field
+        print(f"State ID: {self.state['id']}")
         self.state.message = "Hello from structured flow"
         self.state.counter = 0
 
@@ -231,10 +248,12 @@ class UntructuredExampleFlow(Flow):
         print(f"State after third_method: {self.state}")
 
 
-flow = UntructuredExampleFlow()
+flow = UnstructuredExampleFlow()
 flow.kickoff()
 ```
 
+**Note:** The `id` field is automatically generated and preserved throughout the flow's execution. You don't need to manage or set it manually, and it will be maintained even when updating the state with new data.
+
 **Key Points:**
 
 - **Flexibility:** You can dynamically add attributes to `self.state` without predefined constraints.
@@ -245,12 +264,15 @@ flow.kickoff()
 Structured state management leverages predefined schemas to ensure consistency and type safety across the workflow.
 By using models like Pydantic's `BaseModel`, developers can define the exact shape of the state, enabling better validation and auto-completion in development environments.
 
+Each state in CrewAI Flows automatically receives a unique identifier (UUID) to help track and manage state instances. This ID is automatically generated and managed by the Flow system.
+
 ```python Code
 from crewai.flow.flow import Flow, listen, start
 from pydantic import BaseModel
 
 
 class ExampleState(BaseModel):
+    # Note: 'id' field is automatically added to all states
     counter: int = 0
     message: str = ""
 
@@ -259,6 +281,8 @@ class StructuredExampleFlow(Flow[ExampleState]):
 
     @start()
     def first_method(self):
+        # Access the auto-generated ID if needed
+        print(f"State ID: {self.state.id}")
         self.state.message = "Hello from structured flow"
 
     @listen(first_method)
@@ -628,4 +652,4 @@ Also, check out our YouTube video on how to use flows in CrewAI below!
   allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share"
   referrerpolicy="strict-origin-when-cross-origin"
   allowfullscreen
-></iframe>
+></iframe>

docs/concepts/knowledge.mdx

@@ -4,12 +4,10 @@ description: What is knowledge in CrewAI and how to use it.
 icon: book
 ---
 
-# Using Knowledge in CrewAI
-
 ## What is Knowledge?
 
-Knowledge in CrewAI is a powerful system that allows AI agents to access and utilize external information sources during their tasks. 
-Think of it as giving your agents a reference library they can consult while working. 
+Knowledge in CrewAI is a powerful system that allows AI agents to access and utilize external information sources during their tasks.
+Think of it as giving your agents a reference library they can consult while working.
 
 <Info>
   Key benefits of using Knowledge:
@@ -36,7 +34,20 @@ CrewAI supports various types of knowledge sources out of the box:
   </Card>
 </CardGroup>
 
-## Quick Start
+## Supported Knowledge Parameters
+
+| Parameter                    | Type                                | Required | Description                                                                                                                                           |
+| :--------------------------- | :---------------------------------- | :------- | :---------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `sources`                  | **List[BaseKnowledgeSource]**        | Yes      | List of knowledge sources that provide content to be stored and queried. Can include PDF, CSV, Excel, JSON, text files, or string content.           |
+| `collection_name`          | **str**                              | No       | Name of the collection where the knowledge will be stored. Used to identify different sets of knowledge. Defaults to "knowledge" if not provided.     |
+| `storage`                  | **Optional[KnowledgeStorage]**       | No       | Custom storage configuration for managing how the knowledge is stored and retrieved. If not provided, a default storage will be created.              |
+
+## Quickstart Example
+
+<Tip>
+For file-Based Knowledge Sources, make sure to place your files in a `knowledge` directory at the root of your project. 
+Also, use relative paths from the `knowledge` directory when creating the source.
+</Tip>
 
 Here's an example using string-based knowledge:
 
@@ -47,7 +58,7 @@ from crewai.knowledge.source.string_knowledge_source import StringKnowledgeSourc
 # Create a knowledge source
 content = "Users name is John. He is 30 years old and lives in San Francisco."
 string_source = StringKnowledgeSource(
-    content=content, 
+    content=content,
 )
 
 # Create an LLM with a temperature of 0 to ensure deterministic outputs
@@ -79,39 +90,267 @@ crew = Crew(
 result = crew.kickoff(inputs={"question": "What city does John live in and how old is he?"})
 ```
 
+
+Here's another example with the `CrewDoclingSource`. The CrewDoclingSource is actually quite versatile and can handle multiple file formats including TXT, PDF, DOCX, HTML, and more. 
+
+```python Code
+from crewai import LLM, Agent, Crew, Process, Task
+from crewai.knowledge.source.crew_docling_source import CrewDoclingSource
+
+# Create a knowledge source
+content_source = CrewDoclingSource(
+    file_paths=[
+        "https://lilianweng.github.io/posts/2024-11-28-reward-hacking",
+        "https://lilianweng.github.io/posts/2024-07-07-hallucination",
+    ],
+)
+
+# Create an LLM with a temperature of 0 to ensure deterministic outputs
+llm = LLM(model="gpt-4o-mini", temperature=0)
+
+# Create an agent with the knowledge store
+agent = Agent(
+    role="About papers",
+    goal="You know everything about the papers.",
+    backstory="""You are a master at understanding papers and their content.""",
+    verbose=True,
+    allow_delegation=False,
+    llm=llm,
+)
+task = Task(
+    description="Answer the following questions about the papers: {question}",
+    expected_output="An answer to the question.",
+    agent=agent,
+)
+
+crew = Crew(
+    agents=[agent],
+    tasks=[task],
+    verbose=True,
+    process=Process.sequential,
+    knowledge_sources=[
+        content_source
+    ],  # Enable knowledge by adding the sources here. You can also add more sources to the sources list.
+)
+
+result = crew.kickoff(
+    inputs={
+        "question": "What is the reward hacking paper about? Be sure to provide sources."
+    }
+)
+```
+
+## More Examples
+
+Here are examples of how to use different types of knowledge sources:
+
+### Text File Knowledge Source
+```python
+from crewai.knowledge.source.crew_docling_source import CrewDoclingSource
+
+# Create a text file knowledge source
+text_source = CrewDoclingSource(
+    file_paths=["document.txt", "another.txt"]
+)
+
+# Create crew with text file source on agents or crew level
+agent = Agent(
+    ...
+    knowledge_sources=[text_source]
+)
+
+crew = Crew(
+    ...
+    knowledge_sources=[text_source]
+)
+```
+
+### PDF Knowledge Source
+```python
+from crewai.knowledge.source.pdf_knowledge_source import PDFKnowledgeSource
+
+# Create a PDF knowledge source
+pdf_source = PDFKnowledgeSource(
+    file_paths=["document.pdf", "another.pdf"]
+)
+
+# Create crew with PDF knowledge source on agents or crew level
+agent = Agent(
+    ...
+    knowledge_sources=[pdf_source]
+)
+
+crew = Crew(
+    ...
+    knowledge_sources=[pdf_source]
+)
+```
+
+### CSV Knowledge Source
+```python
+from crewai.knowledge.source.csv_knowledge_source import CSVKnowledgeSource
+
+# Create a CSV knowledge source
+csv_source = CSVKnowledgeSource(
+    file_paths=["data.csv"]
+)
+
+# Create crew with CSV knowledge source or on agent level
+agent = Agent(
+    ...
+    knowledge_sources=[csv_source]
+)
+
+crew = Crew(
+    ...
+    knowledge_sources=[csv_source]
+)
+```
+
+### Excel Knowledge Source
+```python
+from crewai.knowledge.source.excel_knowledge_source import ExcelKnowledgeSource
+
+# Create an Excel knowledge source
+excel_source = ExcelKnowledgeSource(
+    file_paths=["spreadsheet.xlsx"]
+)
+
+# Create crew with Excel knowledge source on agents or crew level
+agent = Agent(
+    ...
+    knowledge_sources=[excel_source]
+)
+
+crew = Crew(
+    ...
+    knowledge_sources=[excel_source]
+)
+```
+
+### JSON Knowledge Source
+```python
+from crewai.knowledge.source.json_knowledge_source import JSONKnowledgeSource
+
+# Create a JSON knowledge source
+json_source = JSONKnowledgeSource(
+    file_paths=["data.json"]
+)
+
+# Create crew with JSON knowledge source on agents or crew level
+agent = Agent(
+    ...
+    knowledge_sources=[json_source]
+)
+
+crew = Crew(
+    ...
+    knowledge_sources=[json_source]
+)
+```
+
 ## Knowledge Configuration
 
 ### Chunking Configuration
 
-Control how content is split for processing by setting the chunk size and overlap.
+Knowledge sources automatically chunk content for better processing. 
+You can configure chunking behavior in your knowledge sources:
 
-```python Code
-knowledge_source = StringKnowledgeSource(
-    content="Long content...",
-    chunk_size=4000,     # Characters per chunk (default)
-    chunk_overlap=200    # Overlap between chunks (default)
+```python
+from crewai.knowledge.source.string_knowledge_source import StringKnowledgeSource
+
+source = StringKnowledgeSource(
+    content="Your content here",
+    chunk_size=4000,      # Maximum size of each chunk (default: 4000)
+    chunk_overlap=200     # Overlap between chunks (default: 200)
 )
 ```
 
-## Embedder Configuration
+The chunking configuration helps in:
+- Breaking down large documents into manageable pieces
+- Maintaining context through chunk overlap
+- Optimizing retrieval accuracy
 
-You can also configure the embedder for the knowledge store. This is useful if you want to use a different embedder for the knowledge store than the one used for the agents.
+### Embeddings Configuration
 
-```python Code
-...
+You can also configure the embedder for the knowledge store. 
+This is useful if you want to use a different embedder for the knowledge store than the one used for the agents.
+The `embedder` parameter supports various embedding model providers that include:
+- `openai`: OpenAI's embedding models
+- `google`: Google's text embedding models
+- `azure`: Azure OpenAI embeddings
+- `ollama`: Local embeddings with Ollama
+- `vertexai`: Google Cloud VertexAI embeddings
+- `cohere`: Cohere's embedding models
+- `bedrock`: AWS Bedrock embeddings
+- `huggingface`: Hugging Face models
+- `watson`: IBM Watson embeddings
+
+Here's an example of how to configure the embedder for the knowledge store using Google's `text-embedding-004` model:
+<CodeGroup>
+```python Example
+from crewai import Agent, Task, Crew, Process, LLM
+from crewai.knowledge.source.string_knowledge_source import StringKnowledgeSource
+import os
+
+# Get the GEMINI API key
+GEMINI_API_KEY = os.environ.get("GEMINI_API_KEY")
+
+# Create a knowledge source
+content = "Users name is John. He is 30 years old and lives in San Francisco."
 string_source = StringKnowledgeSource(
-    content="Users name is John. He is 30 years old and lives in San Francisco.",
+    content=content,
 )
+
+# Create an LLM with a temperature of 0 to ensure deterministic outputs
+gemini_llm = LLM(
+    model="gemini/gemini-1.5-pro-002",
+    api_key=GEMINI_API_KEY,
+    temperature=0,
+)
+
+# Create an agent with the knowledge store
+agent = Agent(
+    role="About User",
+    goal="You know everything about the user.",
+    backstory="""You are a master at understanding people and their preferences.""",
+    verbose=True,
+    allow_delegation=False,
+    llm=gemini_llm,
+)
+
+task = Task(
+    description="Answer the following questions about the user: {question}",
+    expected_output="An answer to the question.",
+    agent=agent,
+)
+
 crew = Crew(
-    ...
+    agents=[agent],
+    tasks=[task],
+    verbose=True,
+    process=Process.sequential,
     knowledge_sources=[string_source],
     embedder={
-        "provider": "openai",
-        "config": {"model": "text-embedding-3-small"},
-    },
+        "provider": "google",
+        "config": {
+            "model": "models/text-embedding-004",
+            "api_key": GEMINI_API_KEY,
+        }
+    }
 )
-```
 
+result = crew.kickoff(inputs={"question": "What city does John live in and how old is he?"})
+```
+```text Output
+# Agent: About User
+## Task: Answer the following questions about the user: What city does John live in and how old is he?
+
+# Agent: About User
+## Final Answer: 
+John is 30 years old and lives in San Francisco.
+```
+</CodeGroup>
 ## Clearing Knowledge
 
 If you need to clear the knowledge stored in CrewAI, you can use the `crewai reset-memories` command with the `--knowledge` option.
@@ -122,6 +361,57 @@ crewai reset-memories --knowledge
 
 This is useful when you've updated your knowledge sources and want to ensure that the agents are using the most recent information.
 
+## Agent-Specific Knowledge
+
+While knowledge can be provided at the crew level using `crew.knowledge_sources`, individual agents can also have their own knowledge sources using the `knowledge_sources` parameter:
+
+```python Code
+from crewai import Agent, Task, Crew
+from crewai.knowledge.source.string_knowledge_source import StringKnowledgeSource
+
+# Create agent-specific knowledge about a product
+product_specs = StringKnowledgeSource(
+    content="""The XPS 13 laptop features:
+    - 13.4-inch 4K display
+    - Intel Core i7 processor
+    - 16GB RAM
+    - 512GB SSD storage
+    - 12-hour battery life""",
+    metadata={"category": "product_specs"}
+)
+
+# Create a support agent with product knowledge
+support_agent = Agent(
+    role="Technical Support Specialist",
+    goal="Provide accurate product information and support.",
+    backstory="You are an expert on our laptop products and specifications.",
+    knowledge_sources=[product_specs]  # Agent-specific knowledge
+)
+
+# Create a task that requires product knowledge
+support_task = Task(
+    description="Answer this customer question: {question}",
+    agent=support_agent
+)
+
+# Create and run the crew
+crew = Crew(
+    agents=[support_agent],
+    tasks=[support_task]
+)
+
+# Get answer about the laptop's specifications
+result = crew.kickoff(
+    inputs={"question": "What is the storage capacity of the XPS 13?"}
+)
+```
+
+<Info>
+  Benefits of agent-specific knowledge:
+  - Give agents specialized information for their roles
+  - Maintain separation of concerns between agents
+  - Combine with crew-level knowledge for layered information access
+</Info>
 
 ## Custom Knowledge Sources
 
@@ -141,10 +431,10 @@ from pydantic import BaseModel, Field
 
 class SpaceNewsKnowledgeSource(BaseKnowledgeSource):
     """Knowledge source that fetches data from Space News API."""
-    
+
     api_endpoint: str = Field(description="API endpoint URL")
     limit: int = Field(default=10, description="Number of articles to fetch")
-    
+
     def load_content(self) -> Dict[Any, str]:
         """Fetch and format space news articles."""
         try:
@@ -152,15 +442,15 @@ class SpaceNewsKnowledgeSource(BaseKnowledgeSource):
                 f"{self.api_endpoint}?limit={self.limit}"
             )
             response.raise_for_status()
-            
+
             data = response.json()
             articles = data.get('results', [])
-            
+
             formatted_data = self._format_articles(articles)
             return {self.api_endpoint: formatted_data}
         except Exception as e:
             raise ValueError(f"Failed to fetch space news: {str(e)}")
-    
+
     def _format_articles(self, articles: list) -> str:
         """Format articles into readable text."""
         formatted = "Space News Articles:\n\n"
@@ -180,7 +470,7 @@ class SpaceNewsKnowledgeSource(BaseKnowledgeSource):
         for _, text in content.items():
             chunks = self._chunk_text(text)
             self.chunks.extend(chunks)
-        
+
         self._save_documents()
 
 # Create knowledge source
@@ -193,7 +483,7 @@ recent_news = SpaceNewsKnowledgeSource(
 space_analyst = Agent(
     role="Space News Analyst",
     goal="Answer questions about space news accurately and comprehensively",
-    backstory="""You are a space industry analyst with expertise in space exploration, 
+    backstory="""You are a space industry analyst with expertise in space exploration,
     satellite technology, and space industry trends. You excel at answering questions
     about space news and providing detailed, accurate information.""",
     knowledge_sources=[recent_news],
@@ -220,13 +510,14 @@ result = crew.kickoff(
     inputs={"user_question": "What are the latest developments in space exploration?"}
 )
 ```
+
 ```output Output
 # Agent: Space News Analyst
 ## Task: Answer this question about space news: What are the latest developments in space exploration?
 
 
 # Agent: Space News Analyst
-## Final Answer: 
+## Final Answer:
 The latest developments in space exploration, based on recent space news articles, include the following:
 
 1. SpaceX has received the final regulatory approvals to proceed with the second integrated Starship/Super Heavy launch, scheduled for as soon as the morning of Nov. 17, 2023. This is a significant step in SpaceX's ambitious plans for space exploration and colonization. [Source: SpaceNews](https://spacenews.com/starship-cleared-for-nov-17-launch/)
@@ -242,11 +533,13 @@ The latest developments in space exploration, based on recent space news article
 6. The National Natural Science Foundation of China has outlined a five-year project for researchers to study the assembly of ultra-large spacecraft. This could lead to significant advancements in spacecraft technology and space exploration capabilities. [Source: SpaceNews](https://spacenews.com/china-researching-challenges-of-kilometer-scale-ultra-large-spacecraft/)
 
 7. The Center for AEroSpace Autonomy Research (CAESAR) at Stanford University is focusing on spacecraft autonomy. The center held a kickoff event on May 22, 2024, to highlight the industry, academia, and government collaboration it seeks to foster. This could lead to significant advancements in autonomous spacecraft technology. [Source: SpaceNews](https://spacenews.com/stanford-center-focuses-on-spacecraft-autonomy/)
-``` 
+```
+
 </CodeGroup>
 #### Key Components Explained
 
 1. **Custom Knowledge Source (`SpaceNewsKnowledgeSource`)**:
+
    - Extends `BaseKnowledgeSource` for integration with CrewAI
    - Configurable API endpoint and article limit
    - Implements three key methods:
@@ -255,10 +548,12 @@ The latest developments in space exploration, based on recent space news article
      - `add()`: Processes and stores the content
 
 2. **Agent Configuration**:
+
    - Specialized role as a Space News Analyst
    - Uses the knowledge source to access space news
 
 3. **Task Setup**:
+
    - Takes a user question as input through `{user_question}`
    - Designed to provide detailed answers based on the knowledge source
 
@@ -267,6 +562,7 @@ The latest developments in space exploration, based on recent space news article
    - Handles input/output through the kickoff method
 
 This example demonstrates how to:
+
 - Create a custom knowledge source that fetches real-time data
 - Process and format external data for AI consumption
 - Use the knowledge source to answer specific user questions
@@ -274,13 +570,15 @@ This example demonstrates how to:
 
 #### About the Spaceflight News API
 
-The example uses the [Spaceflight News API](https://api.spaceflightnewsapi.net/v4/documentation), which:
+The example uses the [Spaceflight News API](https://api.spaceflightnewsapi.net/v4/docs/), which:
+
 - Provides free access to space-related news articles
 - Requires no authentication
 - Returns structured data about space news
 - Supports pagination and filtering
 
 You can customize the API query by modifying the endpoint URL:
+
 ```python
 # Fetch more articles
 recent_news = SpaceNewsKnowledgeSource(
@@ -303,9 +601,9 @@ recent_news = SpaceNewsKnowledgeSource(
     - Consider content overlap for context preservation
     - Organize related information into separate knowledge sources
   </Accordion>
-  
+
   <Accordion title="Performance Tips">
-    - Adjust chunk sizes based on content complexity
+    - Adjust chunk sizes based on content complexity 
     - Configure appropriate embedding models
     - Consider using local embedding providers for faster processing
   </Accordion>

docs/concepts/llms.mdx

@@ -29,7 +29,7 @@ Large Language Models (LLMs) are the core intelligence behind CrewAI agents. The
 
 ## Available Models and Their Capabilities
 
-Here's a detailed breakdown of supported models and their capabilities:
+Here's a detailed breakdown of supported models and their capabilities, you can compare performance at [lmarena.ai](https://lmarena.ai/?leaderboard) and [artificialanalysis.ai](https://artificialanalysis.ai/):
 
 <Tabs>
   <Tab title="OpenAI">
@@ -43,24 +43,128 @@ Here's a detailed breakdown of supported models and their capabilities:
       1 token ≈ 4 characters in English. For example, 8,192 tokens ≈ 32,768 characters or about 6,000 words.
     </Note>
   </Tab>
+  <Tab title="Nvidia NIM">
+    | Model | Context Window | Best For |
+    |-------|---------------|-----------|
+    | nvidia/mistral-nemo-minitron-8b-8k-instruct | 8,192 tokens | State-of-the-art small language model delivering superior accuracy for chatbot, virtual assistants, and content generation. |
+    | nvidia/nemotron-4-mini-hindi-4b-instruct| 4,096 tokens | A bilingual Hindi-English SLM for on-device inference, tailored specifically for Hindi Language. |
+    | "nvidia/llama-3.1-nemotron-70b-instruct | 128k tokens | Llama-3.1-Nemotron-70B-Instruct is a large language model customized by NVIDIA in order to improve the helpfulness of LLM generated responses. |
+    | nvidia/llama3-chatqa-1.5-8b | 128k tokens | Advanced LLM to generate high-quality, context-aware responses for chatbots and search engines. |
+    | nvidia/llama3-chatqa-1.5-70b | 128k tokens | Advanced LLM to generate high-quality, context-aware responses for chatbots and search engines. |
+    | nvidia/vila | 128k tokens | Multi-modal vision-language model that understands text/img/video and creates informative responses |
+    | nvidia/neva-22| 4,096 tokens | Multi-modal vision-language model that understands text/images and generates informative responses |
+    | nvidia/nemotron-mini-4b-instruct | 8,192 tokens | General-purpose tasks |
+    | nvidia/usdcode-llama3-70b-instruct | 128k tokens | State-of-the-art LLM that answers OpenUSD knowledge queries and generates USD-Python code. |
+    | nvidia/nemotron-4-340b-instruct | 4,096 tokens | Creates diverse synthetic data that mimics the characteristics of real-world data. |
+    | meta/codellama-70b | 100k tokens | LLM capable of generating code from natural language and vice versa. |
+    | meta/llama2-70b | 4,096 tokens | Cutting-edge large language AI model capable of generating text and code in response to prompts. |
+    | meta/llama3-8b-instruct | 8,192 tokens | Advanced state-of-the-art LLM with language understanding, superior reasoning, and text generation. |
+    | meta/llama3-70b-instruct | 8,192 tokens | Powers complex conversations with superior contextual understanding, reasoning and text generation. |
+    | meta/llama-3.1-8b-instruct | 128k tokens | Advanced state-of-the-art model with language understanding, superior reasoning, and text generation. |
+    | meta/llama-3.1-70b-instruct | 128k tokens | Powers complex conversations with superior contextual understanding, reasoning and text generation. |
+    | meta/llama-3.1-405b-instruct | 128k tokens | Advanced LLM for synthetic data generation, distillation, and inference for chatbots, coding, and domain-specific tasks. |
+    | meta/llama-3.2-1b-instruct | 128k tokens | Advanced state-of-the-art small language model with language understanding, superior reasoning, and text generation. |
+    | meta/llama-3.2-3b-instruct | 128k tokens | Advanced state-of-the-art small language model with language understanding, superior reasoning, and text generation. |
+    | meta/llama-3.2-11b-vision-instruct | 128k tokens | Advanced state-of-the-art small language model with language understanding, superior reasoning, and text generation. |
+    | meta/llama-3.2-90b-vision-instruct | 128k tokens | Advanced state-of-the-art small language model with language understanding, superior reasoning, and text generation. |
+    | meta/llama-3.1-70b-instruct | 128k tokens | Powers complex conversations with superior contextual understanding, reasoning and text generation. |
+    | google/gemma-7b | 8,192 tokens | Cutting-edge text generation model text understanding, transformation, and code generation. |
+    | google/gemma-2b | 8,192 tokens | Cutting-edge text generation model text understanding, transformation, and code generation. |
+    | google/codegemma-7b | 8,192 tokens | Cutting-edge model built on Google's Gemma-7B specialized for code generation and code completion. |
+    | google/codegemma-1.1-7b | 8,192 tokens | Advanced programming model for code generation, completion, reasoning, and instruction following. |
+    | google/recurrentgemma-2b | 8,192 tokens | Novel recurrent architecture based language model for faster inference when generating long sequences. |
+    | google/gemma-2-9b-it | 8,192 tokens | Cutting-edge text generation model text understanding, transformation, and code generation. |
+    | google/gemma-2-27b-it | 8,192 tokens | Cutting-edge text generation model text understanding, transformation, and code generation. |
+    | google/gemma-2-2b-it | 8,192 tokens | Cutting-edge text generation model text understanding, transformation, and code generation. |
+    | google/deplot | 512 tokens | One-shot visual language understanding model that translates images of plots into tables. |
+    | google/paligemma | 8,192 tokens | Vision language model adept at comprehending text and visual inputs to produce informative responses. |
+    | mistralai/mistral-7b-instruct-v0.2 | 32k tokens | This LLM follows instructions, completes requests, and generates creative text. |
+    | mistralai/mixtral-8x7b-instruct-v0.1 | 8,192 tokens | An MOE LLM that follows instructions, completes requests, and generates creative text. |
+    | mistralai/mistral-large | 4,096 tokens | Creates diverse synthetic data that mimics the characteristics of real-world data. |
+    | mistralai/mixtral-8x22b-instruct-v0.1 | 8,192 tokens | Creates diverse synthetic data that mimics the characteristics of real-world data. |
+    | mistralai/mistral-7b-instruct-v0.3 | 32k tokens | This LLM follows instructions, completes requests, and generates creative text. |
+    | nv-mistralai/mistral-nemo-12b-instruct | 128k tokens | Most advanced language model for reasoning, code, multilingual tasks; runs on a single GPU. |
+    | mistralai/mamba-codestral-7b-v0.1 | 256k tokens | Model for writing and interacting with code across a wide range of programming languages and tasks. |
+    | microsoft/phi-3-mini-128k-instruct | 128K tokens | Lightweight, state-of-the-art open LLM with strong math and logical reasoning skills. |
+    | microsoft/phi-3-mini-4k-instruct | 4,096 tokens | Lightweight, state-of-the-art open LLM with strong math and logical reasoning skills. |
+    | microsoft/phi-3-small-8k-instruct | 8,192 tokens | Lightweight, state-of-the-art open LLM with strong math and logical reasoning skills. |
+    | microsoft/phi-3-small-128k-instruct | 128K tokens | Lightweight, state-of-the-art open LLM with strong math and logical reasoning skills. |
+    | microsoft/phi-3-medium-4k-instruct | 4,096 tokens | Lightweight, state-of-the-art open LLM with strong math and logical reasoning skills. |
+    | microsoft/phi-3-medium-128k-instruct | 128K tokens | Lightweight, state-of-the-art open LLM with strong math and logical reasoning skills. |
+    | microsoft/phi-3.5-mini-instruct | 128K tokens | Lightweight multilingual LLM powering AI applications in latency bound, memory/compute constrained environments |
+    | microsoft/phi-3.5-moe-instruct | 128K tokens | Advanced LLM based on Mixture of Experts architecure to deliver compute efficient content generation |
+    | microsoft/kosmos-2 | 1,024 tokens | Groundbreaking multimodal model designed to understand and reason about visual elements in images. |
+    | microsoft/phi-3-vision-128k-instruct | 128k tokens | Cutting-edge open multimodal model exceling in high-quality reasoning from images. |
+    | microsoft/phi-3.5-vision-instruct | 128k tokens | Cutting-edge open multimodal model exceling in high-quality reasoning from images. |
+    | databricks/dbrx-instruct | 12k tokens | A general-purpose LLM with state-of-the-art performance in language understanding, coding, and RAG. |
+    | snowflake/arctic | 1,024 tokens | Delivers high efficiency inference for enterprise applications focused on SQL generation and coding. |
+    | aisingapore/sea-lion-7b-instruct | 4,096 tokens | LLM to represent and serve the linguistic and cultural diversity of Southeast Asia |
+    | ibm/granite-8b-code-instruct | 4,096 tokens | Software programming LLM for code generation, completion, explanation, and multi-turn conversion. |
+    | ibm/granite-34b-code-instruct | 8,192 tokens | Software programming LLM for code generation, completion, explanation, and multi-turn conversion. |
+    | ibm/granite-3.0-8b-instruct | 4,096 tokens | Advanced Small Language Model supporting RAG, summarization, classification, code, and agentic AI |
+    | ibm/granite-3.0-3b-a800m-instruct | 4,096 tokens | Highly efficient Mixture of Experts model for RAG, summarization, entity extraction, and classification |
+    | mediatek/breeze-7b-instruct | 4,096 tokens | Creates diverse synthetic data that mimics the characteristics of real-world data. |
+    | upstage/solar-10.7b-instruct | 4,096 tokens | Excels in NLP tasks, particularly in instruction-following, reasoning, and mathematics. |
+    | writer/palmyra-med-70b-32k | 32k tokens | Leading LLM for accurate, contextually relevant responses in the medical domain. |
+    | writer/palmyra-med-70b | 32k tokens | Leading LLM for accurate, contextually relevant responses in the medical domain. |
+    | writer/palmyra-fin-70b-32k | 32k tokens | Specialized LLM for financial analysis, reporting, and data processing |
+    | 01-ai/yi-large | 32k tokens | Powerful model trained on English and Chinese for diverse tasks including chatbot and creative writing. |
+    | deepseek-ai/deepseek-coder-6.7b-instruct | 2k tokens | Powerful coding model offering advanced capabilities in code generation, completion, and infilling |
+    | rakuten/rakutenai-7b-instruct | 1,024 tokens | Advanced state-of-the-art LLM with language understanding, superior reasoning, and text generation. |
+    | rakuten/rakutenai-7b-chat | 1,024 tokens | Advanced state-of-the-art LLM with language understanding, superior reasoning, and text generation. |
+    | baichuan-inc/baichuan2-13b-chat | 4,096 tokens | Support Chinese and English chat, coding, math, instruction following, solving quizzes |
+
+    <Note>
+      NVIDIA's NIM support for models is expanding continuously! For the most up-to-date list of available models, please visit build.nvidia.com.
+    </Note>
+  </Tab>
+  <Tab title="Gemini">
+    | Model | Context Window | Best For |
+    |-------|---------------|-----------|
+    | gemini-2.0-flash-exp | 1M tokens | Higher quality at faster speed, multimodal model, good for most tasks |
+    | gemini-1.5-flash | 1M tokens | Balanced multimodal model, good for most tasks |
+    | gemini-1.5-flash-8B | 1M tokens | Fastest, most cost-efficient, good for high-frequency tasks |
+    | gemini-1.5-pro | 2M tokens | Best performing, wide variety of reasoning tasks including logical reasoning, coding, and creative collaboration |
+
+    <Tip>
+      Google's Gemini models are all multimodal, supporting audio, images, video and text, supporting context caching, json schema, function calling, etc.
+
+      These models are available via API_KEY from 
+      [The Gemini API](https://ai.google.dev/gemini-api/docs) and also from 
+      [Google Cloud Vertex](https://cloud.google.com/vertex-ai/generative-ai/docs/migrate/migrate-google-ai) as part of the
+      [Model Garden](https://cloud.google.com/vertex-ai/generative-ai/docs/model-garden/explore-models).
+    </Tip>
+  </Tab>
   <Tab title="Groq">
     | Model | Context Window | Best For |
     |-------|---------------|-----------|
     | Llama 3.1 70B/8B | 131,072 tokens | High-performance, large context tasks |
     | Llama 3.2 Series | 8,192 tokens | General-purpose tasks |
     | Mixtral 8x7B | 32,768 tokens | Balanced performance and context |
-    | Gemma Series | 8,192 tokens | Efficient, smaller-scale tasks |
 
     <Tip>
       Groq is known for its fast inference speeds, making it suitable for real-time applications.
     </Tip>
   </Tab>
+  <Tab title="SambaNova">
+    | Model | Context Window | Best For |
+    |-------|---------------|-----------|
+    | Llama 3.1 70B/8B | Up to 131,072 tokens | High-performance, large context tasks |
+    | Llama 3.1 405B | 8,192 tokens | High-performance and output quality |
+    | Llama 3.2 Series | 8,192 tokens | General-purpose tasks, multimodal |
+    | Llama 3.3 70B | Up to 131,072 tokens | High-performance and output quality|
+    | Qwen2 familly | 8,192 tokens | High-performance and output quality |
+
+    <Tip>
+      [SambaNova](https://cloud.sambanova.ai/) has several models with fast inference speed at full precision.
+    </Tip>
+  </Tab>
   <Tab title="Others">
     | Provider | Context Window | Key Features |
     |----------|---------------|--------------|
     | Deepseek Chat | 128,000 tokens | Specialized in technical discussions |
     | Claude 3 | Up to 200K tokens | Strong reasoning, code understanding |
-    | Gemini | Varies by model | Multimodal capabilities |
+    | Gemma Series | 8,192 tokens | Efficient, smaller-scale tasks |
 
     <Info>
       Provider selection should consider factors like:
@@ -128,10 +232,10 @@ There are three ways to configure LLMs in CrewAI. Choose the method that best fi
         # llm: anthropic/claude-2.1
         # llm: anthropic/claude-2.0
 
-        # Google Models - Good for general tasks
-        # llm: gemini/gemini-pro
+        # Google Models - Strong reasoning, large cachable context window, multimodal
         # llm: gemini/gemini-1.5-pro-latest
-        # llm: gemini/gemini-1.0-pro-latest
+        # llm: gemini/gemini-1.5-flash-latest
+        # llm: gemini/gemini-1.5-flash-8b-latest
 
         # AWS Bedrock Models - Enterprise-grade
         # llm: bedrock/anthropic.claude-3-sonnet-20240229-v1:0
@@ -350,13 +454,18 @@ Learn how to get the most out of your LLM configuration:
 
   <Accordion title="Google">
     ```python Code
+    # Option 1. Gemini accessed with an API key.
+    # https://ai.google.dev/gemini-api/docs/api-key
     GEMINI_API_KEY=<your-api-key>
+
+    # Option 2. Vertex AI IAM credentials for Gemini, Anthropic, and anything in the Model Garden.
+    # https://cloud.google.com/vertex-ai/generative-ai/docs/overview
     ```
 
     Example usage:
     ```python Code
     llm = LLM(
-        model="gemini/gemini-pro",
+        model="gemini/gemini-1.5-pro-latest",
         temperature=0.7
     )
     ```
@@ -412,6 +521,20 @@ Learn how to get the most out of your LLM configuration:
     ```
   </Accordion>
 
+  <Accordion title="Nvidia NIM">
+    ```python Code
+    NVIDIA_API_KEY=<your-api-key>
+    ```
+
+    Example usage:
+    ```python Code
+    llm = LLM(
+        model="nvidia_nim/meta/llama3-70b-instruct",
+        temperature=0.7
+    )
+    ```
+  </Accordion>
+
   <Accordion title="Groq">
     ```python Code
     GROQ_API_KEY=<your-api-key>
@@ -502,20 +625,6 @@ Learn how to get the most out of your LLM configuration:
     ```
   </Accordion>
 
-  <Accordion title="Nvidia NIM">
-    ```python Code
-    NVIDIA_API_KEY=<your-api-key>
-    ```
-
-    Example usage:
-    ```python Code
-    llm = LLM(
-        model="nvidia_nim/meta/llama3-70b-instruct",
-        temperature=0.7
-    )
-    ```
-  </Accordion>
-
   <Accordion title="SambaNova">
     ```python Code
     SAMBANOVA_API_KEY=<your-api-key>

docs/concepts/memory.mdx

@@ -134,6 +134,23 @@ crew = Crew(
 )
 ```
 
+## Memory Configuration Options
+If you want to access a specific organization and project, you can set the `org_id` and `project_id` parameters in the memory configuration.
+
+```python Code
+from crewai import Crew
+
+crew = Crew(
+    agents=[...],
+    tasks=[...],
+    verbose=True,
+    memory=True,
+    memory_config={
+        "provider": "mem0",
+        "config": {"user_id": "john", "org_id": "my_org_id", "project_id": "my_project_id"},
+    },
+)
+```
 
 ## Additional Embedding Providers
 

docs/concepts/planning.mdx

@@ -31,7 +31,7 @@ From this point on, your crew will have planning enabled, and the tasks will be
 
 #### Planning LLM
 
-Now you can define the LLM that will be used to plan the tasks. You can use any ChatOpenAI LLM model available.
+Now you can define the LLM that will be used to plan the tasks. 
 
 When running the base case example, you will see something like the output below, which represents the output of the `AgentPlanner` 
 responsible for creating the step-by-step logic to add to the Agents' tasks.
@@ -39,7 +39,6 @@ responsible for creating the step-by-step logic to add to the Agents' tasks.
 <CodeGroup>
 ```python Code
 from crewai import Crew, Agent, Task, Process
-from langchain_openai import ChatOpenAI
 
 # Assemble your crew with planning capabilities and custom LLM
 my_crew = Crew(
@@ -47,7 +46,7 @@ my_crew = Crew(
     tasks=self.tasks,
     process=Process.sequential,
     planning=True,
-    planning_llm=ChatOpenAI(model="gpt-4o")
+    planning_llm="gpt-4o"
 )
 
 # Run the crew

docs/concepts/processes.mdx

@@ -23,9 +23,7 @@ Processes enable individual agents to operate as a cohesive unit, streamlining t
 To assign a process to a crew, specify the process type upon crew creation to set the execution strategy. For a hierarchical process, ensure to define `manager_llm` or `manager_agent` for the manager agent.
 
 ```python
-from crewai import Crew
-from crewai.process import Process
-from langchain_openai import ChatOpenAI
+from crewai import Crew, Process
 
 # Example: Creating a crew with a sequential process
 crew = Crew(
@@ -40,7 +38,7 @@ crew = Crew(
     agents=my_agents,
     tasks=my_tasks,
     process=Process.hierarchical,
-    manager_llm=ChatOpenAI(model="gpt-4")
+    manager_llm="gpt-4o"
     # or
     # manager_agent=my_manager_agent
 )

docs/concepts/tasks.mdx

@@ -6,7 +6,7 @@ icon: list-check
 
 ## Overview of a Task
 
-In the CrewAI framework, a `Task` is a specific assignment completed by an `Agent`. 
+In the CrewAI framework, a `Task` is a specific assignment completed by an `Agent`.
 
 Tasks provide all necessary details for execution, such as a description, the agent responsible, required tools, and more, facilitating a wide range of action complexities.
 
@@ -263,8 +263,148 @@ analysis_task = Task(
 )
 ```
 
+## Task Guardrails
+
+Task guardrails provide a way to validate and transform task outputs before they
+are passed to the next task. This feature helps ensure data quality and provides
+efeedback to agents when their output doesn't meet specific criteria.
+
+### Using Task Guardrails
+
+To add a guardrail to a task, provide a validation function through the `guardrail` parameter:
+
+```python Code
+from typing import Tuple, Union, Dict, Any
+
+def validate_blog_content(result: str) -> Tuple[bool, Union[Dict[str, Any], str]]:
+    """Validate blog content meets requirements."""
+    try:
+        # Check word count
+        word_count = len(result.split())
+        if word_count > 200:
+            return (False, {
+                "error": "Blog content exceeds 200 words",
+                "code": "WORD_COUNT_ERROR",
+                "context": {"word_count": word_count}
+            })
+
+        # Additional validation logic here
+        return (True, result.strip())
+    except Exception as e:
+        return (False, {
+            "error": "Unexpected error during validation",
+            "code": "SYSTEM_ERROR"
+        })
+
+blog_task = Task(
+    description="Write a blog post about AI",
+    expected_output="A blog post under 200 words",
+    agent=blog_agent,
+    guardrail=validate_blog_content  # Add the guardrail function
+)
+```
+
+### Guardrail Function Requirements
+
+1. **Function Signature**:
+   - Must accept exactly one parameter (the task output)
+   - Should return a tuple of `(bool, Any)`
+   - Type hints are recommended but optional
+
+2. **Return Values**:
+   - Success: Return `(True, validated_result)`
+   - Failure: Return `(False, error_details)`
+
+### Error Handling Best Practices
+
+1. **Structured Error Responses**:
+```python Code
+def validate_with_context(result: str) -> Tuple[bool, Union[Dict[str, Any], str]]:
+    try:
+        # Main validation logic
+        validated_data = perform_validation(result)
+        return (True, validated_data)
+    except ValidationError as e:
+        return (False, {
+            "error": str(e),
+            "code": "VALIDATION_ERROR",
+            "context": {"input": result}
+        })
+    except Exception as e:
+        return (False, {
+            "error": "Unexpected error",
+            "code": "SYSTEM_ERROR"
+        })
+```
+
+2. **Error Categories**:
+   - Use specific error codes
+   - Include relevant context
+   - Provide actionable feedback
+
+3. **Validation Chain**:
+```python Code
+from typing import Any, Dict, List, Tuple, Union
+
+def complex_validation(result: str) -> Tuple[bool, Union[str, Dict[str, Any]]]:
+    """Chain multiple validation steps."""
+    # Step 1: Basic validation
+    if not result:
+        return (False, {"error": "Empty result", "code": "EMPTY_INPUT"})
+
+    # Step 2: Content validation
+    try:
+        validated = validate_content(result)
+        if not validated:
+            return (False, {"error": "Invalid content", "code": "CONTENT_ERROR"})
+
+        # Step 3: Format validation
+        formatted = format_output(validated)
+        return (True, formatted)
+    except Exception as e:
+        return (False, {
+            "error": str(e),
+            "code": "VALIDATION_ERROR",
+            "context": {"step": "content_validation"}
+        })
+```
+
+### Handling Guardrail Results
+
+When a guardrail returns `(False, error)`:
+1. The error is sent back to the agent
+2. The agent attempts to fix the issue
+3. The process repeats until:
+   - The guardrail returns `(True, result)`
+   - Maximum retries are reached
+
+Example with retry handling:
+```python Code
+from typing import Optional, Tuple, Union
+
+def validate_json_output(result: str) -> Tuple[bool, Union[Dict[str, Any], str]]:
+    """Validate and parse JSON output."""
+    try:
+        # Try to parse as JSON
+        data = json.loads(result)
+        return (True, data)
+    except json.JSONDecodeError as e:
+        return (False, {
+            "error": "Invalid JSON format",
+            "code": "JSON_ERROR",
+            "context": {"line": e.lineno, "column": e.colno}
+        })
+
+task = Task(
+    description="Generate a JSON report",
+    expected_output="A valid JSON object",
+    agent=analyst,
+    guardrail=validate_json_output,
+    max_retries=3  # Limit retry attempts
+)
+```
+
 ## Getting Structured Consistent Outputs from Tasks
-When you need to ensure that a task outputs a structured and consistent format, you can use the `output_pydantic` or `output_json` properties on a task. These properties allow you to define the expected output structure, making it easier to parse and utilize the results in your application. 
 
 <Note>
 It's also important to note that the output of the final task of a crew becomes the final output of the actual crew itself.
@@ -608,6 +748,114 @@ While creating and executing tasks, certain validation mechanisms are in place t
 
 These validations help in maintaining the consistency and reliability of task executions within the crewAI framework.
 
+## Task Guardrails
+
+Task guardrails provide a powerful way to validate, transform, or filter task outputs before they are passed to the next task. Guardrails are optional functions that execute before the next task starts, allowing you to ensure that task outputs meet specific requirements or formats.
+
+### Basic Usage
+
+```python Code
+from typing import Tuple, Union
+from crewai import Task
+
+def validate_json_output(result: str) -> Tuple[bool, Union[dict, str]]:
+    """Validate that the output is valid JSON."""
+    try:
+        json_data = json.loads(result)
+        return (True, json_data)
+    except json.JSONDecodeError:
+        return (False, "Output must be valid JSON")
+
+task = Task(
+    description="Generate JSON data",
+    expected_output="Valid JSON object",
+    guardrail=validate_json_output
+)
+```
+
+### How Guardrails Work
+
+1. **Optional Attribute**: Guardrails are an optional attribute at the task level, allowing you to add validation only where needed.
+2. **Execution Timing**: The guardrail function is executed before the next task starts, ensuring valid data flow between tasks.
+3. **Return Format**: Guardrails must return a tuple of `(success, data)`:
+   - If `success` is `True`, `data` is the validated/transformed result
+   - If `success` is `False`, `data` is the error message
+4. **Result Routing**:
+   - On success (`True`), the result is automatically passed to the next task
+   - On failure (`False`), the error is sent back to the agent to generate a new answer
+
+### Common Use Cases
+
+#### Data Format Validation
+```python Code
+def validate_email_format(result: str) -> Tuple[bool, Union[str, str]]:
+    """Ensure the output contains a valid email address."""
+    import re
+    email_pattern = r'^[\w\.-]+@[\w\.-]+\.\w+$'
+    if re.match(email_pattern, result.strip()):
+        return (True, result.strip())
+    return (False, "Output must be a valid email address")
+```
+
+#### Content Filtering
+```python Code
+def filter_sensitive_info(result: str) -> Tuple[bool, Union[str, str]]:
+    """Remove or validate sensitive information."""
+    sensitive_patterns = ['SSN:', 'password:', 'secret:']
+    for pattern in sensitive_patterns:
+        if pattern.lower() in result.lower():
+            return (False, f"Output contains sensitive information ({pattern})")
+    return (True, result)
+```
+
+#### Data Transformation
+```python Code
+def normalize_phone_number(result: str) -> Tuple[bool, Union[str, str]]:
+    """Ensure phone numbers are in a consistent format."""
+    import re
+    digits = re.sub(r'\D', '', result)
+    if len(digits) == 10:
+        formatted = f"({digits[:3]}) {digits[3:6]}-{digits[6:]}"
+        return (True, formatted)
+    return (False, "Output must be a 10-digit phone number")
+```
+
+### Advanced Features
+
+#### Chaining Multiple Validations
+```python Code
+def chain_validations(*validators):
+    """Chain multiple validators together."""
+    def combined_validator(result):
+        for validator in validators:
+            success, data = validator(result)
+            if not success:
+                return (False, data)
+            result = data
+        return (True, result)
+    return combined_validator
+
+# Usage
+task = Task(
+    description="Get user contact info",
+    expected_output="Email and phone",
+    guardrail=chain_validations(
+        validate_email_format,
+        filter_sensitive_info
+    )
+)
+```
+
+#### Custom Retry Logic
+```python Code
+task = Task(
+    description="Generate data",
+    expected_output="Valid data",
+    guardrail=validate_data,
+    max_retries=5  # Override default retry limit
+)
+```
+
 ## Creating Directories when Saving Files
 
 You can now specify if a task should create directories when saving its output to a file. This is particularly useful for organizing outputs and ensuring that file paths are correctly structured.
@@ -629,7 +877,7 @@ save_output_task = Task(
 
 ## Conclusion
 
-Tasks are the driving force behind the actions of agents in CrewAI. 
-By properly defining tasks and their outcomes, you set the stage for your AI agents to work effectively, either independently or as a collaborative unit. 
-Equipping tasks with appropriate tools, understanding the execution process, and following robust validation practices are crucial for maximizing CrewAI's potential, 
+Tasks are the driving force behind the actions of agents in CrewAI.
+By properly defining tasks and their outcomes, you set the stage for your AI agents to work effectively, either independently or as a collaborative unit.
+Equipping tasks with appropriate tools, understanding the execution process, and following robust validation practices are crucial for maximizing CrewAI's potential,
 ensuring agents are effectively prepared for their assignments and that tasks are executed as intended.

docs/concepts/tools.mdx

@@ -150,15 +150,20 @@ There are two main ways for one to create a CrewAI tool:
 
 ```python Code
 from crewai.tools import BaseTool
+from pydantic import BaseModel, Field
 
+class MyToolInput(BaseModel):
+    """Input schema for MyCustomTool."""
+    argument: str = Field(..., description="Description of the argument.")
 
 class MyCustomTool(BaseTool):
     name: str = "Name of my tool"
-    description: str = "Clear description for what this tool is useful for, your agent will need this information to use it."
+    description: str = "What this tool does. It's vital for effective utilization."
+    args_schema: Type[BaseModel] = MyToolInput
 
     def _run(self, argument: str) -> str:
-        # Implementation goes here
-        return "Result from custom tool"
+        # Your tool's logic here
+        return "Tool's result"
 ```
 
 ### Utilizing the `tool` Decorator
@@ -172,6 +177,48 @@ def my_tool(question: str) -> str:
     return "Result from your custom tool"
 ```
 
+### Structured Tools
+
+The `StructuredTool` class wraps functions as tools, providing flexibility and validation while reducing boilerplate. It supports custom schemas and dynamic logic for seamless integration of complex functionalities.
+
+#### Example:
+Using `StructuredTool.from_function`, you can wrap a function that interacts with an external API or system, providing a structured interface. This enables robust validation and consistent execution, making it easier to integrate complex functionalities into your applications as demonstrated in the following example:
+
+```python
+from crewai.tools.structured_tool import CrewStructuredTool
+from pydantic import BaseModel
+
+# Define the schema for the tool's input using Pydantic
+class APICallInput(BaseModel):
+    endpoint: str
+    parameters: dict
+
+# Wrapper function to execute the API call
+def tool_wrapper(*args, **kwargs):
+    # Here, you would typically call the API using the parameters
+    # For demonstration, we'll return a placeholder string
+    return f"Call the API at {kwargs['endpoint']} with parameters {kwargs['parameters']}"
+
+# Create and return the structured tool
+def create_structured_tool():
+    return CrewStructuredTool.from_function(
+        name='Wrapper API',
+        description="A tool to wrap API calls with structured input.",
+        args_schema=APICallInput,
+        func=tool_wrapper,
+    )
+
+# Example usage
+structured_tool = create_structured_tool()
+
+# Execute the tool with structured input
+result = structured_tool._run(**{
+    "endpoint": "https://example.com/api",
+    "parameters": {"key1": "value1", "key2": "value2"}
+})
+print(result)  # Output: Call the API at https://example.com/api with parameters {'key1': 'value1', 'key2': 'value2'}
+```
+
 ### Custom Caching Mechanism
 
 <Tip>

docs/how-to/agentops-observability.mdx

@@ -57,7 +57,7 @@ This feature is useful for debugging and understanding how agents interact with
    <Step title="Install AgentOps">
       Install AgentOps with:
       ```bash
-      pip install crewai[agentops]
+      pip install 'crewai[agentops]'
       ```
       or
       ```bash

docs/how-to/custom-manager-agent.mdx

@@ -73,9 +73,9 @@ result = crew.kickoff()
 If you're using the hierarchical process and don't want to set a custom manager agent, you can specify the language model for the manager:
 
 ```python Code
-from langchain_openai import ChatOpenAI
+from crewai import LLM
 
-manager_llm = ChatOpenAI(model_name="gpt-4")
+manager_llm = LLM(model="gpt-4o")
 
 crew = Crew(
     agents=[researcher, writer],

docs/how-to/llm-connections.mdx

@@ -32,6 +32,7 @@ LiteLLM supports a wide range of providers, including but not limited to:
 - Cloudflare Workers AI
 - DeepInfra
 - Groq
+- SambaNova
 - [NVIDIA NIMs](https://docs.api.nvidia.com/nim/reference/models-1)
 - And many more!
 

docs/how-to/multimodal-agents.mdx

@@ -0,0 +1,138 @@
+---
+title: Using Multimodal Agents
+description: Learn how to enable and use multimodal capabilities in your agents for processing images and other non-text content within the CrewAI framework.
+icon: image
+---
+
+# Using Multimodal Agents
+
+CrewAI supports multimodal agents that can process both text and non-text content like images. This guide will show you how to enable and use multimodal capabilities in your agents.
+
+## Enabling Multimodal Capabilities
+
+To create a multimodal agent, simply set the `multimodal` parameter to `True` when initializing your agent:
+
+```python
+from crewai import Agent
+
+agent = Agent(
+    role="Image Analyst",
+    goal="Analyze and extract insights from images",
+    backstory="An expert in visual content interpretation with years of experience in image analysis",
+    multimodal=True  # This enables multimodal capabilities
+)
+```
+
+When you set `multimodal=True`, the agent is automatically configured with the necessary tools for handling non-text content, including the `AddImageTool`.
+
+## Working with Images
+
+The multimodal agent comes pre-configured with the `AddImageTool`, which allows it to process images. You don't need to manually add this tool - it's automatically included when you enable multimodal capabilities.
+
+Here's a complete example showing how to use a multimodal agent to analyze an image:
+
+```python
+from crewai import Agent, Task, Crew
+
+# Create a multimodal agent
+image_analyst = Agent(
+    role="Product Analyst",
+    goal="Analyze product images and provide detailed descriptions",
+    backstory="Expert in visual product analysis with deep knowledge of design and features",
+    multimodal=True
+)
+
+# Create a task for image analysis
+task = Task(
+    description="Analyze the product image at https://example.com/product.jpg and provide a detailed description",
+    agent=image_analyst
+)
+
+# Create and run the crew
+crew = Crew(
+    agents=[image_analyst],
+    tasks=[task]
+)
+
+result = crew.kickoff()
+```
+
+### Advanced Usage with Context
+
+You can provide additional context or specific questions about the image when creating tasks for multimodal agents. The task description can include specific aspects you want the agent to focus on:
+
+```python
+from crewai import Agent, Task, Crew
+
+# Create a multimodal agent for detailed analysis
+expert_analyst = Agent(
+    role="Visual Quality Inspector",
+    goal="Perform detailed quality analysis of product images",
+    backstory="Senior quality control expert with expertise in visual inspection",
+    multimodal=True  # AddImageTool is automatically included
+)
+
+# Create a task with specific analysis requirements
+inspection_task = Task(
+    description="""
+    Analyze the product image at https://example.com/product.jpg with focus on:
+    1. Quality of materials
+    2. Manufacturing defects
+    3. Compliance with standards
+    Provide a detailed report highlighting any issues found.
+    """,
+    agent=expert_analyst
+)
+
+# Create and run the crew
+crew = Crew(
+    agents=[expert_analyst],
+    tasks=[inspection_task]
+)
+
+result = crew.kickoff()
+```
+
+### Tool Details
+
+When working with multimodal agents, the `AddImageTool` is automatically configured with the following schema:
+
+```python
+class AddImageToolSchema:
+    image_url: str  # Required: The URL or path of the image to process
+    action: Optional[str] = None  # Optional: Additional context or specific questions about the image
+```
+
+The multimodal agent will automatically handle the image processing through its built-in tools, allowing it to:
+- Access images via URLs or local file paths
+- Process image content with optional context or specific questions
+- Provide analysis and insights based on the visual information and task requirements
+
+## Best Practices
+
+When working with multimodal agents, keep these best practices in mind:
+
+1. **Image Access**
+   - Ensure your images are accessible via URLs that the agent can reach
+   - For local images, consider hosting them temporarily or using absolute file paths
+   - Verify that image URLs are valid and accessible before running tasks
+
+2. **Task Description**
+   - Be specific about what aspects of the image you want the agent to analyze
+   - Include clear questions or requirements in the task description
+   - Consider using the optional `action` parameter for focused analysis
+
+3. **Resource Management**
+   - Image processing may require more computational resources than text-only tasks
+   - Some language models may require base64 encoding for image data
+   - Consider batch processing for multiple images to optimize performance
+
+4. **Environment Setup**
+   - Verify that your environment has the necessary dependencies for image processing
+   - Ensure your language model supports multimodal capabilities
+   - Test with small images first to validate your setup
+
+5. **Error Handling**
+   - Implement proper error handling for image loading failures
+   - Have fallback strategies for when image processing fails
+   - Monitor and log image processing operations for debugging

docs/how-to/portkey-observability-and-guardrails.mdx

@@ -0,0 +1,211 @@
+# Portkey Integration with CrewAI
+<img src="https://raw.githubusercontent.com/siddharthsambharia-portkey/Portkey-Product-Images/main/Portkey-CrewAI.png" alt="Portkey CrewAI Header Image" width="70%" />
+
+
+[Portkey](https://portkey.ai/?utm_source=crewai&utm_medium=crewai&utm_campaign=crewai) is a 2-line upgrade to make your CrewAI agents reliable, cost-efficient, and fast.
+
+Portkey adds 4 core production capabilities to any CrewAI agent:
+1. Routing to **200+ LLMs**
+2. Making each LLM call more robust
+3. Full-stack tracing & cost, performance analytics
+4. Real-time guardrails to enforce behavior
+
+
+
+
+
+## Getting Started
+
+1. **Install Required Packages:**
+
+```bash
+pip install -qU crewai portkey-ai
+```
+
+2. **Configure the LLM Client:**
+
+To build CrewAI Agents with Portkey, you'll need two keys:
+- **Portkey API Key**: Sign up on the [Portkey app](https://app.portkey.ai/?utm_source=crewai&utm_medium=crewai&utm_campaign=crewai) and copy your API key
+- **Virtual Key**: Virtual Keys securely manage your LLM API keys in one place. Store your LLM provider API keys securely in Portkey's vault
+
+```python
+from crewai import LLM
+from portkey_ai import createHeaders, PORTKEY_GATEWAY_URL
+
+gpt_llm = LLM(
+    model="gpt-4",
+    base_url=PORTKEY_GATEWAY_URL,
+    api_key="dummy", # We are using Virtual key
+    extra_headers=createHeaders(
+        api_key="YOUR_PORTKEY_API_KEY",
+        virtual_key="YOUR_VIRTUAL_KEY", # Enter your Virtual key from Portkey
+    )
+)
+```
+
+3. **Create and Run Your First Agent:**
+
+```python
+from crewai import Agent, Task, Crew
+
+# Define your agents with roles and goals
+coder = Agent(
+    role='Software developer',
+    goal='Write clear, concise code on demand',
+    backstory='An expert coder with a keen eye for software trends.',
+    llm=gpt_llm
+)
+
+# Create tasks for your agents
+task1 = Task(
+    description="Define the HTML for making a simple website with heading- Hello World! Portkey is working!",
+    expected_output="A clear and concise HTML code",
+    agent=coder
+)
+
+# Instantiate your crew
+crew = Crew(
+    agents=[coder],
+    tasks=[task1],
+)
+
+result = crew.kickoff()
+print(result)
+```
+
+
+## Key Features
+
+| Feature | Description |
+|---------|-------------|
+| 🌐 Multi-LLM Support | Access OpenAI, Anthropic, Gemini, Azure, and 250+ providers through a unified interface |
+| 🛡️ Production Reliability | Implement retries, timeouts, load balancing, and fallbacks |
+| 📊 Advanced Observability | Track 40+ metrics including costs, tokens, latency, and custom metadata |
+| 🔍 Comprehensive Logging | Debug with detailed execution traces and function call logs |
+| 🚧 Security Controls | Set budget limits and implement role-based access control |
+| 🔄 Performance Analytics | Capture and analyze feedback for continuous improvement |
+| 💾 Intelligent Caching | Reduce costs and latency with semantic or simple caching |
+
+
+## Production Features with Portkey Configs
+
+All features mentioned below are through Portkey's Config system. Portkey's Config system allows you to define routing strategies using simple JSON objects in your LLM API calls. You can create and manage Configs directly in your code or through the Portkey Dashboard. Each Config has a unique ID for easy reference.
+
+<Frame>
+    <img src="https://raw.githubusercontent.com/Portkey-AI/docs-core/refs/heads/main/images/libraries/libraries-3.avif"/>
+</Frame>
+
+
+### 1. Use 250+ LLMs
+Access various LLMs like Anthropic, Gemini, Mistral, Azure OpenAI, and more with minimal code changes. Switch between providers or use them together seamlessly. [Learn more about Universal API](https://portkey.ai/docs/product/ai-gateway/universal-api)
+
+
+Easily switch between different LLM providers:
+
+```python
+# Anthropic Configuration
+anthropic_llm = LLM(
+    model="claude-3-5-sonnet-latest",
+    base_url=PORTKEY_GATEWAY_URL,
+    api_key="dummy",
+    extra_headers=createHeaders(
+        api_key="YOUR_PORTKEY_API_KEY",
+        virtual_key="YOUR_ANTHROPIC_VIRTUAL_KEY", #You don't need provider when using Virtual keys
+        trace_id="anthropic_agent"
+    )
+)
+
+# Azure OpenAI Configuration
+azure_llm = LLM(
+    model="gpt-4",
+    base_url=PORTKEY_GATEWAY_URL,
+    api_key="dummy",
+    extra_headers=createHeaders(
+        api_key="YOUR_PORTKEY_API_KEY",
+        virtual_key="YOUR_AZURE_VIRTUAL_KEY", #You don't need provider when using Virtual keys
+        trace_id="azure_agent"
+    )
+)
+```
+
+
+### 2. Caching
+Improve response times and reduce costs with two powerful caching modes:
+- **Simple Cache**: Perfect for exact matches
+- **Semantic Cache**: Matches responses for requests that are semantically similar
+[Learn more about Caching](https://portkey.ai/docs/product/ai-gateway/cache-simple-and-semantic)
+
+```py
+config = {
+    "cache": {
+        "mode": "semantic",  # or "simple" for exact matching
+    }
+}
+```
+
+### 3. Production Reliability
+Portkey provides comprehensive reliability features:
+- **Automatic Retries**: Handle temporary failures gracefully
+- **Request Timeouts**: Prevent hanging operations
+- **Conditional Routing**: Route requests based on specific conditions
+- **Fallbacks**: Set up automatic provider failovers
+- **Load Balancing**: Distribute requests efficiently
+
+[Learn more about Reliability Features](https://portkey.ai/docs/product/ai-gateway/)
+
+
+
+### 4. Metrics
+
+Agent runs are complex. Portkey automatically logs **40+ comprehensive metrics** for your AI agents, including cost, tokens used, latency, etc. Whether you need a broad overview or granular insights into your agent runs, Portkey's customizable filters provide the metrics you need.
+
+
+- Cost per agent interaction
+- Response times and latency
+- Token usage and efficiency
+- Success/failure rates
+- Cache hit rates
+
+<img src="https://github.com/siddharthsambharia-portkey/Portkey-Product-Images/blob/main/Portkey-Dashboard.png?raw=true" width="70%" alt="Portkey Dashboard" />
+
+### 5. Detailed Logging
+Logs are essential for understanding agent behavior, diagnosing issues, and improving performance. They provide a detailed record of agent activities and tool use, which is crucial for debugging and optimizing processes.
+
+
+Access a dedicated section to view records of agent executions, including parameters, outcomes, function calls, and errors. Filter logs based on multiple parameters such as trace ID, model, tokens used, and metadata.
+
+<details>
+  <summary><b>Traces</b></summary>
+  <img src="https://raw.githubusercontent.com/siddharthsambharia-portkey/Portkey-Product-Images/main/Portkey-Traces.png" alt="Portkey Traces" width="70%" />
+</details>
+
+<details>
+  <summary><b>Logs</b></summary>
+  <img src="https://raw.githubusercontent.com/siddharthsambharia-portkey/Portkey-Product-Images/main/Portkey-Logs.png" alt="Portkey Logs" width="70%" />
+</details>
+
+### 6. Enterprise Security Features
+- Set budget limit and rate limts per Virtual Key (disposable API keys)
+- Implement role-based access control
+- Track system changes with audit logs
+- Configure data retention policies
+
+
+
+For detailed information on creating and managing Configs, visit the [Portkey documentation](https://docs.portkey.ai/product/ai-gateway/configs).
+
+## Resources
+
+- [📘 Portkey Documentation](https://docs.portkey.ai)
+- [📊 Portkey Dashboard](https://app.portkey.ai/?utm_source=crewai&utm_medium=crewai&utm_campaign=crewai)
+- [🐦 Twitter](https://twitter.com/portkeyai)
+- [💬 Discord Community](https://discord.gg/DD7vgKK299)
+
+
+
+
+
+
+
+
+

docs/how-to/portkey-observability.mdx

@@ -0,0 +1,202 @@
+---
+title: Portkey Observability and Guardrails
+description: How to use Portkey with CrewAI
+icon: key
+---
+
+<img src="https://raw.githubusercontent.com/siddharthsambharia-portkey/Portkey-Product-Images/main/Portkey-CrewAI.png" alt="Portkey CrewAI Header Image" width="70%" />
+
+
+[Portkey](https://portkey.ai/?utm_source=crewai&utm_medium=crewai&utm_campaign=crewai) is a 2-line upgrade to make your CrewAI agents reliable, cost-efficient, and fast.
+
+Portkey adds 4 core production capabilities to any CrewAI agent:
+1. Routing to **200+ LLMs**
+2. Making each LLM call more robust
+3. Full-stack tracing & cost, performance analytics
+4. Real-time guardrails to enforce behavior
+
+## Getting Started
+
+<Steps>
+    <Step title="Install CrewAI and Portkey">
+    ```bash
+    pip install -qU crewai portkey-ai
+    ```
+    </Step>
+    <Step title="Configure the LLM Client">
+    To build CrewAI Agents with Portkey, you'll need two keys:
+    - **Portkey API Key**: Sign up on the [Portkey app](https://app.portkey.ai/?utm_source=crewai&utm_medium=crewai&utm_campaign=crewai) and copy your API key
+    - **Virtual Key**: Virtual Keys securely manage your LLM API keys in one place. Store your LLM provider API keys securely in Portkey's vault
+
+    ```python
+    from crewai import LLM
+    from portkey_ai import createHeaders, PORTKEY_GATEWAY_URL
+
+    gpt_llm = LLM(
+        model="gpt-4",
+        base_url=PORTKEY_GATEWAY_URL,
+        api_key="dummy", # We are using Virtual key
+        extra_headers=createHeaders(
+            api_key="YOUR_PORTKEY_API_KEY",
+            virtual_key="YOUR_VIRTUAL_KEY", # Enter your Virtual key from Portkey
+        )
+    )
+    ```
+    </Step>
+    <Step title="Create and Run Your First Agent">
+    ```python
+    from crewai import Agent, Task, Crew
+
+    # Define your agents with roles and goals
+    coder = Agent(
+        role='Software developer',
+        goal='Write clear, concise code on demand',
+        backstory='An expert coder with a keen eye for software trends.',
+        llm=gpt_llm
+    )
+
+    # Create tasks for your agents
+    task1 = Task(
+        description="Define the HTML for making a simple website with heading- Hello World! Portkey is working!",
+        expected_output="A clear and concise HTML code",
+        agent=coder
+    )
+
+    # Instantiate your crew
+    crew = Crew(
+        agents=[coder],
+        tasks=[task1],
+    )
+
+    result = crew.kickoff()
+    print(result)
+    ```
+    </Step>
+</Steps>
+
+## Key Features
+
+| Feature | Description |
+|:--------|:------------|
+| 🌐 Multi-LLM Support | Access OpenAI, Anthropic, Gemini, Azure, and 250+ providers through a unified interface |
+| 🛡️ Production Reliability | Implement retries, timeouts, load balancing, and fallbacks |
+| 📊 Advanced Observability | Track 40+ metrics including costs, tokens, latency, and custom metadata |
+| 🔍 Comprehensive Logging | Debug with detailed execution traces and function call logs |
+| 🚧 Security Controls | Set budget limits and implement role-based access control |
+| 🔄 Performance Analytics | Capture and analyze feedback for continuous improvement |
+| 💾 Intelligent Caching | Reduce costs and latency with semantic or simple caching |
+
+
+## Production Features with Portkey Configs
+
+All features mentioned below are through Portkey's Config system. Portkey's Config system allows you to define routing strategies using simple JSON objects in your LLM API calls. You can create and manage Configs directly in your code or through the Portkey Dashboard. Each Config has a unique ID for easy reference.
+
+<Frame>
+    <img src="https://raw.githubusercontent.com/Portkey-AI/docs-core/refs/heads/main/images/libraries/libraries-3.avif"/>
+</Frame>
+
+
+### 1. Use 250+ LLMs
+Access various LLMs like Anthropic, Gemini, Mistral, Azure OpenAI, and more with minimal code changes. Switch between providers or use them together seamlessly. [Learn more about Universal API](https://portkey.ai/docs/product/ai-gateway/universal-api)
+
+
+Easily switch between different LLM providers:
+
+```python
+# Anthropic Configuration
+anthropic_llm = LLM(
+    model="claude-3-5-sonnet-latest",
+    base_url=PORTKEY_GATEWAY_URL,
+    api_key="dummy",
+    extra_headers=createHeaders(
+        api_key="YOUR_PORTKEY_API_KEY",
+        virtual_key="YOUR_ANTHROPIC_VIRTUAL_KEY", #You don't need provider when using Virtual keys
+        trace_id="anthropic_agent"
+    )
+)
+
+# Azure OpenAI Configuration
+azure_llm = LLM(
+    model="gpt-4",
+    base_url=PORTKEY_GATEWAY_URL,
+    api_key="dummy",
+    extra_headers=createHeaders(
+        api_key="YOUR_PORTKEY_API_KEY",
+        virtual_key="YOUR_AZURE_VIRTUAL_KEY", #You don't need provider when using Virtual keys
+        trace_id="azure_agent"
+    )
+)
+```
+
+
+### 2. Caching
+Improve response times and reduce costs with two powerful caching modes:
+- **Simple Cache**: Perfect for exact matches
+- **Semantic Cache**: Matches responses for requests that are semantically similar
+[Learn more about Caching](https://portkey.ai/docs/product/ai-gateway/cache-simple-and-semantic)
+
+```py
+config = {
+    "cache": {
+        "mode": "semantic",  # or "simple" for exact matching
+    }
+}
+```
+
+### 3. Production Reliability
+Portkey provides comprehensive reliability features:
+- **Automatic Retries**: Handle temporary failures gracefully
+- **Request Timeouts**: Prevent hanging operations
+- **Conditional Routing**: Route requests based on specific conditions
+- **Fallbacks**: Set up automatic provider failovers
+- **Load Balancing**: Distribute requests efficiently
+
+[Learn more about Reliability Features](https://portkey.ai/docs/product/ai-gateway/)
+
+
+
+### 4. Metrics
+
+Agent runs are complex. Portkey automatically logs **40+ comprehensive metrics** for your AI agents, including cost, tokens used, latency, etc. Whether you need a broad overview or granular insights into your agent runs, Portkey's customizable filters provide the metrics you need.
+
+
+- Cost per agent interaction
+- Response times and latency
+- Token usage and efficiency
+- Success/failure rates
+- Cache hit rates
+
+<img src="https://github.com/siddharthsambharia-portkey/Portkey-Product-Images/blob/main/Portkey-Dashboard.png?raw=true" width="70%" alt="Portkey Dashboard" />
+
+### 5. Detailed Logging
+Logs are essential for understanding agent behavior, diagnosing issues, and improving performance. They provide a detailed record of agent activities and tool use, which is crucial for debugging and optimizing processes.
+
+
+Access a dedicated section to view records of agent executions, including parameters, outcomes, function calls, and errors. Filter logs based on multiple parameters such as trace ID, model, tokens used, and metadata.
+
+<details>
+  <summary><b>Traces</b></summary>
+  <img src="https://raw.githubusercontent.com/siddharthsambharia-portkey/Portkey-Product-Images/main/Portkey-Traces.png" alt="Portkey Traces" width="70%" />
+</details>
+
+<details>
+  <summary><b>Logs</b></summary>
+  <img src="https://raw.githubusercontent.com/siddharthsambharia-portkey/Portkey-Product-Images/main/Portkey-Logs.png" alt="Portkey Logs" width="70%" />
+</details>
+
+### 6. Enterprise Security Features
+- Set budget limit and rate limts per Virtual Key (disposable API keys)
+- Implement role-based access control
+- Track system changes with audit logs
+- Configure data retention policies
+
+
+
+For detailed information on creating and managing Configs, visit the [Portkey documentation](https://docs.portkey.ai/product/ai-gateway/configs).
+
+## Resources
+
+- [📘 Portkey Documentation](https://docs.portkey.ai)
+- [📊 Portkey Dashboard](https://app.portkey.ai/?utm_source=crewai&utm_medium=crewai&utm_campaign=crewai)
+- [🐦 Twitter](https://twitter.com/portkeyai)
+- [💬 Discord Community](https://discord.gg/DD7vgKK299)

docs/installation.mdx

@@ -7,7 +7,7 @@ icon: wrench
 <Note>
   **Python Version Requirements**
   
-  CrewAI requires `Python >=3.10 and <=3.13`. Here's how to check your version:
+  CrewAI requires `Python >=3.10 and <3.13`. Here's how to check your version:
   ```bash
   python3 --version
   ```

docs/mint.json

@@ -100,7 +100,8 @@
         "how-to/conditional-tasks",
         "how-to/agentops-observability",
         "how-to/langtrace-observability",
-        "how-to/openlit-observability"
+        "how-to/openlit-observability",
+        "how-to/portkey-observability"
       ]
     },
     {

docs/quickstart.mdx

@@ -301,38 +301,166 @@ Use the annotations to properly reference the agent and task in the `crew.py` fi
 
 ### Annotations include:
 
-* `@agent`
-* `@task`
-* `@crew`
-* `@tool`
-* `@before_kickoff`
-* `@after_kickoff`
-* `@callback`
-* `@output_json`
-* `@output_pydantic`
-* `@cache_handler`
+Here are examples of how to use each annotation in your CrewAI project, and when you should use them:
 
-```python crew.py
-# ...
+#### @agent
+Used to define an agent in your crew. Use this when:
+- You need to create a specialized AI agent with a specific role
+- You want the agent to be automatically collected and managed by the crew
+- You need to reuse the same agent configuration across multiple tasks
+
+```python
 @agent
-def email_summarizer(self) -> Agent:
+def research_agent(self) -> Agent:
     return Agent(
-        config=self.agents_config["email_summarizer"],
+        role="Research Analyst",
+        goal="Conduct thorough research on given topics",
+        backstory="Expert researcher with years of experience in data analysis",
+        tools=[SerperDevTool()],
+        verbose=True
     )
-
-@task
-def email_summarizer_task(self) -> Task:
-    return Task(
-        config=self.tasks_config["email_summarizer_task"],
-    )
-# ...
 ```
 
-<Tip>
-In addition to the [sequential process](../how-to/sequential-process), you can use the [hierarchical process](../how-to/hierarchical-process),
-which automatically assigns a manager to the defined crew to properly coordinate the planning and execution of tasks through delegation and validation of results.
-You can learn more about the core concepts [here](/concepts).
-</Tip>
+#### @task
+Used to define a task that can be executed by agents. Use this when:
+- You need to define a specific piece of work for an agent
+- You want tasks to be automatically sequenced and managed
+- You need to establish dependencies between different tasks
+
+```python
+@task
+def research_task(self) -> Task:
+    return Task(
+        description="Research the latest developments in AI technology",
+        expected_output="A comprehensive report on AI advancements",
+        agent=self.research_agent(),
+        output_file="output/research.md"
+    )
+```
+
+#### @crew
+Used to define your crew configuration. Use this when:
+- You want to automatically collect all @agent and @task definitions
+- You need to specify how tasks should be processed (sequential or hierarchical)
+- You want to set up crew-wide configurations
+
+```python
+@crew
+def research_crew(self) -> Crew:
+    return Crew(
+        agents=self.agents,  # Automatically collected from @agent methods
+        tasks=self.tasks,    # Automatically collected from @task methods
+        process=Process.sequential,
+        verbose=True
+    )
+```
+
+#### @tool
+Used to create custom tools for your agents. Use this when:
+- You need to give agents specific capabilities (like web search, data analysis)
+- You want to encapsulate external API calls or complex operations
+- You need to share functionality across multiple agents
+
+```python
+@tool
+def web_search_tool(query: str, max_results: int = 5) -> list[str]:
+    """
+    Search the web for information.
+
+    Args:
+        query: The search query
+        max_results: Maximum number of results to return
+
+    Returns:
+        List of search results
+    """
+    # Implement your search logic here
+    return [f"Result {i} for: {query}" for i in range(max_results)]
+```
+
+#### @before_kickoff
+Used to execute logic before the crew starts. Use this when:
+- You need to validate or preprocess input data
+- You want to set up resources or configurations before execution
+- You need to perform any initialization logic
+
+```python
+@before_kickoff
+def validate_inputs(self, inputs: Optional[Dict[str, Any]]) -> Optional[Dict[str, Any]]:
+    """Validate and preprocess inputs before the crew starts."""
+    if inputs is None:
+        return None
+        
+    if 'topic' not in inputs:
+        raise ValueError("Topic is required")
+    
+    # Add additional context
+    inputs['timestamp'] = datetime.now().isoformat()
+    inputs['topic'] = inputs['topic'].strip().lower()
+    return inputs
+```
+
+#### @after_kickoff
+Used to process results after the crew completes. Use this when:
+- You need to format or transform the final output
+- You want to perform cleanup operations
+- You need to save or log the results in a specific way
+
+```python
+@after_kickoff
+def process_results(self, result: CrewOutput) -> CrewOutput:
+    """Process and format the results after the crew completes."""
+    result.raw = result.raw.strip()
+    result.raw = f"""
+    # Research Results
+    Generated on: {datetime.now().isoformat()}
+    
+    {result.raw}
+    """
+    return result
+```
+
+#### @callback
+Used to handle events during crew execution. Use this when:
+- You need to monitor task progress
+- You want to log intermediate results
+- You need to implement custom progress tracking or metrics
+
+```python
+@callback
+def log_task_completion(self, task: Task, output: str):
+    """Log task completion details for monitoring."""
+    print(f"Task '{task.description}' completed")
+    print(f"Output length: {len(output)} characters")
+    print(f"Agent used: {task.agent.role}")
+    print("-" * 50)
+```
+
+#### @cache_handler
+Used to implement custom caching for task results. Use this when:
+- You want to avoid redundant expensive operations
+- You need to implement custom cache storage or expiration logic
+- You want to persist results between runs
+
+```python
+@cache_handler
+def custom_cache(self, key: str) -> Optional[str]:
+    """Custom cache implementation for storing task results."""
+    cache_file = f"cache/{key}.json"
+    
+    if os.path.exists(cache_file):
+        with open(cache_file, 'r') as f:
+            data = json.load(f)
+            # Check if cache is still valid (e.g., not expired)
+            if datetime.fromisoformat(data['timestamp']) > datetime.now() - timedelta(days=1):
+                return data['result']
+    return None
+```
+
+<Note>
+These decorators are part of the CrewAI framework and help organize your crew's structure by automatically collecting agents, tasks, and handling various lifecycle events. 
+They should be used within a class decorated with `@CrewBase`.
+</Note>
 
 ### Replay Tasks from Latest Crew Kickoff
 

pyproject.toml

@@ -1,35 +1,46 @@
 [project]
 name = "crewai"
-version = "0.86.0"
+version = "0.95.0"
 description = "Cutting-edge framework for orchestrating role-playing, autonomous AI agents. By fostering collaborative intelligence, CrewAI empowers agents to work together seamlessly, tackling complex tasks."
 readme = "README.md"
-requires-python = ">=3.10,<=3.13"
+requires-python = ">=3.10,<3.13"
 authors = [
     { name = "Joao Moura", email = "joao@crewai.com" }
 ]
 dependencies = [
+    # Core Dependencies
     "pydantic>=2.4.2",
     "openai>=1.13.3",
+    "litellm==1.57.4",
+    "instructor>=1.3.3",
+
+    # Text Processing
+    "pdfplumber>=0.11.4",
+    "regex>=2024.9.11",
+
+    # Telemetry and Monitoring
     "opentelemetry-api>=1.22.0",
     "opentelemetry-sdk>=1.22.0",
     "opentelemetry-exporter-otlp-proto-http>=1.22.0",
-    "instructor>=1.3.3",
-    "regex>=2024.9.11",
-    "crewai-tools>=0.17.0",
-    "click>=8.1.7",
+
+    # Data Handling
+    "chromadb>=0.5.23",
+    "openpyxl>=3.1.5",
+    "pyvis>=0.3.2",
+
+    # Authentication and Security
+    "auth0-python>=4.7.1",
     "python-dotenv>=1.0.0",
+
+    # Configuration and Utils
+    "click>=8.1.7",
     "appdirs>=1.4.4",
     "jsonref>=1.1.0",
     "json-repair>=0.25.2",
-    "auth0-python>=4.7.1",
-    "litellm>=1.44.22",
-    "pyvis>=0.3.2",
     "uv>=0.4.25",
     "tomli-w>=1.1.0",
     "tomli>=2.0.2",
-    "chromadb>=0.5.18",
-    "pdfplumber>=0.11.4",
-    "openpyxl>=3.1.5",
+    "blinker>=1.9.0"
 ]
 
 [project.urls]
@@ -38,7 +49,10 @@ Documentation = "https://docs.crewai.com"
 Repository = "https://github.com/crewAIInc/crewAI"
 
 [project.optional-dependencies]
-tools = ["crewai-tools>=0.14.0"]
+tools = ["crewai-tools>=0.25.5"]
+embeddings = [
+    "tiktoken~=0.7.0"
+]
 agentops = ["agentops>=0.3.0"]
 fastembed = ["fastembed>=0.4.1"]
 pdfplumber = [
@@ -51,10 +65,13 @@ openpyxl = [
     "openpyxl>=3.1.5",
 ]
 mem0 = ["mem0ai>=0.1.29"]
+docling = [
+    "docling>=2.12.0",
+]
 
 [tool.uv]
 dev-dependencies = [
-    "ruff>=0.4.10",
+    "ruff>=0.8.2",
     "mypy>=1.10.0",
     "pre-commit>=3.6.0",
     "mkdocs>=1.4.3",
@@ -64,7 +81,6 @@ dev-dependencies = [
     "mkdocs-material-extensions>=1.3.1",
     "pillow>=10.2.0",
     "cairosvg>=2.7.1",
-    "crewai-tools>=0.14.0",
     "pytest>=8.0.0",
     "pytest-vcr>=1.0.2",
     "python-dotenv>=1.0.0",

src/crewai/__init__.py

@@ -14,7 +14,7 @@ warnings.filterwarnings(
     category=UserWarning,
     module="pydantic.main",
 )
-__version__ = "0.86.0"
+__version__ = "0.95.0"
 __all__ = [
     "Agent",
     "Crew",

src/crewai/agent.py

@@ -17,33 +17,27 @@ from crewai.memory.contextual.contextual_memory import ContextualMemory
 from crewai.task import Task
 from crewai.tools import BaseTool
 from crewai.tools.agent_tools.agent_tools import AgentTools
+from crewai.tools.base_tool import Tool
 from crewai.utilities import Converter, Prompts
 from crewai.utilities.constants import TRAINED_AGENTS_DATA_FILE, TRAINING_DATA_FILE
 from crewai.utilities.converter import generate_model_description
+from crewai.utilities.llm_utils import create_llm
 from crewai.utilities.token_counter_callback import TokenCalcHandler
 from crewai.utilities.training_handler import CrewTrainingHandler
 
+agentops = None
 
-def mock_agent_ops_provider():
-    def track_agent(*args, **kwargs):
+try:
+    import agentops  # type: ignore # Name "agentops" is already defined
+    from agentops import track_agent  # type: ignore
+except ImportError:
+
+    def track_agent():
         def noop(f):
             return f
 
         return noop
 
-    return track_agent
-
-
-agentops = None
-
-if os.environ.get("AGENTOPS_API_KEY"):
-    try:
-        from agentops import track_agent
-    except ImportError:
-        track_agent = mock_agent_ops_provider()
-else:
-    track_agent = mock_agent_ops_provider()
-
 
 @track_agent()
 class Agent(BaseAgent):
@@ -92,7 +86,7 @@ class Agent(BaseAgent):
     llm: Union[str, InstanceOf[LLM], Any] = Field(
         description="Language model that will run the agent.", default=None
     )
-    function_calling_llm: Optional[Any] = Field(
+    function_calling_llm: Optional[Union[str, InstanceOf[LLM], Any]] = Field(
         description="Language model that will run the agent.", default=None
     )
     system_template: Optional[str] = Field(
@@ -122,6 +116,10 @@ class Agent(BaseAgent):
         default=2,
         description="Maximum number of retries for an agent to execute a task when an error occurs.",
     )
+    multimodal: bool = Field(
+        default=False,
+        description="Whether the agent is multimodal.",
+    )
     code_execution_mode: Literal["safe", "unsafe"] = Field(
         default="safe",
         description="Mode for code execution: 'safe' (using Docker) or 'unsafe' (direct execution).",
@@ -142,89 +140,10 @@ class Agent(BaseAgent):
     def post_init_setup(self):
         self._set_knowledge()
         self.agent_ops_agent_name = self.role
-        unaccepted_attributes = [
-            "AWS_ACCESS_KEY_ID",
-            "AWS_SECRET_ACCESS_KEY",
-            "AWS_REGION_NAME",
-        ]
 
-        # Handle different cases for self.llm
-        if isinstance(self.llm, str):
-            # If it's a string, create an LLM instance
-            self.llm = LLM(model=self.llm)
-        elif isinstance(self.llm, LLM):
-            # If it's already an LLM instance, keep it as is
-            pass
-        elif self.llm is None:
-            # Determine the model name from environment variables or use default
-            model_name = (
-                os.environ.get("OPENAI_MODEL_NAME")
-                or os.environ.get("MODEL")
-                or "gpt-4o-mini"
-            )
-            llm_params = {"model": model_name}
-
-            api_base = os.environ.get("OPENAI_API_BASE") or os.environ.get(
-                "OPENAI_BASE_URL"
-            )
-            if api_base:
-                llm_params["base_url"] = api_base
-
-            set_provider = model_name.split("/")[0] if "/" in model_name else "openai"
-
-            # Iterate over all environment variables to find matching API keys or use defaults
-            for provider, env_vars in ENV_VARS.items():
-                if provider == set_provider:
-                    for env_var in env_vars:
-                        # Check if the environment variable is set
-                        key_name = env_var.get("key_name")
-                        if key_name and key_name not in unaccepted_attributes:
-                            env_value = os.environ.get(key_name)
-                            if env_value:
-                                key_name = key_name.lower()
-                                for pattern in LITELLM_PARAMS:
-                                    if pattern in key_name:
-                                        key_name = pattern
-                                        break
-                                llm_params[key_name] = env_value
-                        # Check for default values if the environment variable is not set
-                        elif env_var.get("default", False):
-                            for key, value in env_var.items():
-                                if key not in ["prompt", "key_name", "default"]:
-                                    # Only add default if the key is already set in os.environ
-                                    if key in os.environ:
-                                        llm_params[key] = value
-
-            self.llm = LLM(**llm_params)
-        else:
-            # For any other type, attempt to extract relevant attributes
-            llm_params = {
-                "model": getattr(self.llm, "model_name", None)
-                or getattr(self.llm, "deployment_name", None)
-                or str(self.llm),
-                "temperature": getattr(self.llm, "temperature", None),
-                "max_tokens": getattr(self.llm, "max_tokens", None),
-                "logprobs": getattr(self.llm, "logprobs", None),
-                "timeout": getattr(self.llm, "timeout", None),
-                "max_retries": getattr(self.llm, "max_retries", None),
-                "api_key": getattr(self.llm, "api_key", None),
-                "base_url": getattr(self.llm, "base_url", None),
-                "organization": getattr(self.llm, "organization", None),
-            }
-            # Remove None values to avoid passing unnecessary parameters
-            llm_params = {k: v for k, v in llm_params.items() if v is not None}
-            self.llm = LLM(**llm_params)
-
-        # Similar handling for function_calling_llm
-        if self.function_calling_llm:
-            if isinstance(self.function_calling_llm, str):
-                self.function_calling_llm = LLM(model=self.function_calling_llm)
-            elif not isinstance(self.function_calling_llm, LLM):
-                self.function_calling_llm = LLM(
-                    model=getattr(self.function_calling_llm, "model_name", None)
-                    or getattr(self.function_calling_llm, "deployment_name", None)
-                    or str(self.function_calling_llm)
-                )
+        self.llm = create_llm(self.llm)
+        if self.function_calling_llm and not isinstance(self.function_calling_llm, LLM):
+            self.function_calling_llm = create_llm(self.function_calling_llm)
 
         if not self.agent_executor:
             self._setup_agent_executor()
@@ -414,6 +333,11 @@ class Agent(BaseAgent):
         tools = agent_tools.tools()
         return tools
 
+    def get_multimodal_tools(self) -> List[Tool]:
+        from crewai.tools.agent_tools.add_image_tool import AddImageTool
+
+        return [AddImageTool()]
+
     def get_code_execution_tools(self):
         try:
             from crewai_tools import CodeInterpreterTool

src/crewai/agents/agent_builder/base_agent_executor_mixin.py

@@ -19,15 +19,10 @@ class CrewAgentExecutorMixin:
     agent: Optional["BaseAgent"]
     task: Optional["Task"]
     iterations: int
-    have_forced_answer: bool
     max_iter: int
     _i18n: I18N
     _printer: Printer = Printer()
 
-    def _should_force_answer(self) -> bool:
-        """Determine if a forced answer is required based on iteration count."""
-        return (self.iterations >= self.max_iter) and not self.have_forced_answer
-
     def _create_short_term_memory(self, output) -> None:
         """Create and save a short-term memory item if conditions are met."""
         if (

src/crewai/agents/crew_agent_executor.py

@@ -1,7 +1,7 @@
 import json
 import re
 from dataclasses import dataclass
-from typing import Any, Dict, List, Union
+from typing import Any, Callable, Dict, List, Optional, Union
 
 from crewai.agents.agent_builder.base_agent import BaseAgent
 from crewai.agents.agent_builder.base_agent_executor_mixin import CrewAgentExecutorMixin
@@ -50,7 +50,7 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
         original_tools: List[Any] = [],
         function_calling_llm: Any = None,
         respect_context_window: bool = False,
-        request_within_rpm_limit: Any = None,
+        request_within_rpm_limit: Optional[Callable[[], bool]] = None,
         callbacks: List[Any] = [],
     ):
         self._i18n: I18N = I18N()
@@ -77,7 +77,6 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
         self.messages: List[Dict[str, str]] = []
         self.iterations = 0
         self.log_error_after = 3
-        self.have_forced_answer = False
         self.tool_name_to_tool_map: Dict[str, BaseTool] = {
             tool.name: tool for tool in self.tools
         }
@@ -108,93 +107,149 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
         self._create_long_term_memory(formatted_answer)
         return {"output": formatted_answer.output}
 
-    def _invoke_loop(self, formatted_answer=None):
-        try:
-            while not isinstance(formatted_answer, AgentFinish):
-                if not self.request_within_rpm_limit or self.request_within_rpm_limit():
-                    answer = self.llm.call(
-                        self.messages,
-                        callbacks=self.callbacks,
+    def _invoke_loop(self):
+        """
+        Main loop to invoke the agent's thought process until it reaches a conclusion
+        or the maximum number of iterations is reached.
+        """
+        formatted_answer = None
+        while not isinstance(formatted_answer, AgentFinish):
+            try:
+                if self._has_reached_max_iterations():
+                    formatted_answer = self._handle_max_iterations_exceeded(
+                        formatted_answer
+                    )
+                    break
+
+                self._enforce_rpm_limit()
+
+                answer = self._get_llm_response()
+
+                formatted_answer = self._process_llm_response(answer)
+
+                if isinstance(formatted_answer, AgentAction):
+                    tool_result = self._execute_tool_and_check_finality(
+                        formatted_answer
+                    )
+                    formatted_answer = self._handle_agent_action(
+                        formatted_answer, tool_result
                     )
 
-                    if answer is None or answer == "":
-                        self._printer.print(
-                            content="Received None or empty response from LLM call.",
-                            color="red",
-                        )
-                        raise ValueError(
-                            "Invalid response from LLM call - None or empty."
-                        )
+                self._invoke_step_callback(formatted_answer)
+                self._append_message(formatted_answer.text, role="assistant")
 
-                    if not self.use_stop_words:
-                        try:
-                            self._format_answer(answer)
-                        except OutputParserException as e:
-                            if (
-                                FINAL_ANSWER_AND_PARSABLE_ACTION_ERROR_MESSAGE
-                                in e.error
-                            ):
-                                answer = answer.split("Observation:")[0].strip()
+            except OutputParserException as e:
+                formatted_answer = self._handle_output_parser_exception(e)
 
-                    self.iterations += 1
-                    formatted_answer = self._format_answer(answer)
-
-                    if isinstance(formatted_answer, AgentAction):
-                        tool_result = self._execute_tool_and_check_finality(
-                            formatted_answer
-                        )
-                        formatted_answer.text += f"\nObservation: {tool_result.result}"
-                        formatted_answer.result = tool_result.result
-                        if tool_result.result_as_answer:
-                            return AgentFinish(
-                                thought="",
-                                output=tool_result.result,
-                                text=formatted_answer.text,
-                            )
-                        self._show_logs(formatted_answer)
-
-                    if self.step_callback:
-                        self.step_callback(formatted_answer)
-
-                    if self._should_force_answer():
-                        if self.have_forced_answer:
-                            return AgentFinish(
-                                thought="",
-                                output=self._i18n.errors(
-                                    "force_final_answer_error"
-                                ).format(formatted_answer.text),
-                                text=formatted_answer.text,
-                            )
-                        else:
-                            formatted_answer.text += (
-                                f'\n{self._i18n.errors("force_final_answer")}'
-                            )
-                            self.have_forced_answer = True
-                    self.messages.append(
-                        self._format_msg(formatted_answer.text, role="assistant")
-                    )
-
-        except OutputParserException as e:
-            self.messages.append({"role": "user", "content": e.error})
-            if self.iterations > self.log_error_after:
-                self._printer.print(
-                    content=f"Error parsing LLM output, agent will retry: {e.error}",
-                    color="red",
-                )
-            return self._invoke_loop(formatted_answer)
-
-        except Exception as e:
-            if LLMContextLengthExceededException(str(e))._is_context_limit_error(
-                str(e)
-            ):
-                self._handle_context_length()
-                return self._invoke_loop(formatted_answer)
-            else:
-                raise e
+            except Exception as e:
+                if self._is_context_length_exceeded(e):
+                    self._handle_context_length()
+                    continue
 
         self._show_logs(formatted_answer)
         return formatted_answer
 
+    def _has_reached_max_iterations(self) -> bool:
+        """Check if the maximum number of iterations has been reached."""
+        return self.iterations >= self.max_iter
+
+    def _enforce_rpm_limit(self) -> None:
+        """Enforce the requests per minute (RPM) limit if applicable."""
+        if self.request_within_rpm_limit:
+            self.request_within_rpm_limit()
+
+    def _get_llm_response(self) -> str:
+        """Call the LLM and return the response, handling any invalid responses."""
+        answer = self.llm.call(
+            self.messages,
+            callbacks=self.callbacks,
+        )
+
+        if not answer:
+            self._printer.print(
+                content="Received None or empty response from LLM call.",
+                color="red",
+            )
+            raise ValueError("Invalid response from LLM call - None or empty.")
+
+        return answer
+
+    def _process_llm_response(self, answer: str) -> Union[AgentAction, AgentFinish]:
+        """Process the LLM response and format it into an AgentAction or AgentFinish."""
+        if not self.use_stop_words:
+            try:
+                # Preliminary parsing to check for errors.
+                self._format_answer(answer)
+            except OutputParserException as e:
+                if FINAL_ANSWER_AND_PARSABLE_ACTION_ERROR_MESSAGE in e.error:
+                    answer = answer.split("Observation:")[0].strip()
+
+        self.iterations += 1
+        return self._format_answer(answer)
+
+    def _handle_agent_action(
+        self, formatted_answer: AgentAction, tool_result: ToolResult
+    ) -> Union[AgentAction, AgentFinish]:
+        """Handle the AgentAction, execute tools, and process the results."""
+        add_image_tool = self._i18n.tools("add_image")
+        if (
+            isinstance(add_image_tool, dict)
+            and formatted_answer.tool.casefold().strip()
+            == add_image_tool.get("name", "").casefold().strip()
+        ):
+            self.messages.append(tool_result.result)
+            return formatted_answer  # Continue the loop
+
+        if self.step_callback:
+            self.step_callback(tool_result)
+
+        formatted_answer.text += f"\nObservation: {tool_result.result}"
+        formatted_answer.result = tool_result.result
+
+        if tool_result.result_as_answer:
+            return AgentFinish(
+                thought="",
+                output=tool_result.result,
+                text=formatted_answer.text,
+            )
+
+        self._show_logs(formatted_answer)
+        return formatted_answer
+
+    def _invoke_step_callback(self, formatted_answer) -> None:
+        """Invoke the step callback if it exists."""
+        if self.step_callback:
+            self.step_callback(formatted_answer)
+
+    def _append_message(self, text: str, role: str = "assistant") -> None:
+        """Append a message to the message list with the given role."""
+        self.messages.append(self._format_msg(text, role=role))
+
+    def _handle_output_parser_exception(self, e: OutputParserException) -> AgentAction:
+        """Handle OutputParserException by updating messages and formatted_answer."""
+        self.messages.append({"role": "user", "content": e.error})
+
+        formatted_answer = AgentAction(
+            text=e.error,
+            tool="",
+            tool_input="",
+            thought="",
+        )
+
+        if self.iterations > self.log_error_after:
+            self._printer.print(
+                content=f"Error parsing LLM output, agent will retry: {e.error}",
+                color="red",
+            )
+
+        return formatted_answer
+
+    def _is_context_length_exceeded(self, exception: Exception) -> bool:
+        """Check if the exception is due to context length exceeding."""
+        return LLMContextLengthExceededException(
+            str(exception)
+        )._is_context_limit_error(str(exception))
+
     def _show_start_logs(self):
         if self.agent is None:
             raise ValueError("Agent cannot be None")
@@ -259,7 +314,7 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
             agent=self.agent,
             action=agent_action,
         )
-        tool_calling = tool_usage.parse(agent_action.text)
+        tool_calling = tool_usage.parse_tool_calling(agent_action.text)
 
         if isinstance(tool_calling, ToolUsageErrorException):
             tool_result = tool_calling.message
@@ -299,7 +354,7 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
                         self._i18n.slice("summarizer_system_message"), role="system"
                     ),
                     self._format_msg(
-                        self._i18n.slice("sumamrize_instruction").format(group=group),
+                        self._i18n.slice("summarize_instruction").format(group=group),
                     ),
                 ],
                 callbacks=self.callbacks,
@@ -410,7 +465,6 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
         """
         while self.ask_for_human_input:
             human_feedback = self._ask_human_input(formatted_answer.output)
-            print("Human feedback: ", human_feedback)
 
             if self.crew and self.crew._train:
                 self._handle_crew_training_output(formatted_answer, human_feedback)
@@ -475,3 +529,45 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
                 self.ask_for_human_input = False
 
         return formatted_answer
+
+    def _handle_max_iterations_exceeded(self, formatted_answer):
+        """
+        Handles the case when the maximum number of iterations is exceeded.
+        Performs one more LLM call to get the final answer.
+
+        Parameters:
+            formatted_answer: The last formatted answer from the agent.
+
+        Returns:
+            The final formatted answer after exceeding max iterations.
+        """
+        self._printer.print(
+            content="Maximum iterations reached. Requesting final answer.",
+            color="yellow",
+        )
+
+        if formatted_answer and hasattr(formatted_answer, "text"):
+            assistant_message = (
+                formatted_answer.text + f'\n{self._i18n.errors("force_final_answer")}'
+            )
+        else:
+            assistant_message = self._i18n.errors("force_final_answer")
+
+        self.messages.append(self._format_msg(assistant_message, role="assistant"))
+
+        # Perform one more LLM call to get the final answer
+        answer = self.llm.call(
+            self.messages,
+            callbacks=self.callbacks,
+        )
+
+        if answer is None or answer == "":
+            self._printer.print(
+                content="Received None or empty response from LLM call.",
+                color="red",
+            )
+            raise ValueError("Invalid response from LLM call - None or empty.")
+
+        formatted_answer = self._format_answer(answer)
+        # Return the formatted answer, regardless of its type
+        return formatted_answer

src/crewai/agents/parser.py

@@ -1,5 +1,6 @@
 import re
 from typing import Any, Union
+
 from json_repair import repair_json
 
 from crewai.utilities import I18N

src/crewai/cli/authentication/main.py

@@ -5,9 +5,10 @@ from typing import Any, Dict
 import requests
 from rich.console import Console
 
+from crewai.cli.tools.main import ToolCommand
+
 from .constants import AUTH0_AUDIENCE, AUTH0_CLIENT_ID, AUTH0_DOMAIN
 from .utils import TokenManager, validate_token
-from crewai.cli.tools.main import ToolCommand
 
 console = Console()
 
@@ -79,7 +80,9 @@ class AuthenticationCommand:
                         style="yellow",
                     )
 
-                console.print("\n[bold green]Welcome to CrewAI Enterprise![/bold green]\n")
+                console.print(
+                    "\n[bold green]Welcome to CrewAI Enterprise![/bold green]\n"
+                )
                 return
 
             if token_data["error"] not in ("authorization_pending", "slow_down"):

src/crewai/cli/authentication/token.py

@@ -1,10 +1,9 @@
 from .utils import TokenManager
 
+
 def get_auth_token() -> str:
     """Get the authentication token."""
     access_token = TokenManager().get_token()
     if not access_token:
         raise Exception()
     return access_token
-
-

src/crewai/cli/cli.py

@@ -1,11 +1,13 @@
-from typing import Optional
+import os
+from importlib.metadata import version as get_version
+from typing import Optional, Tuple
 
 import click
-import pkg_resources
 
 from crewai.cli.add_crew_to_flow import add_crew_to_flow
 from crewai.cli.create_crew import create_crew
 from crewai.cli.create_flow import create_flow
+from crewai.cli.crew_chat import run_chat
 from crewai.memory.storage.kickoff_task_outputs_storage import (
     KickoffTaskOutputsSQLiteStorage,
 )
@@ -25,7 +27,7 @@ from .update_crew import update_crew
 
 
 @click.group()
-@click.version_option(pkg_resources.get_distribution("crewai").version)
+@click.version_option(get_version("crewai"))
 def crewai():
     """Top-level command group for crewai."""
 
@@ -52,16 +54,16 @@ def create(type, name, provider, skip_provider=False):
 def version(tools):
     """Show the installed version of crewai."""
     try:
-        crewai_version = pkg_resources.get_distribution("crewai").version
+        crewai_version = get_version("crewai")
     except Exception:
         crewai_version = "unknown version"
     click.echo(f"crewai version: {crewai_version}")
 
     if tools:
         try:
-            tools_version = pkg_resources.get_distribution("crewai-tools").version
+            tools_version = get_version("crewai")
             click.echo(f"crewai tools version: {tools_version}")
-        except pkg_resources.DistributionNotFound:
+        except Exception:
             click.echo("crewai tools not installed")
 
 
@@ -342,5 +344,15 @@ def flow_add_crew(crew_name):
     add_crew_to_flow(crew_name)
 
 
+@crewai.command()
+def chat():
+    """
+    Start a conversation with the Crew, collecting user-supplied inputs,
+    and using the Chat LLM to generate responses.
+    """
+    click.echo("Starting a conversation with the Crew")
+    run_chat()
+
+
 if __name__ == "__main__":
     crewai()

src/crewai/cli/command.py

@@ -1,8 +1,9 @@
 import requests
 from requests.exceptions import JSONDecodeError
 from rich.console import Console
-from crewai.cli.plus_api import PlusAPI
+
 from crewai.cli.authentication.token import get_auth_token
+from crewai.cli.plus_api import PlusAPI
 from crewai.telemetry.telemetry import Telemetry
 
 console = Console()

src/crewai/cli/config.py

@@ -1,13 +1,19 @@
 import json
 from pathlib import Path
-from pydantic import BaseModel, Field
 from typing import Optional
 
+from pydantic import BaseModel, Field
+
 DEFAULT_CONFIG_PATH = Path.home() / ".config" / "crewai" / "settings.json"
 
+
 class Settings(BaseModel):
-    tool_repository_username: Optional[str] = Field(None, description="Username for interacting with the Tool Repository")
-    tool_repository_password: Optional[str] = Field(None, description="Password for interacting with the Tool Repository")
+    tool_repository_username: Optional[str] = Field(
+        None, description="Username for interacting with the Tool Repository"
+    )
+    tool_repository_password: Optional[str] = Field(
+        None, description="Password for interacting with the Tool Repository"
+    )
     config_path: Path = Field(default=DEFAULT_CONFIG_PATH, exclude=True)
 
     def __init__(self, config_path: Path = DEFAULT_CONFIG_PATH, **data):

src/crewai/cli/constants.py

@@ -17,6 +17,12 @@ ENV_VARS = {
             "key_name": "GEMINI_API_KEY",
         }
     ],
+    "nvidia_nim": [
+        {
+            "prompt": "Enter your NVIDIA API key (press Enter to skip)",
+            "key_name": "NVIDIA_NIM_API_KEY",
+        }
+    ],
     "groq": [
         {
             "prompt": "Enter your GROQ API key (press Enter to skip)",
@@ -85,6 +91,12 @@ ENV_VARS = {
             "key_name": "CEREBRAS_API_KEY",
         },
     ],
+    "sambanova": [
+        {
+            "prompt": "Enter your SambaNovaCloud API key (press Enter to skip)",
+            "key_name": "SAMBANOVA_API_KEY",
+        }
+    ],
 }
 
 
@@ -92,12 +104,14 @@ PROVIDERS = [
     "openai",
     "anthropic",
     "gemini",
+    "nvidia_nim",
     "groq",
     "ollama",
     "watson",
     "bedrock",
     "azure",
     "cerebras",
+    "sambanova",
 ]
 
 MODELS = {
@@ -114,6 +128,75 @@ MODELS = {
         "gemini/gemini-gemma-2-9b-it",
         "gemini/gemini-gemma-2-27b-it",
     ],
+    "nvidia_nim": [
+        "nvidia_nim/nvidia/mistral-nemo-minitron-8b-8k-instruct",
+        "nvidia_nim/nvidia/nemotron-4-mini-hindi-4b-instruct",
+        "nvidia_nim/nvidia/llama-3.1-nemotron-70b-instruct",
+        "nvidia_nim/nvidia/llama3-chatqa-1.5-8b",
+        "nvidia_nim/nvidia/llama3-chatqa-1.5-70b",
+        "nvidia_nim/nvidia/vila",
+        "nvidia_nim/nvidia/neva-22",
+        "nvidia_nim/nvidia/nemotron-mini-4b-instruct",
+        "nvidia_nim/nvidia/usdcode-llama3-70b-instruct",
+        "nvidia_nim/nvidia/nemotron-4-340b-instruct",
+        "nvidia_nim/meta/codellama-70b",
+        "nvidia_nim/meta/llama2-70b",
+        "nvidia_nim/meta/llama3-8b-instruct",
+        "nvidia_nim/meta/llama3-70b-instruct",
+        "nvidia_nim/meta/llama-3.1-8b-instruct",
+        "nvidia_nim/meta/llama-3.1-70b-instruct",
+        "nvidia_nim/meta/llama-3.1-405b-instruct",
+        "nvidia_nim/meta/llama-3.2-1b-instruct",
+        "nvidia_nim/meta/llama-3.2-3b-instruct",
+        "nvidia_nim/meta/llama-3.2-11b-vision-instruct",
+        "nvidia_nim/meta/llama-3.2-90b-vision-instruct",
+        "nvidia_nim/meta/llama-3.1-70b-instruct",
+        "nvidia_nim/google/gemma-7b",
+        "nvidia_nim/google/gemma-2b",
+        "nvidia_nim/google/codegemma-7b",
+        "nvidia_nim/google/codegemma-1.1-7b",
+        "nvidia_nim/google/recurrentgemma-2b",
+        "nvidia_nim/google/gemma-2-9b-it",
+        "nvidia_nim/google/gemma-2-27b-it",
+        "nvidia_nim/google/gemma-2-2b-it",
+        "nvidia_nim/google/deplot",
+        "nvidia_nim/google/paligemma",
+        "nvidia_nim/mistralai/mistral-7b-instruct-v0.2",
+        "nvidia_nim/mistralai/mixtral-8x7b-instruct-v0.1",
+        "nvidia_nim/mistralai/mistral-large",
+        "nvidia_nim/mistralai/mixtral-8x22b-instruct-v0.1",
+        "nvidia_nim/mistralai/mistral-7b-instruct-v0.3",
+        "nvidia_nim/nv-mistralai/mistral-nemo-12b-instruct",
+        "nvidia_nim/mistralai/mamba-codestral-7b-v0.1",
+        "nvidia_nim/microsoft/phi-3-mini-128k-instruct",
+        "nvidia_nim/microsoft/phi-3-mini-4k-instruct",
+        "nvidia_nim/microsoft/phi-3-small-8k-instruct",
+        "nvidia_nim/microsoft/phi-3-small-128k-instruct",
+        "nvidia_nim/microsoft/phi-3-medium-4k-instruct",
+        "nvidia_nim/microsoft/phi-3-medium-128k-instruct",
+        "nvidia_nim/microsoft/phi-3.5-mini-instruct",
+        "nvidia_nim/microsoft/phi-3.5-moe-instruct",
+        "nvidia_nim/microsoft/kosmos-2",
+        "nvidia_nim/microsoft/phi-3-vision-128k-instruct",
+        "nvidia_nim/microsoft/phi-3.5-vision-instruct",
+        "nvidia_nim/databricks/dbrx-instruct",
+        "nvidia_nim/snowflake/arctic",
+        "nvidia_nim/aisingapore/sea-lion-7b-instruct",
+        "nvidia_nim/ibm/granite-8b-code-instruct",
+        "nvidia_nim/ibm/granite-34b-code-instruct",
+        "nvidia_nim/ibm/granite-3.0-8b-instruct",
+        "nvidia_nim/ibm/granite-3.0-3b-a800m-instruct",
+        "nvidia_nim/mediatek/breeze-7b-instruct",
+        "nvidia_nim/upstage/solar-10.7b-instruct",
+        "nvidia_nim/writer/palmyra-med-70b-32k",
+        "nvidia_nim/writer/palmyra-med-70b",
+        "nvidia_nim/writer/palmyra-fin-70b-32k",
+        "nvidia_nim/01-ai/yi-large",
+        "nvidia_nim/deepseek-ai/deepseek-coder-6.7b-instruct",
+        "nvidia_nim/rakuten/rakutenai-7b-instruct",
+        "nvidia_nim/rakuten/rakutenai-7b-chat",
+        "nvidia_nim/baichuan-inc/baichuan2-13b-chat",
+    ],
     "groq": [
         "groq/llama-3.1-8b-instant",
         "groq/llama-3.1-70b-versatile",
@@ -156,8 +239,23 @@ MODELS = {
         "bedrock/mistral.mistral-7b-instruct-v0:2",
         "bedrock/mistral.mixtral-8x7b-instruct-v0:1",
     ],
+    "sambanova": [
+        "sambanova/Meta-Llama-3.3-70B-Instruct",
+        "sambanova/QwQ-32B-Preview",
+        "sambanova/Qwen2.5-72B-Instruct",
+        "sambanova/Qwen2.5-Coder-32B-Instruct",
+        "sambanova/Meta-Llama-3.1-405B-Instruct",
+        "sambanova/Meta-Llama-3.1-70B-Instruct",
+        "sambanova/Meta-Llama-3.1-8B-Instruct",
+        "sambanova/Llama-3.2-90B-Vision-Instruct",
+        "sambanova/Llama-3.2-11B-Vision-Instruct",
+        "sambanova/Meta-Llama-3.2-3B-Instruct",
+        "sambanova/Meta-Llama-3.2-1B-Instruct",
+    ],
 }
 
+DEFAULT_LLM_MODEL = "gpt-4o-mini"
+
 JSON_URL = "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
 
 

src/crewai/cli/crew_chat.py

@@ -0,0 +1,413 @@
+import json
+import re
+import sys
+from pathlib import Path
+from typing import Any, Dict, List, Optional, Set, Tuple
+
+import click
+import tomli
+
+from crewai.crew import Crew
+from crewai.llm import LLM
+from crewai.types.crew_chat import ChatInputField, ChatInputs
+from crewai.utilities.llm_utils import create_llm
+
+
+def run_chat():
+    """
+    Runs an interactive chat loop using the Crew's chat LLM with function calling.
+    Incorporates crew_name, crew_description, and input fields to build a tool schema.
+    Exits if crew_name or crew_description are missing.
+    """
+    crew, crew_name = load_crew_and_name()
+    chat_llm = initialize_chat_llm(crew)
+    if not chat_llm:
+        return
+
+    crew_chat_inputs = generate_crew_chat_inputs(crew, crew_name, chat_llm)
+    crew_tool_schema = generate_crew_tool_schema(crew_chat_inputs)
+    system_message = build_system_message(crew_chat_inputs)
+
+    # Call the LLM to generate the introductory message
+    introductory_message = chat_llm.call(
+        messages=[{"role": "system", "content": system_message}]
+    )
+    click.secho(f"\nAssistant: {introductory_message}\n", fg="green")
+
+    messages = [
+        {"role": "system", "content": system_message},
+        {"role": "assistant", "content": introductory_message},
+    ]
+
+    available_functions = {
+        crew_chat_inputs.crew_name: create_tool_function(crew, messages),
+    }
+
+    click.secho(
+        "\nEntering an interactive chat loop with function-calling.\n"
+        "Type 'exit' or Ctrl+C to quit.\n",
+        fg="cyan",
+    )
+
+    chat_loop(chat_llm, messages, crew_tool_schema, available_functions)
+
+
+def initialize_chat_llm(crew: Crew) -> Optional[LLM]:
+    """Initializes the chat LLM and handles exceptions."""
+    try:
+        return create_llm(crew.chat_llm)
+    except Exception as e:
+        click.secho(
+            f"Unable to find a Chat LLM. Please make sure you set chat_llm on the crew: {e}",
+            fg="red",
+        )
+        return None
+
+
+def build_system_message(crew_chat_inputs: ChatInputs) -> str:
+    """Builds the initial system message for the chat."""
+    required_fields_str = (
+        ", ".join(
+            f"{field.name} (desc: {field.description or 'n/a'})"
+            for field in crew_chat_inputs.inputs
+        )
+        or "(No required fields detected)"
+    )
+
+    return (
+        "You are a helpful AI assistant for the CrewAI platform. "
+        "Your primary purpose is to assist users with the crew's specific tasks. "
+        "You can answer general questions, but should guide users back to the crew's purpose afterward. "
+        "For example, after answering a general question, remind the user of your main purpose, such as generating a research report, and prompt them to specify a topic or task related to the crew's purpose. "
+        "You have a function (tool) you can call by name if you have all required inputs. "
+        f"Those required inputs are: {required_fields_str}. "
+        "Once you have them, call the function. "
+        "Please keep your responses concise and friendly. "
+        "If a user asks a question outside the crew's scope, provide a brief answer and remind them of the crew's purpose. "
+        "After calling the tool, be prepared to take user feedback and make adjustments as needed. "
+        "If you are ever unsure about a user's request or need clarification, ask the user for more information."
+        "Before doing anything else, introduce yourself with a friendly message like: 'Hey! I'm here to help you with [crew's purpose]. Could you please provide me with [inputs] so we can get started?' "
+        "For example: 'Hey! I'm here to help you with uncovering and reporting cutting-edge developments through thorough research and detailed analysis. Could you please provide me with a topic you're interested in? This will help us generate a comprehensive research report and detailed analysis.'"
+        f"\nCrew Name: {crew_chat_inputs.crew_name}"
+        f"\nCrew Description: {crew_chat_inputs.crew_description}"
+    )
+
+
+def create_tool_function(crew: Crew, messages: List[Dict[str, str]]) -> Any:
+    """Creates a wrapper function for running the crew tool with messages."""
+
+    def run_crew_tool_with_messages(**kwargs):
+        return run_crew_tool(crew, messages, **kwargs)
+
+    return run_crew_tool_with_messages
+
+
+def chat_loop(chat_llm, messages, crew_tool_schema, available_functions):
+    """Main chat loop for interacting with the user."""
+    while True:
+        try:
+            user_input = click.prompt("You", type=str)
+            if user_input.strip().lower() in ["exit", "quit"]:
+                click.echo("Exiting chat. Goodbye!")
+                break
+
+            messages.append({"role": "user", "content": user_input})
+            final_response = chat_llm.call(
+                messages=messages,
+                tools=[crew_tool_schema],
+                available_functions=available_functions,
+            )
+
+            messages.append({"role": "assistant", "content": final_response})
+            click.secho(f"\nAssistant: {final_response}\n", fg="green")
+
+        except KeyboardInterrupt:
+            click.echo("\nExiting chat. Goodbye!")
+            break
+        except Exception as e:
+            click.secho(f"An error occurred: {e}", fg="red")
+            break
+
+
+def generate_crew_tool_schema(crew_inputs: ChatInputs) -> dict:
+    """
+    Dynamically build a Littellm 'function' schema for the given crew.
+
+    crew_name: The name of the crew (used for the function 'name').
+    crew_inputs: A ChatInputs object containing crew_description
+                 and a list of input fields (each with a name & description).
+    """
+    properties = {}
+    for field in crew_inputs.inputs:
+        properties[field.name] = {
+            "type": "string",
+            "description": field.description or "No description provided",
+        }
+
+    required_fields = [field.name for field in crew_inputs.inputs]
+
+    return {
+        "type": "function",
+        "function": {
+            "name": crew_inputs.crew_name,
+            "description": crew_inputs.crew_description or "No crew description",
+            "parameters": {
+                "type": "object",
+                "properties": properties,
+                "required": required_fields,
+            },
+        },
+    }
+
+
+def run_crew_tool(crew: Crew, messages: List[Dict[str, str]], **kwargs):
+    """
+    Runs the crew using crew.kickoff(inputs=kwargs) and returns the output.
+
+    Args:
+        crew (Crew): The crew instance to run.
+        messages (List[Dict[str, str]]): The chat messages up to this point.
+        **kwargs: The inputs collected from the user.
+
+    Returns:
+        str: The output from the crew's execution.
+
+    Raises:
+        SystemExit: Exits the chat if an error occurs during crew execution.
+    """
+    try:
+        # Serialize 'messages' to JSON string before adding to kwargs
+        kwargs["crew_chat_messages"] = json.dumps(messages)
+
+        # Run the crew with the provided inputs
+        crew_output = crew.kickoff(inputs=kwargs)
+
+        # Convert CrewOutput to a string to send back to the user
+        result = str(crew_output)
+
+        return result
+    except Exception as e:
+        # Exit the chat and show the error message
+        click.secho("An error occurred while running the crew:", fg="red")
+        click.secho(str(e), fg="red")
+        sys.exit(1)
+
+
+def load_crew_and_name() -> Tuple[Crew, str]:
+    """
+    Loads the crew by importing the crew class from the user's project.
+
+    Returns:
+        Tuple[Crew, str]: A tuple containing the Crew instance and the name of the crew.
+    """
+    # Get the current working directory
+    cwd = Path.cwd()
+
+    # Path to the pyproject.toml file
+    pyproject_path = cwd / "pyproject.toml"
+    if not pyproject_path.exists():
+        raise FileNotFoundError("pyproject.toml not found in the current directory.")
+
+    # Load the pyproject.toml file using 'tomli'
+    with pyproject_path.open("rb") as f:
+        pyproject_data = tomli.load(f)
+
+    # Get the project name from the 'project' section
+    project_name = pyproject_data["project"]["name"]
+    folder_name = project_name
+
+    # Derive the crew class name from the project name
+    # E.g., if project_name is 'my_project', crew_class_name is 'MyProject'
+    crew_class_name = project_name.replace("_", " ").title().replace(" ", "")
+
+    # Add the 'src' directory to sys.path
+    src_path = cwd / "src"
+    if str(src_path) not in sys.path:
+        sys.path.insert(0, str(src_path))
+
+    # Import the crew module
+    crew_module_name = f"{folder_name}.crew"
+    try:
+        crew_module = __import__(crew_module_name, fromlist=[crew_class_name])
+    except ImportError as e:
+        raise ImportError(f"Failed to import crew module {crew_module_name}: {e}")
+
+    # Get the crew class from the module
+    try:
+        crew_class = getattr(crew_module, crew_class_name)
+    except AttributeError:
+        raise AttributeError(
+            f"Crew class {crew_class_name} not found in module {crew_module_name}"
+        )
+
+    # Instantiate the crew
+    crew_instance = crew_class().crew()
+    return crew_instance, crew_class_name
+
+
+def generate_crew_chat_inputs(crew: Crew, crew_name: str, chat_llm) -> ChatInputs:
+    """
+    Generates the ChatInputs required for the crew by analyzing the tasks and agents.
+
+    Args:
+        crew (Crew): The crew object containing tasks and agents.
+        crew_name (str): The name of the crew.
+        chat_llm: The chat language model to use for AI calls.
+
+    Returns:
+        ChatInputs: An object containing the crew's name, description, and input fields.
+    """
+    # Extract placeholders from tasks and agents
+    required_inputs = fetch_required_inputs(crew)
+
+    # Generate descriptions for each input using AI
+    input_fields = []
+    for input_name in required_inputs:
+        description = generate_input_description_with_ai(input_name, crew, chat_llm)
+        input_fields.append(ChatInputField(name=input_name, description=description))
+
+    # Generate crew description using AI
+    crew_description = generate_crew_description_with_ai(crew, chat_llm)
+
+    return ChatInputs(
+        crew_name=crew_name, crew_description=crew_description, inputs=input_fields
+    )
+
+
+def fetch_required_inputs(crew: Crew) -> Set[str]:
+    """
+    Extracts placeholders from the crew's tasks and agents.
+
+    Args:
+        crew (Crew): The crew object.
+
+    Returns:
+        Set[str]: A set of placeholder names.
+    """
+    placeholder_pattern = re.compile(r"\{(.+?)\}")
+    required_inputs: Set[str] = set()
+
+    # Scan tasks
+    for task in crew.tasks:
+        text = f"{task.description or ''} {task.expected_output or ''}"
+        required_inputs.update(placeholder_pattern.findall(text))
+
+    # Scan agents
+    for agent in crew.agents:
+        text = f"{agent.role or ''} {agent.goal or ''} {agent.backstory or ''}"
+        required_inputs.update(placeholder_pattern.findall(text))
+
+    return required_inputs
+
+
+def generate_input_description_with_ai(input_name: str, crew: Crew, chat_llm) -> str:
+    """
+    Generates an input description using AI based on the context of the crew.
+
+    Args:
+        input_name (str): The name of the input placeholder.
+        crew (Crew): The crew object.
+        chat_llm: The chat language model to use for AI calls.
+
+    Returns:
+        str: A concise description of the input.
+    """
+    # Gather context from tasks and agents where the input is used
+    context_texts = []
+    placeholder_pattern = re.compile(r"\{(.+?)\}")
+
+    for task in crew.tasks:
+        if (
+            f"{{{input_name}}}" in task.description
+            or f"{{{input_name}}}" in task.expected_output
+        ):
+            # Replace placeholders with input names
+            task_description = placeholder_pattern.sub(
+                lambda m: m.group(1), task.description
+            )
+            expected_output = placeholder_pattern.sub(
+                lambda m: m.group(1), task.expected_output
+            )
+            context_texts.append(f"Task Description: {task_description}")
+            context_texts.append(f"Expected Output: {expected_output}")
+    for agent in crew.agents:
+        if (
+            f"{{{input_name}}}" in agent.role
+            or f"{{{input_name}}}" in agent.goal
+            or f"{{{input_name}}}" in agent.backstory
+        ):
+            # Replace placeholders with input names
+            agent_role = placeholder_pattern.sub(lambda m: m.group(1), agent.role)
+            agent_goal = placeholder_pattern.sub(lambda m: m.group(1), agent.goal)
+            agent_backstory = placeholder_pattern.sub(
+                lambda m: m.group(1), agent.backstory
+            )
+            context_texts.append(f"Agent Role: {agent_role}")
+            context_texts.append(f"Agent Goal: {agent_goal}")
+            context_texts.append(f"Agent Backstory: {agent_backstory}")
+
+    context = "\n".join(context_texts)
+    if not context:
+        # If no context is found for the input, raise an exception as per instruction
+        raise ValueError(f"No context found for input '{input_name}'.")
+
+    prompt = (
+        f"Based on the following context, write a concise description (15 words or less) of the input '{input_name}'.\n"
+        "Provide only the description, without any extra text or labels. Do not include placeholders like '{topic}' in the description.\n"
+        "Context:\n"
+        f"{context}"
+    )
+    response = chat_llm.call(messages=[{"role": "user", "content": prompt}])
+    description = response.strip()
+
+    return description
+
+
+def generate_crew_description_with_ai(crew: Crew, chat_llm) -> str:
+    """
+    Generates a brief description of the crew using AI.
+
+    Args:
+        crew (Crew): The crew object.
+        chat_llm: The chat language model to use for AI calls.
+
+    Returns:
+        str: A concise description of the crew's purpose (15 words or less).
+    """
+    # Gather context from tasks and agents
+    context_texts = []
+    placeholder_pattern = re.compile(r"\{(.+?)\}")
+
+    for task in crew.tasks:
+        # Replace placeholders with input names
+        task_description = placeholder_pattern.sub(
+            lambda m: m.group(1), task.description
+        )
+        expected_output = placeholder_pattern.sub(
+            lambda m: m.group(1), task.expected_output
+        )
+        context_texts.append(f"Task Description: {task_description}")
+        context_texts.append(f"Expected Output: {expected_output}")
+    for agent in crew.agents:
+        # Replace placeholders with input names
+        agent_role = placeholder_pattern.sub(lambda m: m.group(1), agent.role)
+        agent_goal = placeholder_pattern.sub(lambda m: m.group(1), agent.goal)
+        agent_backstory = placeholder_pattern.sub(lambda m: m.group(1), agent.backstory)
+        context_texts.append(f"Agent Role: {agent_role}")
+        context_texts.append(f"Agent Goal: {agent_goal}")
+        context_texts.append(f"Agent Backstory: {agent_backstory}")
+
+    context = "\n".join(context_texts)
+    if not context:
+        raise ValueError("No context found for generating crew description.")
+
+    prompt = (
+        "Based on the following context, write a concise, action-oriented description (15 words or less) of the crew's purpose.\n"
+        "Provide only the description, without any extra text or labels. Do not include placeholders like '{topic}' in the description.\n"
+        "Context:\n"
+        f"{context}"
+    )
+    response = chat_llm.call(messages=[{"role": "user", "content": prompt}])
+    crew_description = response.strip()
+
+    return crew_description

src/crewai/cli/plus_api.py

@@ -1,9 +1,11 @@
-from typing import Optional
-import requests
 from os import getenv
-from crewai.cli.version import get_crewai_version
+from typing import Optional
 from urllib.parse import urljoin
 
+import requests
+
+from crewai.cli.version import get_crewai_version
+
 
 class PlusAPI:
     """

src/crewai/cli/reset_memories_command.py

@@ -1,11 +1,12 @@
 import subprocess
+
 import click
 
+from crewai.knowledge.storage.knowledge_storage import KnowledgeStorage
 from crewai.memory.entity.entity_memory import EntityMemory
 from crewai.memory.long_term.long_term_memory import LongTermMemory
 from crewai.memory.short_term.short_term_memory import ShortTermMemory
 from crewai.utilities.task_output_storage_handler import TaskOutputStorageHandler
-from crewai.knowledge.storage.knowledge_storage import KnowledgeStorage
 
 
 def reset_memories_command(

src/crewai/cli/templates/crew/README.md

@@ -4,7 +4,7 @@ Welcome to the {{crew_name}} Crew project, powered by [crewAI](https://crewai.co
 
 ## Installation
 
-Ensure you have Python >=3.10 <=3.13 installed on your system. This project uses [UV](https://docs.astral.sh/uv/) for dependency management and package handling, offering a seamless setup and execution experience.
+Ensure you have Python >=3.10 <3.13 installed on your system. This project uses [UV](https://docs.astral.sh/uv/) for dependency management and package handling, offering a seamless setup and execution experience.
 
 First, if you haven't already, install uv:
 

src/crewai/cli/templates/crew/config/tasks.yaml

@@ -2,7 +2,7 @@ research_task:
   description: >
     Conduct a thorough research about {topic}
     Make sure you find any interesting and relevant information given
-    the current year is 2024.
+    the current year is {current_year}.
   expected_output: >
     A list with 10 bullet points of the most relevant information about {topic}
   agent: researcher
@@ -12,6 +12,6 @@ reporting_task:
     Review the context you got and expand each topic into a full section for a report.
     Make sure the report is detailed and contains any and all relevant information.
   expected_output: >
-    A fully fledge reports with the mains topics, each with a full section of information.
+    A fully fledged report with the main topics, each with a full section of information.
     Formatted as markdown without '```'
   agent: reporting_analyst

src/crewai/cli/templates/crew/main.py

@@ -2,6 +2,8 @@
 import sys
 import warnings
 
+from datetime import datetime
+
 from {{folder_name}}.crew import {{crew_name}}
 
 warnings.filterwarnings("ignore", category=SyntaxWarning, module="pysbd")
@@ -16,9 +18,14 @@ def run():
     Run the crew.
     """
     inputs = {
-        'topic': 'AI LLMs'
+        'topic': 'AI LLMs',
+        'current_year': str(datetime.now().year)
     }
-    {{crew_name}}().crew().kickoff(inputs=inputs)
+    
+    try:
+        {{crew_name}}().crew().kickoff(inputs=inputs)
+    except Exception as e:
+        raise Exception(f"An error occurred while running the crew: {e}")
 
 
 def train():
@@ -55,4 +62,4 @@ def test():
         {{crew_name}}().crew().test(n_iterations=int(sys.argv[1]), openai_model_name=sys.argv[2], inputs=inputs)
 
     except Exception as e:
-        raise Exception(f"An error occurred while replaying the crew: {e}")
+        raise Exception(f"An error occurred while testing the crew: {e}")

src/crewai/cli/templates/crew/pyproject.toml

@@ -3,9 +3,9 @@ name = "{{folder_name}}"
 version = "0.1.0"
 description = "{{name}} using crewAI"
 authors = [{ name = "Your Name", email = "you@example.com" }]
-requires-python = ">=3.10,<=3.13"
+requires-python = ">=3.10,<3.13"
 dependencies = [
-    "crewai[tools]>=0.86.0,<1.0.0"
+    "crewai[tools]>=0.95.0,<1.0.0"
 ]
 
 [project.scripts]
@@ -18,3 +18,6 @@ test = "{{folder_name}}.main:test"
 [build-system]
 requires = ["hatchling"]
 build-backend = "hatchling.build"
+
+[tool.crewai]
+type = "crew"

src/crewai/cli/templates/crew/tools/custom_tool.py

@@ -10,7 +10,7 @@ class MyCustomToolInput(BaseModel):
 class MyCustomTool(BaseTool):
     name: str = "Name of my tool"
     description: str = (
-        "Clear description for what this tool is useful for, you agent will need this information to use it."
+        "Clear description for what this tool is useful for, your agent will need this information to use it."
     )
     args_schema: Type[BaseModel] = MyCustomToolInput
 

src/crewai/cli/templates/flow/README.md

@@ -4,7 +4,7 @@ Welcome to the {{crew_name}} Crew project, powered by [crewAI](https://crewai.co
 
 ## Installation
 
-Ensure you have Python >=3.10 <=3.13 installed on your system. This project uses [UV](https://docs.astral.sh/uv/) for dependency management and package handling, offering a seamless setup and execution experience.
+Ensure you have Python >=3.10 <3.13 installed on your system. This project uses [UV](https://docs.astral.sh/uv/) for dependency management and package handling, offering a seamless setup and execution experience.
 
 First, if you haven't already, install uv:
 

src/crewai/cli/templates/flow/main.py

@@ -5,7 +5,7 @@ from pydantic import BaseModel
 
 from crewai.flow.flow import Flow, listen, start
 
-from .crews.poem_crew.poem_crew import PoemCrew
+from {{folder_name}}.crews.poem_crew.poem_crew import PoemCrew
 
 
 class PoemState(BaseModel):

src/crewai/cli/templates/flow/pyproject.toml

@@ -3,9 +3,9 @@ name = "{{folder_name}}"
 version = "0.1.0"
 description = "{{name}} using crewAI"
 authors = [{ name = "Your Name", email = "you@example.com" }]
-requires-python = ">=3.10,<=3.13"
+requires-python = ">=3.10,<3.13"
 dependencies = [
-    "crewai[tools]>=0.86.0,<1.0.0",
+    "crewai[tools]>=0.95.0,<1.0.0",
 ]
 
 [project.scripts]
@@ -15,3 +15,6 @@ plot = "{{folder_name}}.main:plot"
 [build-system]
 requires = ["hatchling"]
 build-backend = "hatchling.build"
+
+[tool.crewai]
+type = "flow"

src/crewai/cli/templates/flow/tools/custom_tool.py

@@ -13,7 +13,7 @@ class MyCustomToolInput(BaseModel):
 class MyCustomTool(BaseTool):
     name: str = "Name of my tool"
     description: str = (
-        "Clear description for what this tool is useful for, you agent will need this information to use it."
+        "Clear description for what this tool is useful for, your agent will need this information to use it."
     )
     args_schema: Type[BaseModel] = MyCustomToolInput
 

src/crewai/cli/templates/pipeline/pyproject.toml

@@ -1,17 +0,0 @@
-[tool.poetry]
-name = "{{folder_name}}"
-version = "0.1.0"
-description = "{{name}} using crewAI"
-authors = ["Your Name <you@example.com>"]
-
-[tool.poetry.dependencies]
-python = ">=3.10,<=3.13"
-crewai = { extras = ["tools"], version = ">=0.86.0,<1.0.0" }
-asyncio = "*"
-
-[tool.poetry.scripts]
-{{folder_name}} = "{{folder_name}}.main:main"
-
-[build-system]
-requires = ["poetry-core"]
-build-backend = "poetry.core.masonry.api"

src/crewai/cli/templates/tool/README.md

@@ -5,7 +5,7 @@ custom tools to power up your crews.
 
 ## Installing
 
-Ensure you have Python >=3.10 <=3.13 installed on your system. This project
+Ensure you have Python >=3.10 <3.13 installed on your system. This project
 uses [UV](https://docs.astral.sh/uv/) for dependency management and package
 handling, offering a seamless setup and execution experience.
 

src/crewai/cli/templates/tool/pyproject.toml

@@ -3,8 +3,10 @@ name = "{{folder_name}}"
 version = "0.1.0"
 description = "Power up your crews with {{folder_name}}"
 readme = "README.md"
-requires-python = ">=3.10,<=3.13"
+requires-python = ">=3.10,<3.13"
 dependencies = [
-    "crewai[tools]>=0.86.0"
+    "crewai[tools]>=0.95.0"
 ]
 
+[tool.crewai]
+type = "tool"

src/crewai/cli/tools/main.py

@@ -117,7 +117,7 @@ class ToolCommand(BaseCommand, PlusAPIMixin):
 
         published_handle = publish_response.json()["handle"]
         console.print(
-            f"Succesfully published {published_handle} ({project_version}).\nInstall it in other projects with crewai tool install {published_handle}",
+            f"Successfully published {published_handle} ({project_version}).\nInstall it in other projects with crewai tool install {published_handle}",
             style="bold green",
         )
 
@@ -138,7 +138,7 @@ class ToolCommand(BaseCommand, PlusAPIMixin):
 
         self._add_package(get_response.json())
 
-        console.print(f"Succesfully installed {handle}", style="bold green")
+        console.print(f"Successfully installed {handle}", style="bold green")
 
     def login(self):
         login_response = self.plus_api_client.login_to_tool_repository()

src/crewai/cli/utils.py

@@ -33,26 +33,6 @@ def copy_template(src, dst, name, class_name, folder_name):
     click.secho(f"  - Created {dst}", fg="green")
 
 
-# Drop the simple_toml_parser when we move to python3.11
-def simple_toml_parser(content):
-    result = {}
-    current_section = result
-    for line in content.split("\n"):
-        line = line.strip()
-        if line.startswith("[") and line.endswith("]"):
-            # New section
-            section = line[1:-1].split(".")
-            current_section = result
-            for key in section:
-                current_section = current_section.setdefault(key, {})
-        elif "=" in line:
-            key, value = line.split("=", 1)
-            key = key.strip()
-            value = value.strip().strip('"')
-            current_section[key] = value
-    return result
-
-
 def read_toml(file_path: str = "pyproject.toml"):
     """Read the content of a TOML file and return it as a dictionary."""
     with open(file_path, "rb") as f:
@@ -63,7 +43,7 @@ def read_toml(file_path: str = "pyproject.toml"):
 def parse_toml(content):
     if sys.version_info >= (3, 11):
         return tomllib.loads(content)
-    return simple_toml_parser(content)
+    return tomli.loads(content)
 
 
 def get_project_name(

src/crewai/cli/version.py

@@ -1,6 +1,6 @@
 import importlib.metadata
 
+
 def get_crewai_version() -> str:
     """Get the version number of CrewAI running the CLI"""
     return importlib.metadata.version("crewai")
-

src/crewai/crew.py

@@ -1,11 +1,11 @@
 import asyncio
 import json
-import os
+import re
 import uuid
 import warnings
 from concurrent.futures import Future
 from hashlib import md5
-from typing import Any, Callable, Dict, List, Optional, Tuple, Union
+from typing import Any, Callable, Dict, List, Optional, Set, Tuple, Union
 
 from pydantic import (
     UUID4,
@@ -36,6 +36,8 @@ from crewai.tasks.conditional_task import ConditionalTask
 from crewai.tasks.task_output import TaskOutput
 from crewai.telemetry import Telemetry
 from crewai.tools.agent_tools.agent_tools import AgentTools
+from crewai.tools.base_tool import Tool
+from crewai.types.crew_chat import ChatInputs
 from crewai.types.usage_metrics import UsageMetrics
 from crewai.utilities import I18N, FileHandler, Logger, RPMController
 from crewai.utilities.constants import TRAINING_DATA_FILE
@@ -45,16 +47,15 @@ from crewai.utilities.formatter import (
     aggregate_raw_outputs_from_task_outputs,
     aggregate_raw_outputs_from_tasks,
 )
+from crewai.utilities.llm_utils import create_llm
 from crewai.utilities.planning_handler import CrewPlanner
 from crewai.utilities.task_output_storage_handler import TaskOutputStorageHandler
 from crewai.utilities.training_handler import CrewTrainingHandler
 
-agentops = None
-if os.environ.get("AGENTOPS_API_KEY"):
-    try:
-        import agentops  # type: ignore
-    except ImportError:
-        pass
+try:
+    import agentops  # type: ignore
+except ImportError:
+    agentops = None
 
 
 warnings.filterwarnings("ignore", category=SyntaxWarning, module="pysbd")
@@ -149,7 +150,7 @@ class Crew(BaseModel):
     manager_agent: Optional[BaseAgent] = Field(
         description="Custom agent that will be used as manager.", default=None
     )
-    function_calling_llm: Optional[Any] = Field(
+    function_calling_llm: Optional[Union[str, InstanceOf[LLM], Any]] = Field(
         description="Language model that will run the agent.", default=None
     )
     config: Optional[Union[Json, Dict[str, Any]]] = Field(default=None)
@@ -205,6 +206,10 @@ class Crew(BaseModel):
         default=None,
         description="Knowledge sources for the crew. Add knowledge sources to the knowledge object.",
     )
+    chat_llm: Optional[Any] = Field(
+        default=None,
+        description="LLM used to handle chatting with the crew.",
+    )
     _knowledge: Optional[Knowledge] = PrivateAttr(
         default=None,
     )
@@ -241,15 +246,9 @@ class Crew(BaseModel):
         if self.output_log_file:
             self._file_handler = FileHandler(self.output_log_file)
         self._rpm_controller = RPMController(max_rpm=self.max_rpm, logger=self._logger)
-        if self.function_calling_llm:
-            if isinstance(self.function_calling_llm, str):
-                self.function_calling_llm = LLM(model=self.function_calling_llm)
-            elif not isinstance(self.function_calling_llm, LLM):
-                self.function_calling_llm = LLM(
-                    model=getattr(self.function_calling_llm, "model_name", None)
-                    or getattr(self.function_calling_llm, "deployment_name", None)
-                    or str(self.function_calling_llm)
-                )
+        if self.function_calling_llm and not isinstance(self.function_calling_llm, LLM):
+            self.function_calling_llm = create_llm(self.function_calling_llm)
+
         self._telemetry = Telemetry()
         self._telemetry.set_tracer()
         return self
@@ -514,6 +513,8 @@ class Crew(BaseModel):
         inputs: Optional[Dict[str, Any]] = None,
     ) -> CrewOutput:
         for before_callback in self.before_kickoff_callbacks:
+            if inputs is None:
+                inputs = {}
             inputs = before_callback(inputs)
 
         """Starts the crew to work on its assigned tasks."""
@@ -536,9 +537,6 @@ class Crew(BaseModel):
             if not agent.function_calling_llm:  # type: ignore # "BaseAgent" has no attribute "function_calling_llm"
                 agent.function_calling_llm = self.function_calling_llm  # type: ignore # "BaseAgent" has no attribute "function_calling_llm"
 
-            if agent.allow_code_execution:  # type: ignore # BaseAgent" has no attribute "allow_code_execution"
-                agent.tools += agent.get_code_execution_tools()  # type: ignore # "BaseAgent" has no attribute "get_code_execution_tools"; maybe "get_delegation_tools"?
-
             if not agent.step_callback:  # type: ignore # "BaseAgent" has no attribute "step_callback"
                 agent.step_callback = self.step_callback  # type: ignore # "BaseAgent" has no attribute "step_callback"
 
@@ -675,10 +673,10 @@ class Crew(BaseModel):
                 )
                 manager.tools = []
                 raise Exception("Manager agent should not have tools")
-            manager.tools = self.manager_agent.get_delegation_tools(self.agents)
         else:
             self.manager_llm = (
                 getattr(self.manager_llm, "model_name", None)
+                or getattr(self.manager_llm, "model", None)
                 or getattr(self.manager_llm, "deployment_name", None)
                 or self.manager_llm
             )
@@ -687,6 +685,7 @@ class Crew(BaseModel):
                 goal=i18n.retrieve("hierarchical_manager_agent", "goal"),
                 backstory=i18n.retrieve("hierarchical_manager_agent", "backstory"),
                 tools=AgentTools(agents=self.agents).tools(),
+                allow_delegation=True,
                 llm=self.manager_llm,
                 verbose=self.verbose,
             )
@@ -729,7 +728,10 @@ class Crew(BaseModel):
                     f"No agent available for task: {task.description}. Ensure that either the task has an assigned agent or a manager agent is provided."
                 )
 
-            self._prepare_agent_tools(task)
+            # Determine which tools to use - task tools take precedence over agent tools
+            tools_for_task = task.tools or agent_to_use.tools or []
+            tools_for_task = self._prepare_tools(agent_to_use, task, tools_for_task)
+
             self._log_task_start(task, agent_to_use.role)
 
             if isinstance(task, ConditionalTask):
@@ -746,7 +748,7 @@ class Crew(BaseModel):
                 future = task.execute_async(
                     agent=agent_to_use,
                     context=context,
-                    tools=agent_to_use.tools,
+                    tools=tools_for_task,
                 )
                 futures.append((task, future, task_index))
             else:
@@ -758,7 +760,7 @@ class Crew(BaseModel):
                 task_output = task.execute_sync(
                     agent=agent_to_use,
                     context=context,
-                    tools=agent_to_use.tools,
+                    tools=tools_for_task,
                 )
                 task_outputs = [task_output]
                 self._process_task_result(task, task_output)
@@ -795,45 +797,77 @@ class Crew(BaseModel):
             return skipped_task_output
         return None
 
-    def _prepare_agent_tools(self, task: Task):
-        if self.process == Process.hierarchical:
-            if self.manager_agent:
-                self._update_manager_tools(task)
-            else:
-                raise ValueError("Manager agent is required for hierarchical process.")
-        elif task.agent and task.agent.allow_delegation:
-            self._add_delegation_tools(task)
+    def _prepare_tools(
+        self, agent: BaseAgent, task: Task, tools: List[Tool]
+    ) -> List[Tool]:
+        # Add delegation tools if agent allows delegation
+        if agent.allow_delegation:
+            if self.process == Process.hierarchical:
+                if self.manager_agent:
+                    tools = self._update_manager_tools(task, tools)
+                else:
+                    raise ValueError(
+                        "Manager agent is required for hierarchical process."
+                    )
+
+            elif agent and agent.allow_delegation:
+                tools = self._add_delegation_tools(task, tools)
+
+        # Add code execution tools if agent allows code execution
+        if agent.allow_code_execution:
+            tools = self._add_code_execution_tools(agent, tools)
+
+        if agent and agent.multimodal:
+            tools = self._add_multimodal_tools(agent, tools)
+
+        return tools
 
     def _get_agent_to_use(self, task: Task) -> Optional[BaseAgent]:
         if self.process == Process.hierarchical:
             return self.manager_agent
         return task.agent
 
-    def _add_delegation_tools(self, task: Task):
+    def _merge_tools(
+        self, existing_tools: List[Tool], new_tools: List[Tool]
+    ) -> List[Tool]:
+        """Merge new tools into existing tools list, avoiding duplicates by tool name."""
+        if not new_tools:
+            return existing_tools
+
+        # Create mapping of tool names to new tools
+        new_tool_map = {tool.name: tool for tool in new_tools}
+
+        # Remove any existing tools that will be replaced
+        tools = [tool for tool in existing_tools if tool.name not in new_tool_map]
+
+        # Add all new tools
+        tools.extend(new_tools)
+
+        return tools
+
+    def _inject_delegation_tools(
+        self, tools: List[Tool], task_agent: BaseAgent, agents: List[BaseAgent]
+    ):
+        delegation_tools = task_agent.get_delegation_tools(agents)
+        return self._merge_tools(tools, delegation_tools)
+
+    def _add_multimodal_tools(self, agent: BaseAgent, tools: List[Tool]):
+        multimodal_tools = agent.get_multimodal_tools()
+        return self._merge_tools(tools, multimodal_tools)
+
+    def _add_code_execution_tools(self, agent: BaseAgent, tools: List[Tool]):
+        code_tools = agent.get_code_execution_tools()
+        return self._merge_tools(tools, code_tools)
+
+    def _add_delegation_tools(self, task: Task, tools: List[Tool]):
         agents_for_delegation = [agent for agent in self.agents if agent != task.agent]
         if len(self.agents) > 1 and len(agents_for_delegation) > 0 and task.agent:
-            delegation_tools = task.agent.get_delegation_tools(agents_for_delegation)
-
-            # Add tools if they are not already in task.tools
-            for new_tool in delegation_tools:
-                # Find the index of the tool with the same name
-                existing_tool_index = next(
-                    (
-                        index
-                        for index, tool in enumerate(task.tools or [])
-                        if tool.name == new_tool.name
-                    ),
-                    None,
-                )
-                if not task.tools:
-                    task.tools = []
-
-                if existing_tool_index is not None:
-                    # Replace the existing tool
-                    task.tools[existing_tool_index] = new_tool
-                else:
-                    # Add the new tool
-                    task.tools.append(new_tool)
+            if not tools:
+                tools = []
+            tools = self._inject_delegation_tools(
+                tools, task.agent, agents_for_delegation
+            )
+        return tools
 
     def _log_task_start(self, task: Task, role: str = "None"):
         if self.output_log_file:
@@ -841,14 +875,15 @@ class Crew(BaseModel):
                 task_name=task.name, task=task.description, agent=role, status="started"
             )
 
-    def _update_manager_tools(self, task: Task):
+    def _update_manager_tools(self, task: Task, tools: List[Tool]):
         if self.manager_agent:
             if task.agent:
-                self.manager_agent.tools = task.agent.get_delegation_tools([task.agent])
+                tools = self._inject_delegation_tools(tools, task.agent, [task.agent])
             else:
-                self.manager_agent.tools = self.manager_agent.get_delegation_tools(
-                    self.agents
+                tools = self._inject_delegation_tools(
+                    tools, self.manager_agent, self.agents
                 )
+        return tools
 
     def _get_context(self, task: Task, task_outputs: List[TaskOutput]):
         context = (
@@ -960,6 +995,31 @@ class Crew(BaseModel):
             return self._knowledge.query(query)
         return None
 
+    def fetch_inputs(self) -> Set[str]:
+        """
+        Gathers placeholders (e.g., {something}) referenced in tasks or agents.
+        Scans each task's 'description' + 'expected_output', and each agent's
+        'role', 'goal', and 'backstory'.
+
+        Returns a set of all discovered placeholder names.
+        """
+        placeholder_pattern = re.compile(r"\{(.+?)\}")
+        required_inputs: Set[str] = set()
+
+        # Scan tasks for inputs
+        for task in self.tasks:
+            # description and expected_output might contain e.g. {topic}, {user_name}, etc.
+            text = f"{task.description or ''} {task.expected_output or ''}"
+            required_inputs.update(placeholder_pattern.findall(text))
+
+        # Scan agents for inputs
+        for agent in self.agents:
+            # role, goal, backstory might have placeholders like {role_detail}, etc.
+            text = f"{agent.role or ''} {agent.goal or ''} {agent.backstory or ''}"
+            required_inputs.update(placeholder_pattern.findall(text))
+
+        return required_inputs
+
     def copy(self):
         """Create a deep copy of the Crew."""
 
@@ -1015,7 +1075,7 @@ class Crew(BaseModel):
     def _interpolate_inputs(self, inputs: Dict[str, Any]) -> None:
         """Interpolates the inputs in the tasks and agents."""
         [
-            task.interpolate_inputs(
+            task.interpolate_inputs_and_add_conversation_history(
                 # type: ignore # "interpolate_inputs" of "Task" does not return a value (it only ever returns None)
                 inputs
             )
@@ -1032,6 +1092,7 @@ class Crew(BaseModel):
             agentops.end_session(
                 end_state="Success",
                 end_state_reason="Finished Execution",
+                is_auto_end=True,
             )
         self._telemetry.end_crew(self, final_string_output)
 

src/crewai/flow/flow.py

@@ -13,17 +13,70 @@ from typing import (
     Union,
     cast,
 )
+from uuid import uuid4
 
-from pydantic import BaseModel, ValidationError
+from blinker import Signal
+from pydantic import BaseModel, Field, ValidationError
 
+from crewai.flow.flow_events import (
+    FlowFinishedEvent,
+    FlowStartedEvent,
+    MethodExecutionFinishedEvent,
+    MethodExecutionStartedEvent,
+)
 from crewai.flow.flow_visualizer import plot_flow
 from crewai.flow.utils import get_possible_return_constants
 from crewai.telemetry import Telemetry
 
-T = TypeVar("T", bound=Union[BaseModel, Dict[str, Any]])
+
+class FlowState(BaseModel):
+    """Base model for all flow states, ensuring each state has a unique ID."""
+    id: str = Field(default_factory=lambda: str(uuid4()), description="Unique identifier for the flow state")
+
+T = TypeVar("T", bound=Union[FlowState, Dict[str, Any]])
 
 
-def start(condition=None):
+def start(condition: Optional[Union[str, dict, Callable]] = None) -> Callable:
+    """
+    Marks a method as a flow's starting point.
+
+    This decorator designates a method as an entry point for the flow execution.
+    It can optionally specify conditions that trigger the start based on other
+    method executions.
+
+    Parameters
+    ----------
+    condition : Optional[Union[str, dict, Callable]], optional
+        Defines when the start method should execute. Can be:
+        - str: Name of a method that triggers this start
+        - dict: Contains "type" ("AND"/"OR") and "methods" (list of triggers)
+        - Callable: A method reference that triggers this start
+        Default is None, meaning unconditional start.
+
+    Returns
+    -------
+    Callable
+        A decorator function that marks the method as a flow start point.
+
+    Raises
+    ------
+    ValueError
+        If the condition format is invalid.
+
+    Examples
+    --------
+    >>> @start()  # Unconditional start
+    >>> def begin_flow(self):
+    ...     pass
+
+    >>> @start("method_name")  # Start after specific method
+    >>> def conditional_start(self):
+    ...     pass
+
+    >>> @start(and_("method1", "method2"))  # Start after multiple methods
+    >>> def complex_start(self):
+    ...     pass
+    """
     def decorator(func):
         func.__is_start_method__ = True
         if condition is not None:
@@ -48,8 +101,42 @@ def start(condition=None):
 
     return decorator
 
+def listen(condition: Union[str, dict, Callable]) -> Callable:
+    """
+    Creates a listener that executes when specified conditions are met.
 
-def listen(condition):
+    This decorator sets up a method to execute in response to other method
+    executions in the flow. It supports both simple and complex triggering
+    conditions.
+
+    Parameters
+    ----------
+    condition : Union[str, dict, Callable]
+        Specifies when the listener should execute. Can be:
+        - str: Name of a method that triggers this listener
+        - dict: Contains "type" ("AND"/"OR") and "methods" (list of triggers)
+        - Callable: A method reference that triggers this listener
+
+    Returns
+    -------
+    Callable
+        A decorator function that sets up the method as a listener.
+
+    Raises
+    ------
+    ValueError
+        If the condition format is invalid.
+
+    Examples
+    --------
+    >>> @listen("process_data")  # Listen to single method
+    >>> def handle_processed_data(self):
+    ...     pass
+
+    >>> @listen(or_("success", "failure"))  # Listen to multiple methods
+    >>> def handle_completion(self):
+    ...     pass
+    """
     def decorator(func):
         if isinstance(condition, str):
             func.__trigger_methods__ = [condition]
@@ -73,16 +160,103 @@ def listen(condition):
     return decorator
 
 
-def router(method):
+def router(condition: Union[str, dict, Callable]) -> Callable:
+    """
+    Creates a routing method that directs flow execution based on conditions.
+
+    This decorator marks a method as a router, which can dynamically determine
+    the next steps in the flow based on its return value. Routers are triggered
+    by specified conditions and can return constants that determine which path
+    the flow should take.
+
+    Parameters
+    ----------
+    condition : Union[str, dict, Callable]
+        Specifies when the router should execute. Can be:
+        - str: Name of a method that triggers this router
+        - dict: Contains "type" ("AND"/"OR") and "methods" (list of triggers)
+        - Callable: A method reference that triggers this router
+
+    Returns
+    -------
+    Callable
+        A decorator function that sets up the method as a router.
+
+    Raises
+    ------
+    ValueError
+        If the condition format is invalid.
+
+    Examples
+    --------
+    >>> @router("check_status")
+    >>> def route_based_on_status(self):
+    ...     if self.state.status == "success":
+    ...         return SUCCESS
+    ...     return FAILURE
+
+    >>> @router(and_("validate", "process"))
+    >>> def complex_routing(self):
+    ...     if all([self.state.valid, self.state.processed]):
+    ...         return CONTINUE
+    ...     return STOP
+    """
     def decorator(func):
         func.__is_router__ = True
-        func.__router_for__ = method.__name__
+        if isinstance(condition, str):
+            func.__trigger_methods__ = [condition]
+            func.__condition_type__ = "OR"
+        elif (
+            isinstance(condition, dict)
+            and "type" in condition
+            and "methods" in condition
+        ):
+            func.__trigger_methods__ = condition["methods"]
+            func.__condition_type__ = condition["type"]
+        elif callable(condition) and hasattr(condition, "__name__"):
+            func.__trigger_methods__ = [condition.__name__]
+            func.__condition_type__ = "OR"
+        else:
+            raise ValueError(
+                "Condition must be a method, string, or a result of or_() or and_()"
+            )
         return func
 
     return decorator
 
+def or_(*conditions: Union[str, dict, Callable]) -> dict:
+    """
+    Combines multiple conditions with OR logic for flow control.
 
-def or_(*conditions):
+    Creates a condition that is satisfied when any of the specified conditions
+    are met. This is used with @start, @listen, or @router decorators to create
+    complex triggering conditions.
+
+    Parameters
+    ----------
+    *conditions : Union[str, dict, Callable]
+        Variable number of conditions that can be:
+        - str: Method names
+        - dict: Existing condition dictionaries
+        - Callable: Method references
+
+    Returns
+    -------
+    dict
+        A condition dictionary with format:
+        {"type": "OR", "methods": list_of_method_names}
+
+    Raises
+    ------
+    ValueError
+        If any condition is invalid.
+
+    Examples
+    --------
+    >>> @listen(or_("success", "timeout"))
+    >>> def handle_completion(self):
+    ...     pass
+    """
     methods = []
     for condition in conditions:
         if isinstance(condition, dict) and "methods" in condition:
@@ -96,7 +270,39 @@ def or_(*conditions):
     return {"type": "OR", "methods": methods}
 
 
-def and_(*conditions):
+def and_(*conditions: Union[str, dict, Callable]) -> dict:
+    """
+    Combines multiple conditions with AND logic for flow control.
+
+    Creates a condition that is satisfied only when all specified conditions
+    are met. This is used with @start, @listen, or @router decorators to create
+    complex triggering conditions.
+
+    Parameters
+    ----------
+    *conditions : Union[str, dict, Callable]
+        Variable number of conditions that can be:
+        - str: Method names
+        - dict: Existing condition dictionaries
+        - Callable: Method references
+
+    Returns
+    -------
+    dict
+        A condition dictionary with format:
+        {"type": "AND", "methods": list_of_method_names}
+
+    Raises
+    ------
+    ValueError
+        If any condition is invalid.
+
+    Examples
+    --------
+    >>> @listen(and_("validated", "processed"))
+    >>> def handle_complete_data(self):
+    ...     pass
+    """
     methods = []
     for condition in conditions:
         if isinstance(condition, dict) and "methods" in condition:
@@ -116,8 +322,8 @@ class FlowMeta(type):
 
         start_methods = []
         listeners = {}
-        routers = {}
         router_paths = {}
+        routers = set()
 
         for attr_name, attr_value in dct.items():
             if hasattr(attr_value, "__is_start_method__"):
@@ -130,18 +336,11 @@ class FlowMeta(type):
                 methods = attr_value.__trigger_methods__
                 condition_type = getattr(attr_value, "__condition_type__", "OR")
                 listeners[attr_name] = (condition_type, methods)
-
-            elif hasattr(attr_value, "__is_router__"):
-                routers[attr_value.__router_for__] = attr_name
-                possible_returns = get_possible_return_constants(attr_value)
-                if possible_returns:
-                    router_paths[attr_name] = possible_returns
-
-                # Register router as a listener to its triggering method
-                trigger_method_name = attr_value.__router_for__
-                methods = [trigger_method_name]
-                condition_type = "OR"
-                listeners[attr_name] = (condition_type, methods)
+                if hasattr(attr_value, "__is_router__") and attr_value.__is_router__:
+                    routers.add(attr_name)
+                    possible_returns = get_possible_return_constants(attr_value)
+                    if possible_returns:
+                        router_paths[attr_name] = possible_returns
 
         setattr(cls, "_start_methods", start_methods)
         setattr(cls, "_listeners", listeners)
@@ -156,9 +355,10 @@ class Flow(Generic[T], metaclass=FlowMeta):
 
     _start_methods: List[str] = []
     _listeners: Dict[str, tuple[str, List[str]]] = {}
-    _routers: Dict[str, str] = {}
+    _routers: Set[str] = set()
     _router_paths: Dict[str, List[str]] = {}
     initial_state: Union[Type[T], T, None] = None
+    event_emitter = Signal("event_emitter")
 
     def __class_getitem__(cls: Type["Flow"], item: Type[T]) -> Type["Flow"]:
         class _FlowGeneric(cls):  # type: ignore
@@ -183,14 +383,37 @@ class Flow(Generic[T], metaclass=FlowMeta):
                 self._methods[method_name] = getattr(self, method_name)
 
     def _create_initial_state(self) -> T:
+        # Handle case where initial_state is None but we have a type parameter
         if self.initial_state is None and hasattr(self, "_initial_state_T"):
-            return self._initial_state_T()  # type: ignore
+            state_type = getattr(self, "_initial_state_T")
+            if isinstance(state_type, type):
+                if issubclass(state_type, FlowState):
+                    return state_type()  # type: ignore
+                elif issubclass(state_type, BaseModel):
+                    # Create a new type that includes the ID field
+                    class StateWithId(state_type, FlowState):  # type: ignore
+                        pass
+                    return StateWithId()  # type: ignore
+
+        # Handle case where no initial state is provided
         if self.initial_state is None:
-            return {}  # type: ignore
-        elif isinstance(self.initial_state, type):
-            return self.initial_state()
-        else:
-            return self.initial_state
+            return {"id": str(uuid4())}  # type: ignore
+
+        # Handle case where initial_state is a type (class)
+        if isinstance(self.initial_state, type):
+            if issubclass(self.initial_state, FlowState):
+                return self.initial_state()  # type: ignore
+            elif issubclass(self.initial_state, BaseModel):
+                # Create a new type that includes the ID field
+                class StateWithId(self.initial_state, FlowState):  # type: ignore
+                    pass
+                return StateWithId()  # type: ignore
+
+        # Handle dictionary case
+        if isinstance(self.initial_state, dict) and "id" not in self.initial_state:
+            self.initial_state["id"] = str(uuid4())
+
+        return self.initial_state  # type: ignore
 
     @property
     def state(self) -> T:
@@ -202,20 +425,17 @@ class Flow(Generic[T], metaclass=FlowMeta):
         return self._method_outputs
 
     def _initialize_state(self, inputs: Dict[str, Any]) -> None:
-        """
-        Initializes or updates the state with the provided inputs.
-
-        Args:
-            inputs: Dictionary of inputs to initialize or update the state.
-
-        Raises:
-            ValueError: If inputs do not match the structured state model.
-            TypeError: If state is neither a BaseModel instance nor a dictionary.
-        """
-        if isinstance(self._state, BaseModel):
-            # Structured state management
+        if isinstance(self._state, dict):
+            # Preserve the ID when updating unstructured state
+            current_id = self._state.get("id")
+            self._state.update(inputs)
+            if current_id:
+                self._state["id"] = current_id
+            elif "id" not in self._state:
+                self._state["id"] = str(uuid4())
+        elif isinstance(self._state, BaseModel):
+            # Structured state
             try:
-                # Define a function to create the dynamic class
                 def create_model_with_extra_forbid(
                     base_model: Type[BaseModel],
                 ) -> Type[BaseModel]:
@@ -225,50 +445,45 @@ class Flow(Generic[T], metaclass=FlowMeta):
 
                     return ModelWithExtraForbid
 
-                # Create the dynamic class
+                # Get current state as dict, preserving the ID if it exists
+                state_model = cast(BaseModel, self._state)
+                current_state = (
+                    state_model.model_dump()
+                    if hasattr(state_model, "model_dump")
+                    else state_model.dict()
+                    if hasattr(state_model, "dict")
+                    else {
+                        k: v
+                        for k, v in state_model.__dict__.items()
+                        if not k.startswith("_")
+                    }
+                )
+
                 ModelWithExtraForbid = create_model_with_extra_forbid(
                     self._state.__class__
                 )
-
-                # Create a new instance using the combined state and inputs
                 self._state = cast(
-                    T, ModelWithExtraForbid(**{**self._state.model_dump(), **inputs})
+                    T, ModelWithExtraForbid(**{**current_state, **inputs})
                 )
-
             except ValidationError as e:
                 raise ValueError(f"Invalid inputs for structured state: {e}") from e
-        elif isinstance(self._state, dict):
-            # Unstructured state management
-            self._state.update(inputs)
         else:
             raise TypeError("State must be a BaseModel instance or a dictionary.")
 
     def kickoff(self, inputs: Optional[Dict[str, Any]] = None) -> Any:
-        """
-        Starts the execution of the flow synchronously.
+        self.event_emitter.send(
+            self,
+            event=FlowStartedEvent(
+                type="flow_started",
+                flow_name=self.__class__.__name__,
+            ),
+        )
 
-        Args:
-            inputs: Optional dictionary of inputs to initialize or update the state.
-
-        Returns:
-            The final output from the flow execution.
-        """
         if inputs is not None:
             self._initialize_state(inputs)
         return asyncio.run(self.kickoff_async())
 
     async def kickoff_async(self, inputs: Optional[Dict[str, Any]] = None) -> Any:
-        """
-        Starts the execution of the flow asynchronously.
-
-        Args:
-            inputs: Optional dictionary of inputs to initialize or update the state.
-
-        Returns:
-            The final output from the flow execution.
-        """
-        if inputs is not None:
-            self._initialize_state(inputs)
         if not self._start_methods:
             raise ValueError("No start method defined")
 
@@ -276,22 +491,42 @@ class Flow(Generic[T], metaclass=FlowMeta):
             self.__class__.__name__, list(self._methods.keys())
         )
 
-        # Create tasks for all start methods
         tasks = [
             self._execute_start_method(start_method)
             for start_method in self._start_methods
         ]
-
-        # Run all start methods concurrently
         await asyncio.gather(*tasks)
 
-        # Return the final output (from the last executed method)
-        if self._method_outputs:
-            return self._method_outputs[-1]
-        else:
-            return None  # Or raise an exception if no methods were executed
+        final_output = self._method_outputs[-1] if self._method_outputs else None
+
+        self.event_emitter.send(
+            self,
+            event=FlowFinishedEvent(
+                type="flow_finished",
+                flow_name=self.__class__.__name__,
+                result=final_output,
+            ),
+        )
+        return final_output
 
     async def _execute_start_method(self, start_method_name: str) -> None:
+        """
+        Executes a flow's start method and its triggered listeners.
+
+        This internal method handles the execution of methods marked with @start
+        decorator and manages the subsequent chain of listener executions.
+
+        Parameters
+        ----------
+        start_method_name : str
+            The name of the start method to execute.
+
+        Notes
+        -----
+        - Executes the start method and captures its result
+        - Triggers execution of any listeners waiting on this start method
+        - Part of the flow's initialization sequence
+        """
         result = await self._execute_method(
             start_method_name, self._methods[start_method_name]
         )
@@ -305,70 +540,181 @@ class Flow(Generic[T], metaclass=FlowMeta):
             if asyncio.iscoroutinefunction(method)
             else method(*args, **kwargs)
         )
-        self._method_outputs.append(result)  # Store the output
-
-        # Track method execution counts
+        self._method_outputs.append(result)
         self._method_execution_counts[method_name] = (
             self._method_execution_counts.get(method_name, 0) + 1
         )
-
         return result
 
     async def _execute_listeners(self, trigger_method: str, result: Any) -> None:
-        listener_tasks = []
+        """
+        Executes all listeners and routers triggered by a method completion.
 
-        if trigger_method in self._routers:
-            router_method = self._methods[self._routers[trigger_method]]
-            path = await self._execute_method(
-                self._routers[trigger_method], router_method
+        This internal method manages the execution flow by:
+        1. First executing all triggered routers sequentially
+        2. Then executing all triggered listeners in parallel
+
+        Parameters
+        ----------
+        trigger_method : str
+            The name of the method that triggered these listeners.
+        result : Any
+            The result from the triggering method, passed to listeners
+            that accept parameters.
+
+        Notes
+        -----
+        - Routers are executed sequentially to maintain flow control
+        - Each router's result becomes the new trigger_method
+        - Normal listeners are executed in parallel for efficiency
+        - Listeners can receive the trigger method's result as a parameter
+        """
+        # First, handle routers repeatedly until no router triggers anymore
+        while True:
+            routers_triggered = self._find_triggered_methods(
+                trigger_method, router_only=True
             )
-            trigger_method = path
+            if not routers_triggered:
+                break
+            for router_name in routers_triggered:
+                await self._execute_single_listener(router_name, result)
+                # After executing router, the router's result is the path
+                # The last router executed sets the trigger_method
+                # The router result is the last element in self._method_outputs
+                trigger_method = self._method_outputs[-1]
 
+        # Now that no more routers are triggered by current trigger_method,
+        # execute normal listeners
+        listeners_triggered = self._find_triggered_methods(
+            trigger_method, router_only=False
+        )
+        if listeners_triggered:
+            tasks = [
+                self._execute_single_listener(listener_name, result)
+                for listener_name in listeners_triggered
+            ]
+            await asyncio.gather(*tasks)
+
+    def _find_triggered_methods(
+        self, trigger_method: str, router_only: bool
+    ) -> List[str]:
+        """
+        Finds all methods that should be triggered based on conditions.
+
+        This internal method evaluates both OR and AND conditions to determine
+        which methods should be executed next in the flow.
+
+        Parameters
+        ----------
+        trigger_method : str
+            The name of the method that just completed execution.
+        router_only : bool
+            If True, only consider router methods.
+            If False, only consider non-router methods.
+
+        Returns
+        -------
+        List[str]
+            Names of methods that should be triggered.
+
+        Notes
+        -----
+        - Handles both OR and AND conditions:
+          * OR: Triggers if any condition is met
+          * AND: Triggers only when all conditions are met
+        - Maintains state for AND conditions using _pending_and_listeners
+        - Separates router and normal listener evaluation
+        """
+        triggered = []
         for listener_name, (condition_type, methods) in self._listeners.items():
+            is_router = listener_name in self._routers
+
+            if router_only != is_router:
+                continue
+
             if condition_type == "OR":
+                # If the trigger_method matches any in methods, run this
                 if trigger_method in methods:
-                    # Schedule the listener without preventing re-execution
-                    listener_tasks.append(
-                        self._execute_single_listener(listener_name, result)
-                    )
+                    triggered.append(listener_name)
             elif condition_type == "AND":
                 # Initialize pending methods for this listener if not already done
                 if listener_name not in self._pending_and_listeners:
                     self._pending_and_listeners[listener_name] = set(methods)
                 # Remove the trigger method from pending methods
-                self._pending_and_listeners[listener_name].discard(trigger_method)
+                if trigger_method in self._pending_and_listeners[listener_name]:
+                    self._pending_and_listeners[listener_name].discard(trigger_method)
+
                 if not self._pending_and_listeners[listener_name]:
                     # All required methods have been executed
-                    listener_tasks.append(
-                        self._execute_single_listener(listener_name, result)
-                    )
+                    triggered.append(listener_name)
                     # Reset pending methods for this listener
                     self._pending_and_listeners.pop(listener_name, None)
 
-        # Run all listener tasks concurrently and wait for them to complete
-        if listener_tasks:
-            await asyncio.gather(*listener_tasks)
+        return triggered
 
     async def _execute_single_listener(self, listener_name: str, result: Any) -> None:
+        """
+        Executes a single listener method with proper event handling.
+
+        This internal method manages the execution of an individual listener,
+        including parameter inspection, event emission, and error handling.
+
+        Parameters
+        ----------
+        listener_name : str
+            The name of the listener method to execute.
+        result : Any
+            The result from the triggering method, which may be passed
+            to the listener if it accepts parameters.
+
+        Notes
+        -----
+        - Inspects method signature to determine if it accepts the trigger result
+        - Emits events for method execution start and finish
+        - Handles errors gracefully with detailed logging
+        - Recursively triggers listeners of this listener
+        - Supports both parameterized and parameter-less listeners
+
+        Error Handling
+        -------------
+        Catches and logs any exceptions during execution, preventing
+        individual listener failures from breaking the entire flow.
+        """
         try:
             method = self._methods[listener_name]
+
+            self.event_emitter.send(
+                self,
+                event=MethodExecutionStartedEvent(
+                    type="method_execution_started",
+                    method_name=listener_name,
+                    flow_name=self.__class__.__name__,
+                ),
+            )
+
             sig = inspect.signature(method)
             params = list(sig.parameters.values())
-
-            # Exclude 'self' parameter
             method_params = [p for p in params if p.name != "self"]
 
             if method_params:
-                # If listener expects parameters, pass the result
                 listener_result = await self._execute_method(
                     listener_name, method, result
                 )
             else:
-                # If listener does not expect parameters, call without arguments
                 listener_result = await self._execute_method(listener_name, method)
 
-            # Execute listeners of this listener
+            self.event_emitter.send(
+                self,
+                event=MethodExecutionFinishedEvent(
+                    type="method_execution_finished",
+                    method_name=listener_name,
+                    flow_name=self.__class__.__name__,
+                ),
+            )
+
+            # Execute listeners (and possibly routers) of this listener
             await self._execute_listeners(listener_name, listener_result)
+
         except Exception as e:
             print(
                 f"[Flow._execute_single_listener] Error in method {listener_name}: {e}"
@@ -381,5 +727,4 @@ class Flow(Generic[T], metaclass=FlowMeta):
         self._telemetry.flow_plotting_span(
             self.__class__.__name__, list(self._methods.keys())
         )
-
         plot_flow(self, filename)

src/crewai/flow/flow_events.py

@@ -0,0 +1,33 @@
+from dataclasses import dataclass, field
+from datetime import datetime
+from typing import Any, Optional
+
+
+@dataclass
+class Event:
+    type: str
+    flow_name: str
+    timestamp: datetime = field(init=False)
+
+    def __post_init__(self):
+        self.timestamp = datetime.now()
+
+
+@dataclass
+class FlowStartedEvent(Event):
+    pass
+
+
+@dataclass
+class MethodExecutionStartedEvent(Event):
+    method_name: str
+
+
+@dataclass
+class MethodExecutionFinishedEvent(Event):
+    method_name: str
+
+
+@dataclass
+class FlowFinishedEvent(Event):
+    result: Optional[Any] = None

src/crewai/flow/flow_visualizer.py

@@ -1,12 +1,14 @@
 # flow_visualizer.py
 
 import os
+from pathlib import Path
 
 from pyvis.network import Network
 
 from crewai.flow.config import COLORS, NODE_STYLES
 from crewai.flow.html_template_handler import HTMLTemplateHandler
 from crewai.flow.legend_generator import generate_legend_items_html, get_legend_items
+from crewai.flow.path_utils import safe_path_join, validate_path_exists
 from crewai.flow.utils import calculate_node_levels
 from crewai.flow.visualization_utils import (
     add_edges,
@@ -16,89 +18,209 @@ from crewai.flow.visualization_utils import (
 
 
 class FlowPlot:
+    """Handles the creation and rendering of flow visualization diagrams."""
+
     def __init__(self, flow):
+        """
+        Initialize FlowPlot with a flow object.
+
+        Parameters
+        ----------
+        flow : Flow
+            A Flow instance to visualize.
+
+        Raises
+        ------
+        ValueError
+            If flow object is invalid or missing required attributes.
+        """
+        if not hasattr(flow, '_methods'):
+            raise ValueError("Invalid flow object: missing '_methods' attribute")
+        if not hasattr(flow, '_listeners'):
+            raise ValueError("Invalid flow object: missing '_listeners' attribute")
+        if not hasattr(flow, '_start_methods'):
+            raise ValueError("Invalid flow object: missing '_start_methods' attribute")
+            
         self.flow = flow
         self.colors = COLORS
         self.node_styles = NODE_STYLES
 
     def plot(self, filename):
-        net = Network(
-            directed=True,
-            height="750px",
-            width="100%",
-            bgcolor=self.colors["bg"],
-            layout=None,
-        )
-
-        # Set options to disable physics
-        net.set_options(
-            """
-            var options = {
-                "nodes": {
-                    "font": {
-                        "multi": "html"
-                    }
-                },
-                "physics": {
-                    "enabled": false
-                }
-            }
         """
-        )
+        Generate and save an HTML visualization of the flow.
 
-        # Calculate levels for nodes
-        node_levels = calculate_node_levels(self.flow)
+        Parameters
+        ----------
+        filename : str
+            Name of the output file (without extension).
 
-        # Compute positions
-        node_positions = compute_positions(self.flow, node_levels)
+        Raises
+        ------
+        ValueError
+            If filename is invalid or network generation fails.
+        IOError
+            If file operations fail or visualization cannot be generated.
+        RuntimeError
+            If network visualization generation fails.
+        """
+        if not filename or not isinstance(filename, str):
+            raise ValueError("Filename must be a non-empty string")
+            
+        try:
+            # Initialize network
+            net = Network(
+                directed=True,
+                height="750px",
+                width="100%",
+                bgcolor=self.colors["bg"],
+                layout=None,
+            )
 
-        # Add nodes to the network
-        add_nodes_to_network(net, self.flow, node_positions, self.node_styles)
+            # Set options to disable physics
+            net.set_options(
+                """
+                var options = {
+                    "nodes": {
+                        "font": {
+                            "multi": "html"
+                        }
+                    },
+                    "physics": {
+                        "enabled": false
+                    }
+                }
+            """
+            )
 
-        # Add edges to the network
-        add_edges(net, self.flow, node_positions, self.colors)
+            # Calculate levels for nodes
+            try:
+                node_levels = calculate_node_levels(self.flow)
+            except Exception as e:
+                raise ValueError(f"Failed to calculate node levels: {str(e)}")
 
-        network_html = net.generate_html()
-        final_html_content = self._generate_final_html(network_html)
+            # Compute positions
+            try:
+                node_positions = compute_positions(self.flow, node_levels)
+            except Exception as e:
+                raise ValueError(f"Failed to compute node positions: {str(e)}")
 
-        # Save the final HTML content to the file
-        with open(f"{filename}.html", "w", encoding="utf-8") as f:
-            f.write(final_html_content)
-        print(f"Plot saved as {filename}.html")
+            # Add nodes to the network
+            try:
+                add_nodes_to_network(net, self.flow, node_positions, self.node_styles)
+            except Exception as e:
+                raise RuntimeError(f"Failed to add nodes to network: {str(e)}")
 
-        self._cleanup_pyvis_lib()
+            # Add edges to the network
+            try:
+                add_edges(net, self.flow, node_positions, self.colors)
+            except Exception as e:
+                raise RuntimeError(f"Failed to add edges to network: {str(e)}")
+
+            # Generate HTML
+            try:
+                network_html = net.generate_html()
+                final_html_content = self._generate_final_html(network_html)
+            except Exception as e:
+                raise RuntimeError(f"Failed to generate network visualization: {str(e)}")
+
+            # Save the final HTML content to the file
+            try:
+                with open(f"{filename}.html", "w", encoding="utf-8") as f:
+                    f.write(final_html_content)
+                print(f"Plot saved as {filename}.html")
+            except IOError as e:
+                raise IOError(f"Failed to save flow visualization to {filename}.html: {str(e)}")
+
+        except (ValueError, RuntimeError, IOError) as e:
+            raise e
+        except Exception as e:
+            raise RuntimeError(f"Unexpected error during flow visualization: {str(e)}")
+        finally:
+            self._cleanup_pyvis_lib()
 
     def _generate_final_html(self, network_html):
-        # Extract just the body content from the generated HTML
-        current_dir = os.path.dirname(__file__)
-        template_path = os.path.join(
-            current_dir, "assets", "crewai_flow_visual_template.html"
-        )
-        logo_path = os.path.join(current_dir, "assets", "crewai_logo.svg")
+        """
+        Generate the final HTML content with network visualization and legend.
 
-        html_handler = HTMLTemplateHandler(template_path, logo_path)
-        network_body = html_handler.extract_body_content(network_html)
+        Parameters
+        ----------
+        network_html : str
+            HTML content generated by pyvis Network.
 
-        # Generate the legend items HTML
-        legend_items = get_legend_items(self.colors)
-        legend_items_html = generate_legend_items_html(legend_items)
-        final_html_content = html_handler.generate_final_html(
-            network_body, legend_items_html
-        )
-        return final_html_content
+        Returns
+        -------
+        str
+            Complete HTML content with styling and legend.
+
+        Raises
+        ------
+        IOError
+            If template or logo files cannot be accessed.
+        ValueError
+            If network_html is invalid.
+        """
+        if not network_html:
+            raise ValueError("Invalid network HTML content")
+
+        try:
+            # Extract just the body content from the generated HTML
+            current_dir = os.path.dirname(__file__)
+            template_path = safe_path_join("assets", "crewai_flow_visual_template.html", root=current_dir)
+            logo_path = safe_path_join("assets", "crewai_logo.svg", root=current_dir)
+
+            if not os.path.exists(template_path):
+                raise IOError(f"Template file not found: {template_path}")
+            if not os.path.exists(logo_path):
+                raise IOError(f"Logo file not found: {logo_path}")
+
+            html_handler = HTMLTemplateHandler(template_path, logo_path)
+            network_body = html_handler.extract_body_content(network_html)
+
+            # Generate the legend items HTML
+            legend_items = get_legend_items(self.colors)
+            legend_items_html = generate_legend_items_html(legend_items)
+            final_html_content = html_handler.generate_final_html(
+                network_body, legend_items_html
+            )
+            return final_html_content
+        except Exception as e:
+            raise IOError(f"Failed to generate visualization HTML: {str(e)}")
 
     def _cleanup_pyvis_lib(self):
-        # Clean up the generated lib folder
-        lib_folder = os.path.join(os.getcwd(), "lib")
+        """
+        Clean up the generated lib folder from pyvis.
+
+        This method safely removes the temporary lib directory created by pyvis
+        during network visualization generation.
+        """
         try:
+            lib_folder = safe_path_join("lib", root=os.getcwd())
             if os.path.exists(lib_folder) and os.path.isdir(lib_folder):
                 import shutil
-
                 shutil.rmtree(lib_folder)
+        except ValueError as e:
+            print(f"Error validating lib folder path: {e}")
         except Exception as e:
-            print(f"Error cleaning up {lib_folder}: {e}")
+            print(f"Error cleaning up lib folder: {e}")
 
 
 def plot_flow(flow, filename="flow_plot"):
+    """
+    Convenience function to create and save a flow visualization.
+
+    Parameters
+    ----------
+    flow : Flow
+        Flow instance to visualize.
+    filename : str, optional
+        Output filename without extension, by default "flow_plot".
+
+    Raises
+    ------
+    ValueError
+        If flow object or filename is invalid.
+    IOError
+        If file operations fail.
+    """
     visualizer = FlowPlot(flow)
     visualizer.plot(filename)

src/crewai/flow/html_template_handler.py

@@ -1,26 +1,53 @@
 import base64
 import re
+from pathlib import Path
+
+from crewai.flow.path_utils import safe_path_join, validate_path_exists
 
 
 class HTMLTemplateHandler:
+    """Handles HTML template processing and generation for flow visualization diagrams."""
+
     def __init__(self, template_path, logo_path):
-        self.template_path = template_path
-        self.logo_path = logo_path
+        """
+        Initialize HTMLTemplateHandler with validated template and logo paths.
+
+        Parameters
+        ----------
+        template_path : str
+            Path to the HTML template file.
+        logo_path : str
+            Path to the logo image file.
+
+        Raises
+        ------
+        ValueError
+            If template or logo paths are invalid or files don't exist.
+        """
+        try:
+            self.template_path = validate_path_exists(template_path, "file")
+            self.logo_path = validate_path_exists(logo_path, "file")
+        except ValueError as e:
+            raise ValueError(f"Invalid template or logo path: {e}")
 
     def read_template(self):
+        """Read and return the HTML template file contents."""
         with open(self.template_path, "r", encoding="utf-8") as f:
             return f.read()
 
     def encode_logo(self):
+        """Convert the logo SVG file to base64 encoded string."""
         with open(self.logo_path, "rb") as logo_file:
             logo_svg_data = logo_file.read()
             return base64.b64encode(logo_svg_data).decode("utf-8")
 
     def extract_body_content(self, html):
+        """Extract and return content between body tags from HTML string."""
         match = re.search("<body.*?>(.*?)</body>", html, re.DOTALL)
         return match.group(1) if match else ""
 
     def generate_legend_items_html(self, legend_items):
+        """Generate HTML markup for the legend items."""
         legend_items_html = ""
         for item in legend_items:
             if "border" in item:
@@ -48,6 +75,7 @@ class HTMLTemplateHandler:
         return legend_items_html
 
     def generate_final_html(self, network_body, legend_items_html, title="Flow Plot"):
+        """Combine all components into final HTML document with network visualization."""
         html_template = self.read_template()
         logo_svg_base64 = self.encode_logo()
 

src/crewai/flow/legend_generator.py

@@ -1,3 +1,4 @@
+
 def get_legend_items(colors):
     return [
         {"label": "Start Method", "color": colors["start"]},

src/crewai/flow/path_utils.py

@@ -0,0 +1,135 @@
+"""
+Path utilities for secure file operations in CrewAI flow module.
+
+This module provides utilities for secure path handling to prevent directory
+traversal attacks and ensure paths remain within allowed boundaries.
+"""
+
+import os
+from pathlib import Path
+from typing import List, Union
+
+
+def safe_path_join(*parts: str, root: Union[str, Path, None] = None) -> str:
+    """
+    Safely join path components and ensure the result is within allowed boundaries.
+
+    Parameters
+    ----------
+    *parts : str
+        Variable number of path components to join.
+    root : Union[str, Path, None], optional
+        Root directory to use as base. If None, uses current working directory.
+
+    Returns
+    -------
+    str
+        String representation of the resolved path.
+
+    Raises
+    ------
+    ValueError
+        If the resulting path would be outside the root directory
+        or if any path component is invalid.
+    """
+    if not parts:
+        raise ValueError("No path components provided")
+
+    try:
+        # Convert all parts to strings and clean them
+        clean_parts = [str(part).strip() for part in parts if part]
+        if not clean_parts:
+            raise ValueError("No valid path components provided")
+
+        # Establish root directory
+        root_path = Path(root).resolve() if root else Path.cwd()
+        
+        # Join and resolve the full path
+        full_path = Path(root_path, *clean_parts).resolve()
+        
+        # Check if the resolved path is within root
+        if not str(full_path).startswith(str(root_path)):
+            raise ValueError(
+                f"Invalid path: Potential directory traversal. Path must be within {root_path}"
+            )
+            
+        return str(full_path)
+        
+    except Exception as e:
+        if isinstance(e, ValueError):
+            raise
+        raise ValueError(f"Invalid path components: {str(e)}")
+
+
+def validate_path_exists(path: Union[str, Path], file_type: str = "file") -> str:
+    """
+    Validate that a path exists and is of the expected type.
+
+    Parameters
+    ----------
+    path : Union[str, Path]
+        Path to validate.
+    file_type : str, optional
+        Expected type ('file' or 'directory'), by default 'file'.
+
+    Returns
+    -------
+    str
+        Validated path as string.
+
+    Raises
+    ------
+    ValueError
+        If path doesn't exist or is not of expected type.
+    """
+    try:
+        path_obj = Path(path).resolve()
+        
+        if not path_obj.exists():
+            raise ValueError(f"Path does not exist: {path}")
+            
+        if file_type == "file" and not path_obj.is_file():
+            raise ValueError(f"Path is not a file: {path}")
+        elif file_type == "directory" and not path_obj.is_dir():
+            raise ValueError(f"Path is not a directory: {path}")
+            
+        return str(path_obj)
+        
+    except Exception as e:
+        if isinstance(e, ValueError):
+            raise
+        raise ValueError(f"Invalid path: {str(e)}")
+
+
+def list_files(directory: Union[str, Path], pattern: str = "*") -> List[str]:
+    """
+    Safely list files in a directory matching a pattern.
+
+    Parameters
+    ----------
+    directory : Union[str, Path]
+        Directory to search in.
+    pattern : str, optional
+        Glob pattern to match files against, by default "*".
+
+    Returns
+    -------
+    List[str]
+        List of matching file paths.
+
+    Raises
+    ------
+    ValueError
+        If directory is invalid or inaccessible.
+    """
+    try:
+        dir_path = Path(directory).resolve()
+        if not dir_path.is_dir():
+            raise ValueError(f"Not a directory: {directory}")
+            
+        return [str(p) for p in dir_path.glob(pattern) if p.is_file()]
+        
+    except Exception as e:
+        if isinstance(e, ValueError):
+            raise
+        raise ValueError(f"Error listing files: {str(e)}")

src/crewai/flow/utils.py

@@ -1,9 +1,25 @@
+"""
+Utility functions for flow visualization and dependency analysis.
+
+This module provides core functionality for analyzing and manipulating flow structures,
+including node level calculation, ancestor tracking, and return value analysis.
+Functions in this module are primarily used by the visualization system to create
+accurate and informative flow diagrams.
+
+Example
+-------
+>>> flow = Flow()
+>>> node_levels = calculate_node_levels(flow)
+>>> ancestors = build_ancestor_dict(flow)
+"""
+
 import ast
 import inspect
 import textwrap
+from typing import Any, Dict, List, Optional, Set, Union
 
 
-def get_possible_return_constants(function):
+def get_possible_return_constants(function: Any) -> Optional[List[str]]:
     try:
         source = inspect.getsource(function)
     except OSError:
@@ -31,23 +47,80 @@ def get_possible_return_constants(function):
         print(f"Source code:\n{source}")
         return None
 
-    return_values = []
+    return_values = set()
+    dict_definitions = {}
+
+    class DictionaryAssignmentVisitor(ast.NodeVisitor):
+        def visit_Assign(self, node):
+            # Check if this assignment is assigning a dictionary literal to a variable
+            if isinstance(node.value, ast.Dict) and len(node.targets) == 1:
+                target = node.targets[0]
+                if isinstance(target, ast.Name):
+                    var_name = target.id
+                    dict_values = []
+                    # Extract string values from the dictionary
+                    for val in node.value.values:
+                        if isinstance(val, ast.Constant) and isinstance(val.value, str):
+                            dict_values.append(val.value)
+                        # If non-string, skip or just ignore
+                    if dict_values:
+                        dict_definitions[var_name] = dict_values
+            self.generic_visit(node)
 
     class ReturnVisitor(ast.NodeVisitor):
         def visit_Return(self, node):
-            # Check if the return value is a constant (Python 3.8+)
-            if isinstance(node.value, ast.Constant):
-                return_values.append(node.value.value)
+            # Direct string return
+            if isinstance(node.value, ast.Constant) and isinstance(
+                node.value.value, str
+            ):
+                return_values.add(node.value.value)
+            # Dictionary-based return, like return paths[result]
+            elif isinstance(node.value, ast.Subscript):
+                # Check if we're subscripting a known dictionary variable
+                if isinstance(node.value.value, ast.Name):
+                    var_name = node.value.value.id
+                    if var_name in dict_definitions:
+                        # Add all possible dictionary values
+                        for v in dict_definitions[var_name]:
+                            return_values.add(v)
+            self.generic_visit(node)
 
+    # First pass: identify dictionary assignments
+    DictionaryAssignmentVisitor().visit(code_ast)
+    # Second pass: identify returns
     ReturnVisitor().visit(code_ast)
-    return return_values
+
+    return list(return_values) if return_values else None
 
 
-def calculate_node_levels(flow):
-    levels = {}
-    queue = []
-    visited = set()
-    pending_and_listeners = {}
+def calculate_node_levels(flow: Any) -> Dict[str, int]:
+    """
+    Calculate the hierarchical level of each node in the flow.
+
+    Performs a breadth-first traversal of the flow graph to assign levels
+    to nodes, starting with start methods at level 0.
+
+    Parameters
+    ----------
+    flow : Any
+        The flow instance containing methods, listeners, and router configurations.
+
+    Returns
+    -------
+    Dict[str, int]
+        Dictionary mapping method names to their hierarchical levels.
+
+    Notes
+    -----
+    - Start methods are assigned level 0
+    - Each subsequent connected node is assigned level = parent_level + 1
+    - Handles both OR and AND conditions for listeners
+    - Processes router paths separately
+    """
+    levels: Dict[str, int] = {}
+    queue: List[str] = []
+    visited: Set[str] = set()
+    pending_and_listeners: Dict[str, Set[str]] = {}
 
     # Make all start methods at level 0
     for method_name, method in flow._methods.items():
@@ -61,10 +134,7 @@ def calculate_node_levels(flow):
         current_level = levels[current]
         visited.add(current)
 
-        for listener_name, (
-            condition_type,
-            trigger_methods,
-        ) in flow._listeners.items():
+        for listener_name, (condition_type, trigger_methods) in flow._listeners.items():
             if condition_type == "OR":
                 if current in trigger_methods:
                     if (
@@ -89,7 +159,7 @@ def calculate_node_levels(flow):
                             queue.append(listener_name)
 
         # Handle router connections
-        if current in flow._routers.values():
+        if current in flow._routers:
             router_method_name = current
             paths = flow._router_paths.get(router_method_name, [])
             for path in paths:
@@ -105,10 +175,24 @@ def calculate_node_levels(flow):
                             levels[listener_name] = current_level + 1
                             if listener_name not in visited:
                                 queue.append(listener_name)
+
     return levels
 
 
-def count_outgoing_edges(flow):
+def count_outgoing_edges(flow: Any) -> Dict[str, int]:
+    """
+    Count the number of outgoing edges for each method in the flow.
+
+    Parameters
+    ----------
+    flow : Any
+        The flow instance to analyze.
+
+    Returns
+    -------
+    Dict[str, int]
+        Dictionary mapping method names to their outgoing edge count.
+    """
     counts = {}
     for method_name in flow._methods:
         counts[method_name] = 0
@@ -120,16 +204,53 @@ def count_outgoing_edges(flow):
     return counts
 
 
-def build_ancestor_dict(flow):
-    ancestors = {node: set() for node in flow._methods}
-    visited = set()
+def build_ancestor_dict(flow: Any) -> Dict[str, Set[str]]:
+    """
+    Build a dictionary mapping each node to its ancestor nodes.
+
+    Parameters
+    ----------
+    flow : Any
+        The flow instance to analyze.
+
+    Returns
+    -------
+    Dict[str, Set[str]]
+        Dictionary mapping each node to a set of its ancestor nodes.
+    """
+    ancestors: Dict[str, Set[str]] = {node: set() for node in flow._methods}
+    visited: Set[str] = set()
     for node in flow._methods:
         if node not in visited:
             dfs_ancestors(node, ancestors, visited, flow)
     return ancestors
 
 
-def dfs_ancestors(node, ancestors, visited, flow):
+def dfs_ancestors(
+    node: str,
+    ancestors: Dict[str, Set[str]],
+    visited: Set[str],
+    flow: Any
+) -> None:
+    """
+    Perform depth-first search to build ancestor relationships.
+
+    Parameters
+    ----------
+    node : str
+        Current node being processed.
+    ancestors : Dict[str, Set[str]]
+        Dictionary tracking ancestor relationships.
+    visited : Set[str]
+        Set of already visited nodes.
+    flow : Any
+        The flow instance being analyzed.
+
+    Notes
+    -----
+    This function modifies the ancestors dictionary in-place to build
+    the complete ancestor graph.
+    """
     if node in visited:
         return
     visited.add(node)
@@ -142,7 +263,7 @@ def dfs_ancestors(node, ancestors, visited, flow):
             dfs_ancestors(listener_name, ancestors, visited, flow)
 
     # Handle router methods separately
-    if node in flow._routers.values():
+    if node in flow._routers:
         router_method_name = node
         paths = flow._router_paths.get(router_method_name, [])
         for path in paths:
@@ -153,12 +274,48 @@ def dfs_ancestors(node, ancestors, visited, flow):
                     dfs_ancestors(listener_name, ancestors, visited, flow)
 
 
-def is_ancestor(node, ancestor_candidate, ancestors):
+def is_ancestor(node: str, ancestor_candidate: str, ancestors: Dict[str, Set[str]]) -> bool:
+    """
+    Check if one node is an ancestor of another.
+
+    Parameters
+    ----------
+    node : str
+        The node to check ancestors for.
+    ancestor_candidate : str
+        The potential ancestor node.
+    ancestors : Dict[str, Set[str]]
+        Dictionary containing ancestor relationships.
+
+    Returns
+    -------
+    bool
+        True if ancestor_candidate is an ancestor of node, False otherwise.
+    """
     return ancestor_candidate in ancestors.get(node, set())
 
 
-def build_parent_children_dict(flow):
-    parent_children = {}
+def build_parent_children_dict(flow: Any) -> Dict[str, List[str]]:
+    """
+    Build a dictionary mapping parent nodes to their children.
+
+    Parameters
+    ----------
+    flow : Any
+        The flow instance to analyze.
+
+    Returns
+    -------
+    Dict[str, List[str]]
+        Dictionary mapping parent method names to lists of their child method names.
+
+    Notes
+    -----
+    - Maps listeners to their trigger methods
+    - Maps router methods to their paths and listeners
+    - Children lists are sorted for consistent ordering
+    """
+    parent_children: Dict[str, List[str]] = {}
 
     # Map listeners to their trigger methods
     for listener_name, (_, trigger_methods) in flow._listeners.items():
@@ -182,7 +339,24 @@ def build_parent_children_dict(flow):
     return parent_children
 
 
-def get_child_index(parent, child, parent_children):
+def get_child_index(parent: str, child: str, parent_children: Dict[str, List[str]]) -> int:
+    """
+    Get the index of a child node in its parent's sorted children list.
+
+    Parameters
+    ----------
+    parent : str
+        The parent node name.
+    child : str
+        The child node name to find the index for.
+    parent_children : Dict[str, List[str]]
+        Dictionary mapping parents to their children lists.
+
+    Returns
+    -------
+    int
+        Zero-based index of the child in its parent's sorted children list.
+    """
     children = parent_children.get(parent, [])
     children.sort()
     return children.index(child)

src/crewai/flow/visualization_utils.py

@@ -1,5 +1,23 @@
+"""
+Utilities for creating visual representations of flow structures.
+
+This module provides functions for generating network visualizations of flows,
+including node placement, edge creation, and visual styling. It handles the
+conversion of flow structures into visual network graphs with appropriate
+styling and layout.
+
+Example
+-------
+>>> flow = Flow()
+>>> net = Network(directed=True)
+>>> node_positions = compute_positions(flow, node_levels)
+>>> add_nodes_to_network(net, flow, node_positions, node_styles)
+>>> add_edges(net, flow, node_positions, colors)
+"""
+
 import ast
 import inspect
+from typing import Any, Dict, List, Optional, Tuple, Union
 
 from .utils import (
     build_ancestor_dict,
@@ -9,8 +27,25 @@ from .utils import (
 )
 
 
-def method_calls_crew(method):
-    """Check if the method calls `.crew()`."""
+def method_calls_crew(method: Any) -> bool:
+    """
+    Check if the method contains a call to `.crew()`.
+
+    Parameters
+    ----------
+    method : Any
+        The method to analyze for crew() calls.
+
+    Returns
+    -------
+    bool
+        True if the method calls .crew(), False otherwise.
+
+    Notes
+    -----
+    Uses AST analysis to detect method calls, specifically looking for
+    attribute access of 'crew'.
+    """
     try:
         source = inspect.getsource(method)
         source = inspect.cleandoc(source)
@@ -20,6 +55,7 @@ def method_calls_crew(method):
         return False
 
     class CrewCallVisitor(ast.NodeVisitor):
+        """AST visitor to detect .crew() method calls."""
         def __init__(self):
             self.found = False
 
@@ -34,7 +70,34 @@ def method_calls_crew(method):
     return visitor.found
 
 
-def add_nodes_to_network(net, flow, node_positions, node_styles):
+def add_nodes_to_network(
+    net: Any,
+    flow: Any,
+    node_positions: Dict[str, Tuple[float, float]],
+    node_styles: Dict[str, Dict[str, Any]]
+) -> None:
+    """
+    Add nodes to the network visualization with appropriate styling.
+
+    Parameters
+    ----------
+    net : Any
+        The pyvis Network instance to add nodes to.
+    flow : Any
+        The flow instance containing method information.
+    node_positions : Dict[str, Tuple[float, float]]
+        Dictionary mapping node names to their (x, y) positions.
+    node_styles : Dict[str, Dict[str, Any]]
+        Dictionary containing style configurations for different node types.
+
+    Notes
+    -----
+    Node types include:
+    - Start methods
+    - Router methods
+    - Crew methods
+    - Regular methods
+    """
     def human_friendly_label(method_name):
         return method_name.replace("_", " ").title()
 
@@ -73,9 +136,33 @@ def add_nodes_to_network(net, flow, node_positions, node_styles):
         )
 
 
-def compute_positions(flow, node_levels, y_spacing=150, x_spacing=150):
-    level_nodes = {}
-    node_positions = {}
+def compute_positions(
+    flow: Any,
+    node_levels: Dict[str, int],
+    y_spacing: float = 150,
+    x_spacing: float = 150
+) -> Dict[str, Tuple[float, float]]:
+    """
+    Compute the (x, y) positions for each node in the flow graph.
+
+    Parameters
+    ----------
+    flow : Any
+        The flow instance to compute positions for.
+    node_levels : Dict[str, int]
+        Dictionary mapping node names to their hierarchical levels.
+    y_spacing : float, optional
+        Vertical spacing between levels, by default 150.
+    x_spacing : float, optional
+        Horizontal spacing between nodes, by default 150.
+
+    Returns
+    -------
+    Dict[str, Tuple[float, float]]
+        Dictionary mapping node names to their (x, y) coordinates.
+    """
+    level_nodes: Dict[int, List[str]] = {}
+    node_positions: Dict[str, Tuple[float, float]] = {}
 
     for method_name, level in node_levels.items():
         level_nodes.setdefault(level, []).append(method_name)
@@ -90,16 +177,44 @@ def compute_positions(flow, node_levels, y_spacing=150, x_spacing=150):
     return node_positions
 
 
-def add_edges(net, flow, node_positions, colors):
+def add_edges(
+    net: Any,
+    flow: Any,
+    node_positions: Dict[str, Tuple[float, float]],
+    colors: Dict[str, str]
+) -> None:
+    edge_smooth: Dict[str, Union[str, float]] = {"type": "continuous"}  # Default value
+    """
+    Add edges to the network visualization with appropriate styling.
+
+    Parameters
+    ----------
+    net : Any
+        The pyvis Network instance to add edges to.
+    flow : Any
+        The flow instance containing edge information.
+    node_positions : Dict[str, Tuple[float, float]]
+        Dictionary mapping node names to their positions.
+    colors : Dict[str, str]
+        Dictionary mapping edge types to their colors.
+
+    Notes
+    -----
+    - Handles both normal listener edges and router edges
+    - Applies appropriate styling (color, dashes) based on edge type
+    - Adds curvature to edges when needed (cycles or multiple children)
+    """
     ancestors = build_ancestor_dict(flow)
     parent_children = build_parent_children_dict(flow)
 
+    # Edges for normal listeners
     for method_name in flow._listeners:
         condition_type, trigger_methods = flow._listeners[method_name]
         is_and_condition = condition_type == "AND"
 
         for trigger in trigger_methods:
-            if trigger in flow._methods or trigger in flow._routers.values():
+            # Check if nodes exist before adding edges
+            if trigger in node_positions and method_name in node_positions:
                 is_router_edge = any(
                     trigger in paths for paths in flow._router_paths.values()
                 )
@@ -124,7 +239,7 @@ def add_edges(net, flow, node_positions, colors):
                     else:
                         edge_smooth = {"type": "cubicBezier"}
                 else:
-                    edge_smooth = False
+                    edge_smooth.update({"type": "continuous"})
 
                 edge_style = {
                     "color": edge_color,
@@ -135,7 +250,22 @@ def add_edges(net, flow, node_positions, colors):
                 }
 
                 net.add_edge(trigger, method_name, **edge_style)
+            else:
+                # Nodes not found in node_positions. Check if it's a known router outcome and a known method.
+                is_router_edge = any(
+                    trigger in paths for paths in flow._router_paths.values()
+                )
+                # Check if method_name is a known method
+                method_known = method_name in flow._methods
 
+                # If it's a known router edge and the method is known, don't warn.
+                # This means the path is legitimate, just not reflected as nodes here.
+                if not (is_router_edge and method_known):
+                    print(
+                        f"Warning: No node found for '{trigger}' or '{method_name}'. Skipping edge."
+                    )
+
+    # Edges for router return paths
     for router_method_name, paths in flow._router_paths.items():
         for path in paths:
             for listener_name, (
@@ -143,36 +273,49 @@ def add_edges(net, flow, node_positions, colors):
                 trigger_methods,
             ) in flow._listeners.items():
                 if path in trigger_methods:
-                    is_cycle_edge = is_ancestor(trigger, method_name, ancestors)
-                    parent_has_multiple_children = (
-                        len(parent_children.get(router_method_name, [])) > 1
-                    )
-                    needs_curvature = is_cycle_edge or parent_has_multiple_children
+                    if (
+                        router_method_name in node_positions
+                        and listener_name in node_positions
+                    ):
+                        is_cycle_edge = is_ancestor(
+                            router_method_name, listener_name, ancestors
+                        )
+                        parent_has_multiple_children = (
+                            len(parent_children.get(router_method_name, [])) > 1
+                        )
+                        needs_curvature = is_cycle_edge or parent_has_multiple_children
 
-                    if needs_curvature:
-                        source_pos = node_positions.get(router_method_name)
-                        target_pos = node_positions.get(listener_name)
+                        if needs_curvature:
+                            source_pos = node_positions.get(router_method_name)
+                            target_pos = node_positions.get(listener_name)
 
-                        if source_pos and target_pos:
-                            dx = target_pos[0] - source_pos[0]
-                            smooth_type = "curvedCCW" if dx <= 0 else "curvedCW"
-                            index = get_child_index(
-                                router_method_name, listener_name, parent_children
-                            )
-                            edge_smooth = {
-                                "type": smooth_type,
-                                "roundness": 0.2 + (0.1 * index),
-                            }
+                            if source_pos and target_pos:
+                                dx = target_pos[0] - source_pos[0]
+                                smooth_type = "curvedCCW" if dx <= 0 else "curvedCW"
+                                index = get_child_index(
+                                    router_method_name, listener_name, parent_children
+                                )
+                                edge_smooth = {
+                                    "type": smooth_type,
+                                    "roundness": 0.2 + (0.1 * index),
+                                }
+                            else:
+                                edge_smooth = {"type": "cubicBezier"}
                         else:
-                            edge_smooth = {"type": "cubicBezier"}
-                    else:
-                        edge_smooth = False
+                            edge_smooth.update({"type": "continuous"})
 
-                    edge_style = {
-                        "color": colors["router_edge"],
-                        "width": 2,
-                        "arrows": "to",
-                        "dashes": True,
-                        "smooth": edge_smooth,
-                    }
-                    net.add_edge(router_method_name, listener_name, **edge_style)
+                        edge_style = {
+                            "color": colors["router_edge"],
+                            "width": 2,
+                            "arrows": "to",
+                            "dashes": True,
+                            "smooth": edge_smooth,
+                        }
+                        net.add_edge(router_method_name, listener_name, **edge_style)
+                    else:
+                        # Same check here: known router edge and known method?
+                        method_known = listener_name in flow._methods
+                        if not method_known:
+                            print(
+                                f"Warning: No node found for '{router_method_name}' or '{listener_name}'. Skipping edge."
+                            )

src/crewai/knowledge/knowledge.py

@@ -14,13 +14,13 @@ class Knowledge(BaseModel):
     Knowledge is a collection of sources and setup for the vector store to save and query relevant context.
     Args:
         sources: List[BaseKnowledgeSource] = Field(default_factory=list)
-        storage: KnowledgeStorage = Field(default_factory=KnowledgeStorage)
+        storage: Optional[KnowledgeStorage] = Field(default=None)
         embedder_config: Optional[Dict[str, Any]] = None
     """
 
     sources: List[BaseKnowledgeSource] = Field(default_factory=list)
     model_config = ConfigDict(arbitrary_types_allowed=True)
-    storage: KnowledgeStorage = Field(default_factory=KnowledgeStorage)
+    storage: Optional[KnowledgeStorage] = Field(default=None)
     embedder_config: Optional[Dict[str, Any]] = None
     collection_name: Optional[str] = None
 
@@ -49,8 +49,13 @@ class Knowledge(BaseModel):
         """
         Query across all knowledge sources to find the most relevant information.
         Returns the top_k most relevant chunks.
+        
+        Raises:
+            ValueError: If storage is not initialized.
         """
-
+        if self.storage is None:
+            raise ValueError("Storage is not initialized.")
+            
         results = self.storage.search(
             query,
             limit,

src/crewai/knowledge/source/base_file_knowledge_source.py

@@ -1,8 +1,8 @@
 from abc import ABC, abstractmethod
 from pathlib import Path
-from typing import Dict, List, Union
+from typing import Dict, List, Optional, Union
 
-from pydantic import Field
+from pydantic import Field, field_validator
 
 from crewai.knowledge.source.base_knowledge_source import BaseKnowledgeSource
 from crewai.knowledge.storage.knowledge_storage import KnowledgeStorage
@@ -14,17 +14,29 @@ class BaseFileKnowledgeSource(BaseKnowledgeSource, ABC):
     """Base class for knowledge sources that load content from files."""
 
     _logger: Logger = Logger(verbose=True)
-    file_path: Union[Path, List[Path], str, List[str]] = Field(
-        ..., description="The path to the file"
+    file_path: Optional[Union[Path, List[Path], str, List[str]]] = Field(
+        default=None,
+        description="[Deprecated] The path to the file. Use file_paths instead.",
+    )
+    file_paths: Optional[Union[Path, List[Path], str, List[str]]] = Field(
+        default_factory=list, description="The path to the file"
     )
     content: Dict[Path, str] = Field(init=False, default_factory=dict)
-    storage: KnowledgeStorage = Field(default_factory=KnowledgeStorage)
+    storage: Optional[KnowledgeStorage] = Field(default=None)
     safe_file_paths: List[Path] = Field(default_factory=list)
 
+    @field_validator("file_path", "file_paths", mode="before")
+    def validate_file_path(cls, v, info):
+        """Validate that at least one of file_path or file_paths is provided."""
+        # Single check if both are None, O(1) instead of nested conditions
+        if v is None and info.data.get("file_path" if info.field_name == "file_paths" else "file_paths") is None:
+            raise ValueError("Either file_path or file_paths must be provided")
+        return v
+
     def model_post_init(self, _):
         """Post-initialization method to load content."""
         self.safe_file_paths = self._process_file_paths()
-        self.validate_paths()
+        self.validate_content()
         self.content = self.load_content()
 
     @abstractmethod
@@ -32,13 +44,13 @@ class BaseFileKnowledgeSource(BaseKnowledgeSource, ABC):
         """Load and preprocess file content. Should be overridden by subclasses. Assume that the file path is relative to the project root in the knowledge directory."""
         pass
 
-    def validate_paths(self):
+    def validate_content(self):
         """Validate the paths."""
         for path in self.safe_file_paths:
             if not path.exists():
                 self._logger.log(
                     "error",
-                    f"File not found: {path}. Try adding sources to the knowledge directory. If its inside the knowledge directory, use the relative path.",
+                    f"File not found: {path}. Try adding sources to the knowledge directory. If it's inside the knowledge directory, use the relative path.",
                     color="red",
                 )
                 raise FileNotFoundError(f"File not found: {path}")
@@ -51,7 +63,10 @@ class BaseFileKnowledgeSource(BaseKnowledgeSource, ABC):
 
     def _save_documents(self):
         """Save the documents to the storage."""
-        self.storage.save(self.chunks)
+        if self.storage:
+            self.storage.save(self.chunks)
+        else:
+            raise ValueError("No storage found to save documents.")
 
     def convert_to_path(self, path: Union[Path, str]) -> Path:
         """Convert a path to a Path object."""
@@ -59,13 +74,30 @@ class BaseFileKnowledgeSource(BaseKnowledgeSource, ABC):
 
     def _process_file_paths(self) -> List[Path]:
         """Convert file_path to a list of Path objects."""
-        paths = (
-            [self.file_path]
-            if isinstance(self.file_path, (str, Path))
-            else self.file_path
+
+        if hasattr(self, "file_path") and self.file_path is not None:
+            self._logger.log(
+                "warning",
+                "The 'file_path' attribute is deprecated and will be removed in a future version. Please use 'file_paths' instead.",
+                color="yellow",
+            )
+            self.file_paths = self.file_path
+
+        if self.file_paths is None:
+            raise ValueError("Your source must be provided with a file_paths: []")
+
+        # Convert single path to list
+        path_list: List[Union[Path, str]] = (
+            [self.file_paths]
+            if isinstance(self.file_paths, (str, Path))
+            else list(self.file_paths)
+            if isinstance(self.file_paths, list)
+            else []
         )
 
-        if not isinstance(paths, list):
-            raise ValueError("file_path must be a Path, str, or a list of these types")
+        if not path_list:
+            raise ValueError(
+                "file_path/file_paths must be a Path, str, or a list of these types"
+            )
 
-        return [self.convert_to_path(path) for path in paths]
+        return [self.convert_to_path(path) for path in path_list]

src/crewai/knowledge/source/base_knowledge_source.py

@@ -16,12 +16,12 @@ class BaseKnowledgeSource(BaseModel, ABC):
     chunk_embeddings: List[np.ndarray] = Field(default_factory=list)
 
     model_config = ConfigDict(arbitrary_types_allowed=True)
-    storage: KnowledgeStorage = Field(default_factory=KnowledgeStorage)
+    storage: Optional[KnowledgeStorage] = Field(default=None)
     metadata: Dict[str, Any] = Field(default_factory=dict)  # Currently unused
     collection_name: Optional[str] = Field(default=None)
 
     @abstractmethod
-    def load_content(self) -> Dict[Any, str]:
+    def validate_content(self) -> Any:
         """Load and preprocess content from the source."""
         pass
 
@@ -46,4 +46,7 @@ class BaseKnowledgeSource(BaseModel, ABC):
         Save the documents to the storage.
         This method should be called after the chunks and embeddings are generated.
         """
-        self.storage.save(self.chunks)
+        if self.storage:
+            self.storage.save(self.chunks)
+        else:
+            raise ValueError("No storage found to save documents.")

src/crewai/knowledge/source/crew_docling_source.py

@@ -0,0 +1,133 @@
+from pathlib import Path
+from typing import Iterator, List, Optional, Union
+from urllib.parse import urlparse
+
+try:
+    from docling.datamodel.base_models import InputFormat
+    from docling.document_converter import DocumentConverter
+    from docling.exceptions import ConversionError
+    from docling_core.transforms.chunker.hierarchical_chunker import HierarchicalChunker
+    from docling_core.types.doc.document import DoclingDocument
+    DOCLING_AVAILABLE = True
+except ImportError:
+    DOCLING_AVAILABLE = False
+
+from pydantic import Field
+
+from crewai.knowledge.source.base_knowledge_source import BaseKnowledgeSource
+from crewai.utilities.constants import KNOWLEDGE_DIRECTORY
+from crewai.utilities.logger import Logger
+
+
+class CrewDoclingSource(BaseKnowledgeSource):
+    """Default Source class for converting documents to markdown or json
+    This will auto support PDF, DOCX, and TXT, XLSX, Images, and HTML files without any additional dependencies and follows the docling package as the source of truth.
+    """
+
+    def __init__(self, *args, **kwargs):
+        if not DOCLING_AVAILABLE:
+            raise ImportError(
+                "The docling package is required to use CrewDoclingSource. "
+                "Please install it using: uv add docling"
+            )
+        super().__init__(*args, **kwargs)
+
+    _logger: Logger = Logger(verbose=True)
+
+    file_path: Optional[List[Union[Path, str]]] = Field(default=None)
+    file_paths: List[Union[Path, str]] = Field(default_factory=list)
+    chunks: List[str] = Field(default_factory=list)
+    safe_file_paths: List[Union[Path, str]] = Field(default_factory=list)
+    content: List[DoclingDocument] = Field(default_factory=list)
+    document_converter: DocumentConverter = Field(
+        default_factory=lambda: DocumentConverter(
+            allowed_formats=[
+                InputFormat.MD,
+                InputFormat.ASCIIDOC,
+                InputFormat.PDF,
+                InputFormat.DOCX,
+                InputFormat.HTML,
+                InputFormat.IMAGE,
+                InputFormat.XLSX,
+                InputFormat.PPTX,
+            ]
+        )
+    )
+
+    def model_post_init(self, _) -> None:
+        if self.file_path:
+            self._logger.log(
+                "warning",
+                "The 'file_path' attribute is deprecated and will be removed in a future version. Please use 'file_paths' instead.",
+                color="yellow",
+            )
+            self.file_paths = self.file_path
+        self.safe_file_paths = self.validate_content()
+        self.content = self._load_content()
+
+    def _load_content(self) -> List[DoclingDocument]:
+        try:
+            return self._convert_source_to_docling_documents()
+        except ConversionError as e:
+            self._logger.log(
+                "error",
+                f"Error loading content: {e}. Supported formats: {self.document_converter.allowed_formats}",
+                "red",
+            )
+            raise e
+        except Exception as e:
+            self._logger.log("error", f"Error loading content: {e}")
+            raise e
+
+    def add(self) -> None:
+        if self.content is None:
+            return
+        for doc in self.content:
+            new_chunks_iterable = self._chunk_doc(doc)
+            self.chunks.extend(list(new_chunks_iterable))
+        self._save_documents()
+
+    def _convert_source_to_docling_documents(self) -> List[DoclingDocument]:
+        conv_results_iter = self.document_converter.convert_all(self.safe_file_paths)
+        return [result.document for result in conv_results_iter]
+
+    def _chunk_doc(self, doc: DoclingDocument) -> Iterator[str]:
+        chunker = HierarchicalChunker()
+        for chunk in chunker.chunk(doc):
+            yield chunk.text
+
+    def validate_content(self) -> List[Union[Path, str]]:
+        processed_paths: List[Union[Path, str]] = []
+        for path in self.file_paths:
+            if isinstance(path, str):
+                if path.startswith(("http://", "https://")):
+                    try:
+                        if self._validate_url(path):
+                            processed_paths.append(path)
+                        else:
+                            raise ValueError(f"Invalid URL format: {path}")
+                    except Exception as e:
+                        raise ValueError(f"Invalid URL: {path}. Error: {str(e)}")
+                else:
+                    local_path = Path(KNOWLEDGE_DIRECTORY + "/" + path)
+                    if local_path.exists():
+                        processed_paths.append(local_path)
+                    else:
+                        raise FileNotFoundError(f"File not found: {local_path}")
+            else:
+                # this is an instance of Path
+                processed_paths.append(path)
+        return processed_paths
+
+    def _validate_url(self, url: str) -> bool:
+        try:
+            result = urlparse(url)
+            return all(
+                [
+                    result.scheme in ("http", "https"),
+                    result.netloc,
+                    len(result.netloc.split(".")) >= 2,  # Ensure domain has TLD
+                ]
+            )
+        except Exception:
+            return False

src/crewai/knowledge/source/string_knowledge_source.py

@@ -13,9 +13,9 @@ class StringKnowledgeSource(BaseKnowledgeSource):
 
     def model_post_init(self, _):
         """Post-initialization method to validate content."""
-        self.load_content()
+        self.validate_content()
 
-    def load_content(self):
+    def validate_content(self):
         """Validate string content."""
         if not isinstance(self.content, str):
             raise ValueError("StringKnowledgeSource only accepts string content")

src/crewai/knowledge/storage/base_knowledge_storage.py

@@ -1,5 +1,5 @@
 from abc import ABC, abstractmethod
-from typing import Dict, Any, List, Optional
+from typing import Any, Dict, List, Optional
 
 
 class BaseKnowledgeStorage(ABC):

src/crewai/knowledge/storage/knowledge_storage.py

@@ -3,6 +3,7 @@ import hashlib
 import io
 import logging
 import os
+import shutil
 from typing import Any, Dict, List, Optional, Union, cast
 
 import chromadb
@@ -13,6 +14,7 @@ from chromadb.config import Settings
 
 from crewai.knowledge.storage.base_knowledge_storage import BaseKnowledgeStorage
 from crewai.utilities import EmbeddingConfigurator
+from crewai.utilities.constants import KNOWLEDGE_DIRECTORY
 from crewai.utilities.logger import Logger
 from crewai.utilities.paths import db_storage_path
 
@@ -105,58 +107,77 @@ class KnowledgeStorage(BaseKnowledgeStorage):
             raise Exception("Failed to create or get collection")
 
     def reset(self):
-        if self.app:
-            self.app.reset()
-        else:
-            base_path = os.path.join(db_storage_path(), "knowledge")
+        base_path = os.path.join(db_storage_path(), KNOWLEDGE_DIRECTORY)
+        if not self.app:
             self.app = chromadb.PersistentClient(
                 path=base_path,
                 settings=Settings(allow_reset=True),
             )
-            self.app.reset()
+
+        self.app.reset()
+        shutil.rmtree(base_path)
+        self.app = None
+        self.collection = None
 
     def save(
         self,
         documents: List[str],
         metadata: Optional[Union[Dict[str, Any], List[Dict[str, Any]]]] = None,
     ):
-        if self.collection:
-            try:
-                if metadata is None:
-                    metadatas: Optional[OneOrMany[chromadb.Metadata]] = None
-                elif isinstance(metadata, list):
-                    metadatas = [cast(chromadb.Metadata, m) for m in metadata]
-                else:
-                    metadatas = cast(chromadb.Metadata, metadata)
-
-                ids = [
-                    hashlib.sha256(doc.encode("utf-8")).hexdigest() for doc in documents
-                ]
-
-                self.collection.upsert(
-                    documents=documents,
-                    metadatas=metadatas,
-                    ids=ids,
-                )
-            except chromadb.errors.InvalidDimensionException as e:
-                Logger(verbose=True).log(
-                    "error",
-                    "Embedding dimension mismatch. This usually happens when mixing different embedding models. Try resetting the collection using `crewai reset-memories -a`",
-                    "red",
-                )
-                raise ValueError(
-                    "Embedding dimension mismatch. Make sure you're using the same embedding model "
-                    "across all operations with this collection."
-                    "Try resetting the collection using `crewai reset-memories -a`"
-                ) from e
-            except Exception as e:
-                Logger(verbose=True).log(
-                    "error", f"Failed to upsert documents: {e}", "red"
-                )
-                raise
-        else:
+        if not self.collection:
             raise Exception("Collection not initialized")
 
+        try:
+            # Create a dictionary to store unique documents
+            unique_docs = {}
+
+            # Generate IDs and create a mapping of id -> (document, metadata)
+            for idx, doc in enumerate(documents):
+                doc_id = hashlib.sha256(doc.encode("utf-8")).hexdigest()
+                doc_metadata = None
+                if metadata is not None:
+                    if isinstance(metadata, list):
+                        doc_metadata = metadata[idx]
+                    else:
+                        doc_metadata = metadata
+                unique_docs[doc_id] = (doc, doc_metadata)
+
+            # Prepare filtered lists for ChromaDB
+            filtered_docs = []
+            filtered_metadata = []
+            filtered_ids = []
+
+            # Build the filtered lists
+            for doc_id, (doc, meta) in unique_docs.items():
+                filtered_docs.append(doc)
+                filtered_metadata.append(meta)
+                filtered_ids.append(doc_id)
+
+            # If we have no metadata at all, set it to None
+            final_metadata: Optional[OneOrMany[chromadb.Metadata]] = (
+                None if all(m is None for m in filtered_metadata) else filtered_metadata
+            )
+
+            self.collection.upsert(
+                documents=filtered_docs,
+                metadatas=final_metadata,
+                ids=filtered_ids,
+            )
+        except chromadb.errors.InvalidDimensionException as e:
+            Logger(verbose=True).log(
+                "error",
+                "Embedding dimension mismatch. This usually happens when mixing different embedding models. Try resetting the collection using `crewai reset-memories -a`",
+                "red",
+            )
+            raise ValueError(
+                "Embedding dimension mismatch. Make sure you're using the same embedding model "
+                "across all operations with this collection."
+                "Try resetting the collection using `crewai reset-memories -a`"
+            ) from e
+        except Exception as e:
+            Logger(verbose=True).log("error", f"Failed to upsert documents: {e}", "red")
+            raise
+
     def _create_default_embedding_function(self):
         from chromadb.utils.embedding_functions.openai_embedding_function import (
             OpenAIEmbeddingFunction,

src/crewai/llm.py

@@ -1,18 +1,27 @@
+import json
 import logging
 import os
 import sys
 import threading
 import warnings
 from contextlib import contextmanager
-from typing import Any, Dict, List, Optional, Union
+from typing import Any, Dict, List, Optional, Union, cast
+
+from dotenv import load_dotenv
+
+with warnings.catch_warnings():
+    warnings.simplefilter("ignore", UserWarning)
+    import litellm
+    from litellm import Choices, get_supported_openai_params
+    from litellm.types.utils import ModelResponse
 
-import litellm
-from litellm import get_supported_openai_params
 
 from crewai.utilities.exceptions.context_window_exceeding_exception import (
     LLMContextLengthExceededException,
 )
 
+load_dotenv()
+
 
 class FilteredStream:
     def __init__(self, original_stream):
@@ -21,6 +30,7 @@ class FilteredStream:
 
     def write(self, s) -> int:
         with self._lock:
+            # Filter out extraneous messages from LiteLLM
             if (
                 "Give Feedback / Get Help: https://github.com/BerriAI/litellm/issues/new"
                 in s
@@ -43,6 +53,11 @@ LLM_CONTEXT_WINDOW_SIZES = {
     "gpt-4-turbo": 128000,
     "o1-preview": 128000,
     "o1-mini": 128000,
+    # gemini
+    "gemini-2.0-flash": 1048576,
+    "gemini-1.5-pro": 2097152,
+    "gemini-1.5-flash": 1048576,
+    "gemini-1.5-flash-8b": 1048576,
     # deepseek
     "deepseek-chat": 128000,
     # groq
@@ -59,24 +74,42 @@ LLM_CONTEXT_WINDOW_SIZES = {
     "llama3-70b-8192": 8192,
     "llama3-8b-8192": 8192,
     "mixtral-8x7b-32768": 32768,
+    "llama-3.3-70b-versatile": 128000,
+    "llama-3.3-70b-instruct": 128000,
+    # sambanova
+    "Meta-Llama-3.3-70B-Instruct": 131072,
+    "QwQ-32B-Preview": 8192,
+    "Qwen2.5-72B-Instruct": 8192,
+    "Qwen2.5-Coder-32B-Instruct": 8192,
+    "Meta-Llama-3.1-405B-Instruct": 8192,
+    "Meta-Llama-3.1-70B-Instruct": 131072,
+    "Meta-Llama-3.1-8B-Instruct": 131072,
+    "Llama-3.2-90B-Vision-Instruct": 16384,
+    "Llama-3.2-11B-Vision-Instruct": 16384,
+    "Meta-Llama-3.2-3B-Instruct": 4096,
+    "Meta-Llama-3.2-1B-Instruct": 16384,
 }
 
+DEFAULT_CONTEXT_WINDOW_SIZE = 8192
+CONTEXT_WINDOW_USAGE_RATIO = 0.75
+
 
 @contextmanager
 def suppress_warnings():
     with warnings.catch_warnings():
         warnings.filterwarnings("ignore")
+        warnings.filterwarnings(
+            "ignore", message="open_text is deprecated*", category=DeprecationWarning
+        )
 
         # Redirect stdout and stderr
         old_stdout = sys.stdout
         old_stderr = sys.stderr
         sys.stdout = FilteredStream(old_stdout)
         sys.stderr = FilteredStream(old_stderr)
-
         try:
             yield
         finally:
-            # Restore stdout and stderr
             sys.stdout = old_stdout
             sys.stderr = old_stderr
 
@@ -97,13 +130,12 @@ class LLM:
         logit_bias: Optional[Dict[int, float]] = None,
         response_format: Optional[Dict[str, Any]] = None,
         seed: Optional[int] = None,
-        logprobs: Optional[bool] = None,
+        logprobs: Optional[int] = None,
         top_logprobs: Optional[int] = None,
         base_url: Optional[str] = None,
         api_version: Optional[str] = None,
         api_key: Optional[str] = None,
         callbacks: List[Any] = [],
-        **kwargs,
     ):
         self.model = model
         self.timeout = timeout
@@ -124,19 +156,41 @@ class LLM:
         self.api_version = api_version
         self.api_key = api_key
         self.callbacks = callbacks
-        self.kwargs = kwargs
+        self.context_window_size = 0
 
         litellm.drop_params = True
-        litellm.set_verbose = False
+
         self.set_callbacks(callbacks)
         self.set_env_callbacks()
 
-    def call(self, messages: List[Dict[str, str]], callbacks: List[Any] = []) -> str:
+    def call(
+        self,
+        messages: List[Dict[str, str]],
+        tools: Optional[List[dict]] = None,
+        callbacks: Optional[List[Any]] = None,
+        available_functions: Optional[Dict[str, Any]] = None,
+    ) -> str:
+        """
+        High-level call method that:
+          1) Calls litellm.completion
+          2) Checks for function/tool calls
+          3) If a tool call is found:
+               a) executes the function
+               b) returns the result
+          4) If no tool call, returns the text response
+
+        :param messages: The conversation messages
+        :param tools: Optional list of function schemas for function calling
+        :param callbacks: Optional list of callbacks
+        :param available_functions: A dictionary mapping function_name -> actual Python function
+        :return: Final text response from the LLM or the tool result
+        """
         with suppress_warnings():
             if callbacks and len(callbacks) > 0:
                 self.set_callbacks(callbacks)
 
             try:
+                # --- 1) Make the completion call
                 params = {
                     "model": self.model,
                     "messages": messages,
@@ -157,21 +211,58 @@ class LLM:
                     "api_version": self.api_version,
                     "api_key": self.api_key,
                     "stream": False,
-                    **self.kwargs,
+                    "tools": tools,  # pass the tool schema
                 }
 
-                # Remove None values to avoid passing unnecessary parameters
                 params = {k: v for k, v in params.items() if v is not None}
 
                 response = litellm.completion(**params)
-                return response["choices"][0]["message"]["content"]
+                response_message = cast(Choices, cast(ModelResponse, response).choices)[
+                    0
+                ].message
+                text_response = response_message.content or ""
+                tool_calls = getattr(response_message, "tool_calls", [])
+
+                # --- 2) If no tool calls, return the text response
+                if not tool_calls or not available_functions:
+                    return text_response
+
+                # --- 3) Handle the tool call
+                tool_call = tool_calls[0]
+                function_name = tool_call.function.name
+
+                if function_name in available_functions:
+                    try:
+                        function_args = json.loads(tool_call.function.arguments)
+                    except json.JSONDecodeError as e:
+                        logging.warning(f"Failed to parse function arguments: {e}")
+                        return text_response
+
+                    fn = available_functions[function_name]
+                    try:
+                        # Call the actual tool function
+                        result = fn(**function_args)
+
+                        return result
+
+                    except Exception as e:
+                        logging.error(
+                            f"Error executing function '{function_name}': {e}"
+                        )
+                        return text_response
+
+                else:
+                    logging.warning(
+                        f"Tool call requested unknown function '{function_name}'"
+                    )
+                    return text_response
+
             except Exception as e:
                 if not LLMContextLengthExceededException(
                     str(e)
                 )._is_context_limit_error(str(e)):
                     logging.error(f"LiteLLM call failed: {str(e)}")
-
-                raise  # Re-raise the exception after logging
+                raise
 
     def supports_function_calling(self) -> bool:
         try:
@@ -190,20 +281,37 @@ class LLM:
             return False
 
     def get_context_window_size(self) -> int:
-        # Only using 75% of the context window size to avoid cutting the message in the middle
-        return int(LLM_CONTEXT_WINDOW_SIZES.get(self.model, 8192) * 0.75)
+        """
+        Returns the context window size, using 75% of the maximum to avoid
+        cutting off messages mid-thread.
+        """
+        if self.context_window_size != 0:
+            return self.context_window_size
+
+        self.context_window_size = int(
+            DEFAULT_CONTEXT_WINDOW_SIZE * CONTEXT_WINDOW_USAGE_RATIO
+        )
+        for key, value in LLM_CONTEXT_WINDOW_SIZES.items():
+            if self.model.startswith(key):
+                self.context_window_size = int(value * CONTEXT_WINDOW_USAGE_RATIO)
+        return self.context_window_size
 
     def set_callbacks(self, callbacks: List[Any]):
-        callback_types = [type(callback) for callback in callbacks]
-        for callback in litellm.success_callback[:]:
-            if type(callback) in callback_types:
-                litellm.success_callback.remove(callback)
+        """
+        Attempt to keep a single set of callbacks in litellm by removing old
+        duplicates and adding new ones.
+        """
+        with suppress_warnings():
+            callback_types = [type(callback) for callback in callbacks]
+            for callback in litellm.success_callback[:]:
+                if type(callback) in callback_types:
+                    litellm.success_callback.remove(callback)
 
-        for callback in litellm._async_success_callback[:]:
-            if type(callback) in callback_types:
-                litellm._async_success_callback.remove(callback)
+            for callback in litellm._async_success_callback[:]:
+                if type(callback) in callback_types:
+                    litellm._async_success_callback.remove(callback)
 
-        litellm.callbacks = callbacks
+            litellm.callbacks = callbacks
 
     def set_env_callbacks(self):
         """
@@ -224,19 +332,20 @@ class LLM:
         This will set `litellm.success_callback` to ["langfuse", "langsmith"] and
         `litellm.failure_callback` to ["langfuse"].
         """
-        success_callbacks_str = os.environ.get("LITELLM_SUCCESS_CALLBACKS", "")
-        success_callbacks = []
-        if success_callbacks_str:
-            success_callbacks = [
-                callback.strip() for callback in success_callbacks_str.split(",")
-            ]
+        with suppress_warnings():
+            success_callbacks_str = os.environ.get("LITELLM_SUCCESS_CALLBACKS", "")
+            success_callbacks = []
+            if success_callbacks_str:
+                success_callbacks = [
+                    cb.strip() for cb in success_callbacks_str.split(",") if cb.strip()
+                ]
 
-        failure_callbacks_str = os.environ.get("LITELLM_FAILURE_CALLBACKS", "")
-        failure_callbacks = []
-        if failure_callbacks_str:
-            failure_callbacks = [
-                callback.strip() for callback in failure_callbacks_str.split(",")
-            ]
+            failure_callbacks_str = os.environ.get("LITELLM_FAILURE_CALLBACKS", "")
+            failure_callbacks = []
+            if failure_callbacks_str:
+                failure_callbacks = [
+                    cb.strip() for cb in failure_callbacks_str.split(",") if cb.strip()
+                ]
 
-        litellm.success_callback = success_callbacks
-        litellm.failure_callback = failure_callbacks
+                litellm.success_callback = success_callbacks
+                litellm.failure_callback = failure_callbacks

src/crewai/memory/contextual/contextual_memory.py

@@ -1,4 +1,4 @@
-from typing import Optional, Dict, Any
+from typing import Any, Dict, Optional
 
 from crewai.memory import EntityMemory, LongTermMemory, ShortTermMemory, UserMemory
 

src/crewai/memory/memory.py

@@ -1,4 +1,4 @@
-from typing import Any, Dict, Optional, List
+from typing import Any, Dict, List, Optional
 
 from crewai.memory.storage.rag_storage import RAGStorage
 

src/crewai/memory/short_term/short_term_memory.py

@@ -1,4 +1,5 @@
 from typing import Any, Dict, Optional
+
 from crewai.memory.memory import Memory
 from crewai.memory.short_term.short_term_memory_item import ShortTermMemoryItem
 from crewai.memory.storage.rag_storage import RAGStorage
@@ -32,7 +33,10 @@ class ShortTermMemory(Memory):
                 storage
                 if storage
                 else RAGStorage(
-                    type="short_term", embedder_config=embedder_config, crew=crew, path=path
+                    type="short_term",
+                    embedder_config=embedder_config,
+                    crew=crew,
+                    path=path,
                 )
             )
         super().__init__(storage)

src/crewai/memory/storage/mem0_storage.py

@@ -2,6 +2,7 @@ import os
 from typing import Any, Dict, List
 
 from mem0 import MemoryClient
+
 from crewai.memory.storage.interface import Storage
 
 
@@ -26,10 +27,18 @@ class Mem0Storage(Storage):
             raise ValueError("User ID is required for user memory type")
 
         # API key in memory config overrides the environment variable
-        mem0_api_key = self.memory_config.get("config", {}).get("api_key") or os.getenv(
-            "MEM0_API_KEY"
-        )
-        self.memory = MemoryClient(api_key=mem0_api_key)
+        config = self.memory_config.get("config", {})
+        mem0_api_key = config.get("api_key") or os.getenv("MEM0_API_KEY")
+        mem0_org_id = config.get("org_id")
+        mem0_project_id = config.get("project_id")
+
+        # Initialize MemoryClient with available parameters
+        if mem0_org_id and mem0_project_id:
+            self.memory = MemoryClient(
+                api_key=mem0_api_key, org_id=mem0_org_id, project_id=mem0_project_id
+            )
+        else:
+            self.memory = MemoryClient(api_key=mem0_api_key)
 
     def _sanitize_role(self, role: str) -> str:
         """
@@ -56,7 +65,7 @@ class Mem0Storage(Storage):
                 metadata={"type": "long_term", **metadata},
             )
         elif self.memory_type == "entities":
-            entity_name = None
+            entity_name = self._get_agent_name()
             self.memory.add(
                 value, user_id=entity_name, metadata={"type": "entity", **metadata}
             )

src/crewai/memory/storage/rag_storage.py

@@ -150,9 +150,11 @@ class RAGStorage(BaseRAGStorage):
 
     def reset(self) -> None:
         try:
-            shutil.rmtree(f"{db_storage_path()}/{self.type}")
             if self.app:
                 self.app.reset()
+                shutil.rmtree(f"{db_storage_path()}/{self.type}")
+                self.app = None
+                self.collection = None
         except Exception as e:
             if "attempt to write a readonly database" in str(e):
                 # Ignore this specific error

src/crewai/memory/user/user_memory.py

@@ -37,7 +37,7 @@ class UserMemory(Memory):
         limit: int = 3,
         score_threshold: float = 0.35,
     ):
-        results = super().search(
+        results = self.storage.search(
             query=query,
             limit=limit,
             score_threshold=score_threshold,

src/crewai/project/annotations.py

@@ -4,18 +4,23 @@ from typing import Callable
 from crewai import Crew
 from crewai.project.utils import memoize
 
+"""Decorators for defining crew components and their behaviors."""
+
 
 def before_kickoff(func):
+    """Marks a method to execute before crew kickoff."""
     func.is_before_kickoff = True
     return func
 
 
 def after_kickoff(func):
+    """Marks a method to execute after crew kickoff."""
     func.is_after_kickoff = True
     return func
 
 
 def task(func):
+    """Marks a method as a crew task."""
     func.is_task = True
 
     @wraps(func)
@@ -29,43 +34,53 @@ def task(func):
 
 
 def agent(func):
+    """Marks a method as a crew agent."""
     func.is_agent = True
     func = memoize(func)
     return func
 
 
 def llm(func):
+    """Marks a method as an LLM provider."""
     func.is_llm = True
     func = memoize(func)
     return func
 
 
 def output_json(cls):
+    """Marks a class as JSON output format."""
     cls.is_output_json = True
     return cls
 
 
 def output_pydantic(cls):
+    """Marks a class as Pydantic output format."""
     cls.is_output_pydantic = True
     return cls
 
 
 def tool(func):
+    """Marks a method as a crew tool."""
     func.is_tool = True
     return memoize(func)
 
 
 def callback(func):
+    """Marks a method as a crew callback."""
     func.is_callback = True
     return memoize(func)
 
 
 def cache_handler(func):
+    """Marks a method as a cache handler."""
     func.is_cache_handler = True
     return memoize(func)
 
 
 def crew(func) -> Callable[..., Crew]:
+    """Marks a method as the main crew execution point."""
+
+    @wraps(func)
     def wrapper(self, *args, **kwargs) -> Crew:
         instantiated_tasks = []
         instantiated_agents = []

src/crewai/project/crew_base.py

@@ -1,4 +1,5 @@
 import inspect
+import logging
 from pathlib import Path
 from typing import Any, Callable, Dict, TypeVar, cast
 
@@ -7,10 +8,16 @@ from dotenv import load_dotenv
 
 load_dotenv()
 
+logging.basicConfig(level=logging.WARNING)
+
 T = TypeVar("T", bound=type)
 
+"""Base decorator for creating crew classes with configuration and function management."""
+
 
 def CrewBase(cls: T) -> T:
+    """Wraps a class with crew functionality and configuration management."""
+
     class WrappedClass(cls):  # type: ignore
         is_crew_class: bool = True  # type: ignore
 
@@ -24,16 +31,9 @@ def CrewBase(cls: T) -> T:
 
         def __init__(self, *args, **kwargs):
             super().__init__(*args, **kwargs)
-
-            agents_config_path = self.base_directory / self.original_agents_config_path
-            tasks_config_path = self.base_directory / self.original_tasks_config_path
-
-            self.agents_config = self.load_yaml(agents_config_path)
-            self.tasks_config = self.load_yaml(tasks_config_path)
-
+            self.load_configurations()
             self.map_all_agent_variables()
             self.map_all_task_variables()
-
             # Preserve all decorated functions
             self._original_functions = {
                 name: method
@@ -49,7 +49,6 @@ def CrewBase(cls: T) -> T:
                     ]
                 )
             }
-
             # Store specific function types
             self._original_tasks = self._filter_functions(
                 self._original_functions, "is_task"
@@ -67,6 +66,44 @@ def CrewBase(cls: T) -> T:
                 self._original_functions, "is_kickoff"
             )
 
+        def load_configurations(self):
+            """Load agent and task configurations from YAML files."""
+            if isinstance(self.original_agents_config_path, str):
+                agents_config_path = (
+                    self.base_directory / self.original_agents_config_path
+                )
+                try:
+                    self.agents_config = self.load_yaml(agents_config_path)
+                except FileNotFoundError:
+                    logging.warning(
+                        f"Agent config file not found at {agents_config_path}. "
+                        "Proceeding with empty agent configurations."
+                    )
+                    self.agents_config = {}
+            else:
+                logging.warning(
+                    "No agent configuration path provided. Proceeding with empty agent configurations."
+                )
+                self.agents_config = {}
+
+            if isinstance(self.original_tasks_config_path, str):
+                tasks_config_path = (
+                    self.base_directory / self.original_tasks_config_path
+                )
+                try:
+                    self.tasks_config = self.load_yaml(tasks_config_path)
+                except FileNotFoundError:
+                    logging.warning(
+                        f"Task config file not found at {tasks_config_path}. "
+                        "Proceeding with empty task configurations."
+                    )
+                    self.tasks_config = {}
+            else:
+                logging.warning(
+                    "No task configuration path provided. Proceeding with empty task configurations."
+                )
+                self.tasks_config = {}
+
         @staticmethod
         def load_yaml(config_path: Path):
             try:
@@ -213,4 +250,8 @@ def CrewBase(cls: T) -> T:
                     callback_functions[callback]() for callback in callbacks
                 ]
 
+    # Include base class (qual)name in the wrapper class (qual)name.
+    WrappedClass.__name__ = CrewBase.__name__ + "(" + cls.__name__ + ")"
+    WrappedClass.__qualname__ = CrewBase.__qualname__ + "(" + cls.__name__ + ")"
+
     return cast(T, WrappedClass)

src/crewai/project/utils.py

@@ -1,11 +1,14 @@
+from functools import wraps
+
+
 def memoize(func):
     cache = {}
 
+    @wraps(func)
     def memoized_func(*args, **kwargs):
         key = (args, tuple(kwargs.items()))
         if key not in cache:
             cache[key] = func(*args, **kwargs)
         return cache[key]
 
-    memoized_func.__dict__.update(func.__dict__)
     return memoized_func

src/crewai/task.py

@@ -1,12 +1,25 @@
 import datetime
+import inspect
 import json
-import os
+import logging
 import threading
 import uuid
 from concurrent.futures import Future
 from copy import copy
 from hashlib import md5
-from typing import Any, Dict, List, Optional, Set, Tuple, Type, Union
+from pathlib import Path
+from typing import (
+    Any,
+    Callable,
+    ClassVar,
+    Dict,
+    List,
+    Optional,
+    Set,
+    Tuple,
+    Type,
+    Union,
+)
 
 from opentelemetry.trace import Span
 from pydantic import (
@@ -20,6 +33,7 @@ from pydantic import (
 from pydantic_core import PydanticCustomError
 
 from crewai.agents.agent_builder.base_agent import BaseAgent
+from crewai.tasks.guardrail_result import GuardrailResult
 from crewai.tasks.output_format import OutputFormat
 from crewai.tasks.task_output import TaskOutput
 from crewai.telemetry.telemetry import Telemetry
@@ -27,6 +41,7 @@ from crewai.tools.base_tool import BaseTool
 from crewai.utilities.config import process_config
 from crewai.utilities.converter import Converter, convert_to_model
 from crewai.utilities.i18n import I18N
+from crewai.utilities.printer import Printer
 
 
 class Task(BaseModel):
@@ -49,6 +64,7 @@ class Task(BaseModel):
     """
 
     __hash__ = object.__hash__  # type: ignore
+    logger: ClassVar[logging.Logger] = logging.getLogger(__name__)
     used_tools: int = 0
     tools_errors: int = 0
     delegations: int = 0
@@ -110,13 +126,69 @@ class Task(BaseModel):
         default=None,
     )
     processed_by_agents: Set[str] = Field(default_factory=set)
+    guardrail: Optional[Callable[[TaskOutput], Tuple[bool, Any]]] = Field(
+        default=None,
+        description="Function to validate task output before proceeding to next task",
+    )
+    max_retries: int = Field(
+        default=3, description="Maximum number of retries when guardrail fails"
+    )
+    retry_count: int = Field(default=0, description="Current number of retries")
+    start_time: Optional[datetime.datetime] = Field(
+        default=None, description="Start time of the task execution"
+    )
+    end_time: Optional[datetime.datetime] = Field(
+        default=None, description="End time of the task execution"
+    )
+
+    @field_validator("guardrail")
+    @classmethod
+    def validate_guardrail_function(cls, v: Optional[Callable]) -> Optional[Callable]:
+        """Validate that the guardrail function has the correct signature and behavior.
+
+        While type hints provide static checking, this validator ensures runtime safety by:
+        1. Verifying the function accepts exactly one parameter (the TaskOutput)
+        2. Checking return type annotations match Tuple[bool, Any] if present
+        3. Providing clear, immediate error messages for debugging
+
+        This runtime validation is crucial because:
+        - Type hints are optional and can be ignored at runtime
+        - Function signatures need immediate validation before task execution
+        - Clear error messages help users debug guardrail implementation issues
+
+        Args:
+            v: The guardrail function to validate
+
+        Returns:
+            The validated guardrail function
+
+        Raises:
+            ValueError: If the function signature is invalid or return annotation
+                       doesn't match Tuple[bool, Any]
+        """
+        if v is not None:
+            sig = inspect.signature(v)
+            if len(sig.parameters) != 1:
+                raise ValueError("Guardrail function must accept exactly one parameter")
+
+            # Check return annotation if present, but don't require it
+            return_annotation = sig.return_annotation
+            if return_annotation != inspect.Signature.empty:
+                if not (
+                    return_annotation == Tuple[bool, Any]
+                    or str(return_annotation) == "Tuple[bool, Any]"
+                ):
+                    raise ValueError(
+                        "If return type is annotated, it must be Tuple[bool, Any]"
+                    )
+        return v
 
     _telemetry: Telemetry = PrivateAttr(default_factory=Telemetry)
     _execution_span: Optional[Span] = PrivateAttr(default=None)
     _original_description: Optional[str] = PrivateAttr(default=None)
     _original_expected_output: Optional[str] = PrivateAttr(default=None)
+    _original_output_file: Optional[str] = PrivateAttr(default=None)
     _thread: Optional[threading.Thread] = PrivateAttr(default=None)
-    _execution_time: Optional[float] = PrivateAttr(default=None)
 
     @model_validator(mode="before")
     @classmethod
@@ -141,16 +213,54 @@ class Task(BaseModel):
                 "may_not_set_field", "This field is not to be set by the user.", {}
             )
 
-    def _set_start_execution_time(self) -> float:
-        return datetime.datetime.now().timestamp()
-
-    def _set_end_execution_time(self, start_time: float) -> None:
-        self._execution_time = datetime.datetime.now().timestamp() - start_time
-
     @field_validator("output_file")
     @classmethod
-    def output_file_validation(cls, value: str) -> str:
-        """Validate the output file path by removing the / from the beginning of the path."""
+    def output_file_validation(cls, value: Optional[str]) -> Optional[str]:
+        """Validate the output file path.
+
+        Args:
+            value: The output file path to validate. Can be None or a string.
+                  If the path contains template variables (e.g. {var}), leading slashes are preserved.
+                  For regular paths, leading slashes are stripped.
+
+        Returns:
+            The validated and potentially modified path, or None if no path was provided.
+
+        Raises:
+            ValueError: If the path contains invalid characters, path traversal attempts,
+                      or other security concerns.
+        """
+        if value is None:
+            return None
+
+        # Basic security checks
+        if ".." in value:
+            raise ValueError(
+                "Path traversal attempts are not allowed in output_file paths"
+            )
+
+        # Check for shell expansion first
+        if value.startswith("~") or value.startswith("$"):
+            raise ValueError(
+                "Shell expansion characters are not allowed in output_file paths"
+            )
+
+        # Then check other shell special characters
+        if any(char in value for char in ["|", ">", "<", "&", ";"]):
+            raise ValueError(
+                "Shell special characters are not allowed in output_file paths"
+            )
+
+        # Don't strip leading slash if it's a template path with variables
+        if "{" in value or "}" in value:
+            # Validate template variable format
+            template_vars = [part.split("}")[0] for part in value.split("{")[1:]]
+            for var in template_vars:
+                if not var.isidentifier():
+                    raise ValueError(f"Invalid template variable name: {var}")
+            return value
+
+        # Strip leading slash for regular paths
         if value.startswith("/"):
             return value[1:]
         return value
@@ -199,6 +309,12 @@ class Task(BaseModel):
 
         return md5("|".join(source).encode(), usedforsecurity=False).hexdigest()
 
+    @property
+    def execution_duration(self) -> float | None:
+        if not self.start_time or not self.end_time:
+            return None
+        return (self.end_time - self.start_time).total_seconds()
+
     def execute_async(
         self,
         agent: BaseAgent | None = None,
@@ -239,7 +355,7 @@ class Task(BaseModel):
                 f"The task '{self.description}' has no agent assigned, therefore it can't be executed directly and should be executed in a Crew using a specific process that support that, like hierarchical."
             )
 
-        start_time = self._set_start_execution_time()
+        self.start_time = datetime.datetime.now()
         self._execution_span = self._telemetry.task_started(crew=agent.crew, task=self)
 
         self.prompt_context = context
@@ -254,7 +370,6 @@ class Task(BaseModel):
         )
 
         pydantic_output, json_output = self._export_output(result)
-
         task_output = TaskOutput(
             name=self.name,
             description=self.description,
@@ -265,9 +380,46 @@ class Task(BaseModel):
             agent=agent.role,
             output_format=self._get_output_format(),
         )
-        self.output = task_output
 
-        self._set_end_execution_time(start_time)
+        if self.guardrail:
+            guardrail_result = GuardrailResult.from_tuple(self.guardrail(task_output))
+            if not guardrail_result.success:
+                if self.retry_count >= self.max_retries:
+                    raise Exception(
+                        f"Task failed guardrail validation after {self.max_retries} retries. "
+                        f"Last error: {guardrail_result.error}"
+                    )
+
+                self.retry_count += 1
+                context = self.i18n.errors("validation_error").format(
+                    guardrail_result_error=guardrail_result.error,
+                    task_output=task_output.raw,
+                )
+                printer = Printer()
+                printer.print(
+                    content=f"Guardrail blocked, retrying, due to: {guardrail_result.error}\n",
+                    color="yellow",
+                )
+                return self._execute_core(agent, context, tools)
+
+            if guardrail_result.result is None:
+                raise Exception(
+                    "Task guardrail returned None as result. This is not allowed."
+                )
+
+            if isinstance(guardrail_result.result, str):
+                task_output.raw = guardrail_result.result
+                pydantic_output, json_output = self._export_output(
+                    guardrail_result.result
+                )
+                task_output.pydantic = pydantic_output
+                task_output.json_dict = json_output
+            elif isinstance(guardrail_result.result, TaskOutput):
+                task_output = guardrail_result.result
+
+        self.output = task_output
+        self.end_time = datetime.datetime.now()
+
         if self.callback:
             self.callback(self.output)
 
@@ -299,16 +451,127 @@ class Task(BaseModel):
         tasks_slices = [self.description, output]
         return "\n".join(tasks_slices)
 
-    def interpolate_inputs(self, inputs: Dict[str, Any]) -> None:
-        """Interpolate inputs into the task description and expected output."""
+    def interpolate_inputs_and_add_conversation_history(
+        self, inputs: Dict[str, Union[str, int, float]]
+    ) -> None:
+        """Interpolate inputs into the task description, expected output, and output file path.
+           Add conversation history if present.
+
+        Args:
+            inputs: Dictionary mapping template variables to their values.
+                   Supported value types are strings, integers, and floats.
+
+        Raises:
+            ValueError: If a required template variable is missing from inputs.
+        """
         if self._original_description is None:
             self._original_description = self.description
         if self._original_expected_output is None:
             self._original_expected_output = self.expected_output
+        if self.output_file is not None and self._original_output_file is None:
+            self._original_output_file = self.output_file
 
-        if inputs:
+        if not inputs:
+            return
+
+        try:
             self.description = self._original_description.format(**inputs)
-            self.expected_output = self._original_expected_output.format(**inputs)
+        except KeyError as e:
+            raise ValueError(
+                f"Missing required template variable '{e.args[0]}' in description"
+            ) from e
+        except ValueError as e:
+            raise ValueError(f"Error interpolating description: {str(e)}") from e
+
+        try:
+            self.expected_output = self.interpolate_only(
+                input_string=self._original_expected_output, inputs=inputs
+            )
+        except (KeyError, ValueError) as e:
+            raise ValueError(f"Error interpolating expected_output: {str(e)}") from e
+
+        if self.output_file is not None:
+            try:
+                self.output_file = self.interpolate_only(
+                    input_string=self._original_output_file, inputs=inputs
+                )
+            except (KeyError, ValueError) as e:
+                raise ValueError(
+                    f"Error interpolating output_file path: {str(e)}"
+                ) from e
+
+        if "crew_chat_messages" in inputs and inputs["crew_chat_messages"]:
+            conversation_instruction = self.i18n.slice(
+                "conversation_history_instruction"
+            )
+
+            crew_chat_messages_json = str(inputs["crew_chat_messages"])
+
+            try:
+                crew_chat_messages = json.loads(crew_chat_messages_json)
+            except json.JSONDecodeError as e:
+                print("An error occurred while parsing crew chat messages:", e)
+                raise
+
+            conversation_history = "\n".join(
+                f"{msg['role'].capitalize()}: {msg['content']}"
+                for msg in crew_chat_messages
+                if isinstance(msg, dict) and "role" in msg and "content" in msg
+            )
+
+            self.description += (
+                f"\n\n{conversation_instruction}\n\n{conversation_history}"
+            )
+
+    def interpolate_only(
+        self, input_string: Optional[str], inputs: Dict[str, Union[str, int, float]]
+    ) -> str:
+        """Interpolate placeholders (e.g., {key}) in a string while leaving JSON untouched.
+
+        Args:
+            input_string: The string containing template variables to interpolate.
+                         Can be None or empty, in which case an empty string is returned.
+            inputs: Dictionary mapping template variables to their values.
+                   Supported value types are strings, integers, and floats.
+                   If input_string is empty or has no placeholders, inputs can be empty.
+
+        Returns:
+            The interpolated string with all template variables replaced with their values.
+            Empty string if input_string is None or empty.
+
+        Raises:
+            ValueError: If a required template variable is missing from inputs.
+            KeyError: If a template variable is not found in the inputs dictionary.
+        """
+        if input_string is None or not input_string:
+            return ""
+        if "{" not in input_string and "}" not in input_string:
+            return input_string
+        if not inputs:
+            raise ValueError(
+                "Inputs dictionary cannot be empty when interpolating variables"
+            )
+
+        try:
+            # Validate input types
+            for key, value in inputs.items():
+                if not isinstance(value, (str, int, float)):
+                    raise ValueError(
+                        f"Value for key '{key}' must be a string, integer, or float, got {type(value).__name__}"
+                    )
+
+            escaped_string = input_string.replace("{", "{{").replace("}", "}}")
+
+            for key in inputs.keys():
+                escaped_string = escaped_string.replace(f"{{{{{key}}}}}", f"{{{key}}}")
+
+            return escaped_string.format(**inputs)
+        except KeyError as e:
+            raise KeyError(
+                f"Template variable '{e.args[0]}' not found in inputs dictionary"
+            ) from e
+        except ValueError as e:
+            raise ValueError(f"Error during string interpolation: {str(e)}") from e
 
     def increment_tools_errors(self) -> None:
         """Increment the tools errors counter."""
@@ -390,21 +653,34 @@ class Task(BaseModel):
         return OutputFormat.RAW
 
     def _save_file(self, result: Any) -> None:
+        """Save task output to a file.
+
+        Args:
+            result: The result to save to the file. Can be a dict or any stringifiable object.
+
+        Raises:
+            ValueError: If output_file is not set
+            RuntimeError: If there is an error writing to the file
+        """
         if self.output_file is None:
             raise ValueError("output_file is not set.")
 
-        directory = os.path.dirname(self.output_file)  # type: ignore # Value of type variable "AnyOrLiteralStr" of "dirname" cannot be "str | None"
+        try:
+            resolved_path = Path(self.output_file).expanduser().resolve()
+            directory = resolved_path.parent
 
-        if directory and not os.path.exists(directory):
-            os.makedirs(directory)
+            if not directory.exists():
+                directory.mkdir(parents=True, exist_ok=True)
 
-        with open(self.output_file, "w", encoding="utf-8") as file:
-            if isinstance(result, dict):
-                import json
+            with resolved_path.open("w", encoding="utf-8") as file:
+                if isinstance(result, dict):
+                    import json
 
-                json.dump(result, file, ensure_ascii=False, indent=2)
-            else:
-                file.write(str(result))
+                    json.dump(result, file, ensure_ascii=False, indent=2)
+                else:
+                    file.write(str(result))
+        except (OSError, IOError) as e:
+            raise RuntimeError(f"Failed to save output file: {e}")
         return None
 
     def __repr__(self):

src/crewai/tasks/guardrail_result.py

@@ -0,0 +1,56 @@
+"""
+Module for handling task guardrail validation results.
+
+This module provides the GuardrailResult class which standardizes
+the way task guardrails return their validation results.
+"""
+
+from typing import Any, Optional, Tuple, Union
+
+from pydantic import BaseModel, field_validator
+
+
+class GuardrailResult(BaseModel):
+    """Result from a task guardrail execution.
+
+    This class standardizes the return format of task guardrails,
+    converting tuple responses into a structured format that can
+    be easily handled by the task execution system.
+
+    Attributes:
+        success (bool): Whether the guardrail validation passed
+        result (Any, optional): The validated/transformed result if successful
+        error (str, optional): Error message if validation failed
+    """
+    success: bool
+    result: Optional[Any] = None
+    error: Optional[str] = None
+
+    @field_validator("result", "error")
+    @classmethod
+    def validate_result_error_exclusivity(cls, v: Any, info) -> Any:
+        values = info.data
+        if "success" in values:
+            if values["success"] and v and "error" in values and values["error"]:
+                raise ValueError("Cannot have both result and error when success is True")
+            if not values["success"] and v and "result" in values and values["result"]:
+                raise ValueError("Cannot have both result and error when success is False")
+        return v
+
+    @classmethod
+    def from_tuple(cls, result: Tuple[bool, Union[Any, str]]) -> "GuardrailResult":
+        """Create a GuardrailResult from a validation tuple.
+
+        Args:
+            result: A tuple of (success, data) where data is either
+                   the validated result or error message.
+
+        Returns:
+            GuardrailResult: A new instance with the tuple data.
+        """
+        success, data = result
+        return cls(
+            success=success,
+            result=data if success else None,
+            error=data if not success else None
+        )

src/crewai/telemetry/telemetry.py

@@ -6,6 +6,7 @@ import os
 import platform
 import warnings
 from contextlib import contextmanager
+from importlib.metadata import version
 from typing import TYPE_CHECKING, Any, Optional
 
 
@@ -16,12 +17,10 @@ def suppress_warnings():
         yield
 
 
-with suppress_warnings():
-    import pkg_resources
-
-
 from opentelemetry import trace  # noqa: E402
-from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter # noqa: E402
+from opentelemetry.exporter.otlp.proto.http.trace_exporter import (
+    OTLPSpanExporter,  # noqa: E402
+)
 from opentelemetry.sdk.resources import SERVICE_NAME, Resource  # noqa: E402
 from opentelemetry.sdk.trace import TracerProvider  # noqa: E402
 from opentelemetry.sdk.trace.export import BatchSpanProcessor  # noqa: E402
@@ -104,7 +103,7 @@ class Telemetry:
             self._add_attribute(
                 span,
                 "crewai_version",
-                pkg_resources.get_distribution("crewai").version,
+                version("crewai"),
             )
             self._add_attribute(span, "python_version", platform.python_version())
             self._add_attribute(span, "crew_key", crew.key)
@@ -306,7 +305,7 @@ class Telemetry:
             self._add_attribute(
                 span,
                 "crewai_version",
-                pkg_resources.get_distribution("crewai").version,
+                version("crewai"),
             )
             self._add_attribute(span, "tool_name", tool_name)
             self._add_attribute(span, "attempts", attempts)
@@ -326,7 +325,7 @@ class Telemetry:
             self._add_attribute(
                 span,
                 "crewai_version",
-                pkg_resources.get_distribution("crewai").version,
+                version("crewai"),
             )
             self._add_attribute(span, "tool_name", tool_name)
             self._add_attribute(span, "attempts", attempts)
@@ -346,7 +345,7 @@ class Telemetry:
             self._add_attribute(
                 span,
                 "crewai_version",
-                pkg_resources.get_distribution("crewai").version,
+                version("crewai"),
             )
             if llm:
                 self._add_attribute(span, "llm", llm.model)
@@ -365,7 +364,7 @@ class Telemetry:
             self._add_attribute(
                 span,
                 "crewai_version",
-                pkg_resources.get_distribution("crewai").version,
+                version("crewai"),
             )
             self._add_attribute(span, "crew_key", crew.key)
             self._add_attribute(span, "crew_id", str(crew.id))
@@ -391,7 +390,7 @@ class Telemetry:
             self._add_attribute(
                 span,
                 "crewai_version",
-                pkg_resources.get_distribution("crewai").version,
+                version("crewai"),
             )
             self._add_attribute(span, "crew_key", crew.key)
             self._add_attribute(span, "crew_id", str(crew.id))
@@ -472,7 +471,7 @@ class Telemetry:
             self._add_attribute(
                 span,
                 "crewai_version",
-                pkg_resources.get_distribution("crewai").version,
+                version("crewai"),
             )
             self._add_attribute(span, "crew_key", crew.key)
             self._add_attribute(span, "crew_id", str(crew.id))
@@ -541,7 +540,7 @@ class Telemetry:
             self._add_attribute(
                 crew._execution_span,
                 "crewai_version",
-                pkg_resources.get_distribution("crewai").version,
+                version("crewai"),
             )
             self._add_attribute(
                 crew._execution_span, "crew_output", final_string_output

src/crewai/tools/agent_tools/add_image_tool.py

@@ -0,0 +1,45 @@
+from typing import Dict, Optional, Union
+
+from pydantic import BaseModel, Field
+
+from crewai.tools.base_tool import BaseTool
+from crewai.utilities import I18N
+
+i18n = I18N()
+
+class AddImageToolSchema(BaseModel):
+    image_url: str = Field(..., description="The URL or path of the image to add")
+    action: Optional[str] = Field(
+        default=None,
+        description="Optional context or question about the image"
+    )
+
+
+class AddImageTool(BaseTool):
+    """Tool for adding images to the content"""
+
+    name: str = Field(default_factory=lambda: i18n.tools("add_image")["name"])  # type: ignore
+    description: str = Field(default_factory=lambda: i18n.tools("add_image")["description"])  # type: ignore
+    args_schema: type[BaseModel] = AddImageToolSchema
+
+    def _run(
+        self,
+        image_url: str,
+        action: Optional[str] = None,
+        **kwargs,
+    ) -> dict:
+        action = action or i18n.tools("add_image")["default_action"]  # type: ignore
+        content = [
+            {"type": "text", "text": action},
+            {
+                "type": "image_url",
+                "image_url": {
+                    "url": image_url,
+                },
+            }
+        ]
+
+        return {
+            "role": "user",
+            "content": content
+        }

src/crewai/tools/agent_tools/agent_tools.py

@@ -1,9 +1,9 @@
-from crewai.tools.base_tool import BaseTool
 from crewai.agents.agent_builder.base_agent import BaseAgent
+from crewai.tools.base_tool import BaseTool
 from crewai.utilities import I18N
 
-from .delegate_work_tool import DelegateWorkTool
 from .ask_question_tool import AskQuestionTool
+from .delegate_work_tool import DelegateWorkTool
 
 
 class AgentTools:
@@ -20,13 +20,13 @@ class AgentTools:
         delegate_tool = DelegateWorkTool(
             agents=self.agents,
             i18n=self.i18n,
-            description=self.i18n.tools("delegate_work").format(coworkers=coworkers),
+            description=self.i18n.tools("delegate_work").format(coworkers=coworkers),  # type: ignore
         )
 
         ask_tool = AskQuestionTool(
             agents=self.agents,
             i18n=self.i18n,
-            description=self.i18n.tools("ask_question").format(coworkers=coworkers),
+            description=self.i18n.tools("ask_question").format(coworkers=coworkers),  # type: ignore
         )
 
         return [delegate_tool, ask_tool]

src/crewai/tools/agent_tools/ask_question_tool.py

@@ -1,7 +1,9 @@
-from crewai.tools.agent_tools.base_agent_tools import BaseAgentTool
 from typing import Optional
+
 from pydantic import BaseModel, Field
 
+from crewai.tools.agent_tools.base_agent_tools import BaseAgentTool
+
 
 class AskQuestionToolSchema(BaseModel):
     question: str = Field(..., description="The question to ask")

src/crewai/tools/agent_tools/base_agent_tools.py

@@ -1,11 +1,15 @@
-from typing import Optional, Union
+import logging
+from typing import Optional
+
 from pydantic import Field
 
-from crewai.tools.base_tool import BaseTool
 from crewai.agents.agent_builder.base_agent import BaseAgent
 from crewai.task import Task
+from crewai.tools.base_tool import BaseTool
 from crewai.utilities import I18N
 
+logger = logging.getLogger(__name__)
+
 
 class BaseAgentTool(BaseTool):
     """Base class for agent-related tools"""
@@ -15,6 +19,25 @@ class BaseAgentTool(BaseTool):
         default_factory=I18N, description="Internationalization settings"
     )
 
+    def sanitize_agent_name(self, name: str) -> str:
+        """
+        Sanitize agent role name by normalizing whitespace and setting to lowercase.
+        Converts all whitespace (including newlines) to single spaces and removes quotes.
+
+        Args:
+            name (str): The agent role name to sanitize
+
+        Returns:
+            str: The sanitized agent role name, with whitespace normalized,
+                 converted to lowercase, and quotes removed
+        """
+        if not name:
+            return ""
+        # Normalize all whitespace (including newlines) to single spaces
+        normalized = " ".join(name.split())
+        # Remove quotes and convert to lowercase
+        return normalized.replace('"', "").casefold()
+
     def _get_coworker(self, coworker: Optional[str], **kwargs) -> Optional[str]:
         coworker = coworker or kwargs.get("co_worker") or kwargs.get("coworker")
         if coworker:
@@ -24,11 +47,27 @@ class BaseAgentTool(BaseTool):
         return coworker
 
     def _execute(
-        self, agent_name: Union[str, None], task: str, context: Union[str, None]
+        self,
+        agent_name: Optional[str],
+        task: str,
+        context: Optional[str] = None
     ) -> str:
+        """
+        Execute delegation to an agent with case-insensitive and whitespace-tolerant matching.
+
+        Args:
+            agent_name: Name/role of the agent to delegate to (case-insensitive)
+            task: The specific question or task to delegate
+            context: Optional additional context for the task execution
+
+        Returns:
+            str: The execution result from the delegated agent or an error message
+                 if the agent cannot be found
+        """
         try:
             if agent_name is None:
                 agent_name = ""
+                logger.debug("No agent name provided, using empty string")
 
             # It is important to remove the quotes from the agent name.
             # The reason we have to do this is because less-powerful LLM's
@@ -37,31 +76,49 @@ class BaseAgentTool(BaseTool):
             # {"task": "....", "coworker": "....
             # when it should look like this:
             # {"task": "....", "coworker": "...."}
-            agent_name = agent_name.casefold().replace('"', "").replace("\n", "")
+            sanitized_name = self.sanitize_agent_name(agent_name)
+            logger.debug(f"Sanitized agent name from '{agent_name}' to '{sanitized_name}'")
+
+            available_agents = [agent.role for agent in self.agents]
+            logger.debug(f"Available agents: {available_agents}")
+
             agent = [  # type: ignore # Incompatible types in assignment (expression has type "list[BaseAgent]", variable has type "str | None")
                 available_agent
                 for available_agent in self.agents
-                if available_agent.role.casefold().replace("\n", "") == agent_name
+                if self.sanitize_agent_name(available_agent.role) == sanitized_name
             ]
-        except Exception as _:
-            return self.i18n.errors("agent_tool_unexsiting_coworker").format(
+            logger.debug(f"Found {len(agent)} matching agents for role '{sanitized_name}'")
+        except (AttributeError, ValueError) as e:
+            # Handle specific exceptions that might occur during role name processing
+            return self.i18n.errors("agent_tool_unexisting_coworker").format(
                 coworkers="\n".join(
-                    [f"- {agent.role.casefold()}" for agent in self.agents]
-                )
+                    [f"- {self.sanitize_agent_name(agent.role)}" for agent in self.agents]
+                ),
+                error=str(e)
             )
 
         if not agent:
-            return self.i18n.errors("agent_tool_unexsiting_coworker").format(
+            # No matching agent found after sanitization
+            return self.i18n.errors("agent_tool_unexisting_coworker").format(
                 coworkers="\n".join(
-                    [f"- {agent.role.casefold()}" for agent in self.agents]
-                )
+                    [f"- {self.sanitize_agent_name(agent.role)}" for agent in self.agents]
+                ),
+                error=f"No agent found with role '{sanitized_name}'"
             )
 
         agent = agent[0]
-        task_with_assigned_agent = Task(  # type: ignore # Incompatible types in assignment (expression has type "Task", variable has type "str")
-            description=task,
-            agent=agent,
-            expected_output=agent.i18n.slice("manager_request"),
-            i18n=agent.i18n,
-        )
-        return agent.execute_task(task_with_assigned_agent, context)
+        try:
+            task_with_assigned_agent = Task(
+                description=task,
+                agent=agent,
+                expected_output=agent.i18n.slice("manager_request"),
+                i18n=agent.i18n,
+            )
+            logger.debug(f"Created task for agent '{self.sanitize_agent_name(agent.role)}': {task}")
+            return agent.execute_task(task_with_assigned_agent, context)
+        except Exception as e:
+            # Handle task creation or execution errors
+            return self.i18n.errors("agent_tool_execution_error").format(
+                agent_role=self.sanitize_agent_name(agent.role),
+                error=str(e)
+            )

src/crewai/tools/agent_tools/delegate_work_tool.py

@@ -1,8 +1,9 @@
-from crewai.tools.agent_tools.base_agent_tools import BaseAgentTool
 from typing import Optional
 
 from pydantic import BaseModel, Field
 
+from crewai.tools.agent_tools.base_agent_tools import BaseAgentTool
+
 
 class DelegateWorkToolSchema(BaseModel):
     task: str = Field(..., description="The task to delegate")

src/crewai/tools/base_tool.py

@@ -1,12 +1,23 @@
+import warnings
 from abc import ABC, abstractmethod
 from inspect import signature
 from typing import Any, Callable, Type, get_args, get_origin
 
-from pydantic import BaseModel, ConfigDict, Field, create_model, validator
+from pydantic import (
+    BaseModel,
+    ConfigDict,
+    Field,
+    PydanticDeprecatedSince20,
+    create_model,
+    validator,
+)
 from pydantic import BaseModel as PydanticBaseModel
 
 from crewai.tools.structured_tool import CrewStructuredTool
 
+# Ignore all "PydanticDeprecatedSince20" warnings globally
+warnings.filterwarnings("ignore", category=PydanticDeprecatedSince20)
+
 
 class BaseTool(BaseModel, ABC):
     class _ArgsSchemaPlaceholder(PydanticBaseModel):

src/crewai/tools/tool_usage.py

@@ -1,28 +1,37 @@
 import ast
 import datetime
-import os
+import json
+import re
 import time
 from difflib import SequenceMatcher
 from textwrap import dedent
-from typing import Any, List, Union
+from typing import Any, Dict, List, Union
+
+from json_repair import repair_json
 
 import crewai.utilities.events as events
 from crewai.agents.tools_handler import ToolsHandler
 from crewai.task import Task
 from crewai.telemetry import Telemetry
 from crewai.tools import BaseTool
+from crewai.tools.structured_tool import CrewStructuredTool
 from crewai.tools.tool_calling import InstructorToolCalling, ToolCalling
 from crewai.tools.tool_usage_events import ToolUsageError, ToolUsageFinished
 from crewai.utilities import I18N, Converter, ConverterError, Printer
 
-agentops = None
-if os.environ.get("AGENTOPS_API_KEY"):
-    try:
-        import agentops  # type: ignore
-    except ImportError:
-        pass
-
-OPENAI_BIGGER_MODELS = ["gpt-4", "gpt-4o", "o1-preview", "o1-mini"]
+try:
+    import agentops  # type: ignore
+except ImportError:
+    agentops = None
+OPENAI_BIGGER_MODELS = [
+    "gpt-4",
+    "gpt-4o",
+    "o1-preview",
+    "o1-mini",
+    "o1",
+    "o3",
+    "o3-mini",
+]
 
 
 class ToolUsageErrorException(Exception):
@@ -83,7 +92,7 @@ class ToolUsage:
             self._max_parsing_attempts = 2
             self._remember_format_after_usages = 4
 
-    def parse(self, tool_string: str):
+    def parse_tool_calling(self, tool_string: str):
         """Parse the tool string and return the tool calling."""
         return self._tool_calling(tool_string)
 
@@ -97,7 +106,6 @@ class ToolUsage:
             self.task.increment_tools_errors()
             return error
 
-        # BUG? The code below seems to be unreachable
         try:
             tool = self._select_tool(calling.tool_name)
         except Exception as e:
@@ -106,7 +114,20 @@ class ToolUsage:
             if self.agent.verbose:
                 self._printer.print(content=f"\n\n{error}\n", color="red")
             return error
-        return f"{self._use(tool_string=tool_string, tool=tool, calling=calling)}"  # type: ignore # BUG?: "_use" of "ToolUsage" does not return a value (it only ever returns None)
+
+        if isinstance(tool, CrewStructuredTool) and tool.name == self._i18n.tools("add_image")["name"]:  # type: ignore
+            try:
+                result = self._use(tool_string=tool_string, tool=tool, calling=calling)
+                return result
+
+            except Exception as e:
+                error = getattr(e, "message", str(e))
+                self.task.increment_tools_errors()
+                if self.agent.verbose:
+                    self._printer.print(content=f"\n\n{error}\n", color="red")
+                return error
+
+        return f"{self._use(tool_string=tool_string, tool=tool, calling=calling)}"
 
     def _use(
         self,
@@ -159,7 +180,7 @@ class ToolUsage:
 
                 if calling.arguments:
                     try:
-                        acceptable_args = tool.args_schema.schema()["properties"].keys()  # type: ignore # Item "None" of "type[BaseModel] | None" has no attribute "schema"
+                        acceptable_args = tool.args_schema.model_json_schema()["properties"].keys()  # type: ignore
                         arguments = {
                             k: v
                             for k, v in calling.arguments.items()
@@ -339,13 +360,13 @@ class ToolUsage:
         tool_name = self.action.tool
         tool = self._select_tool(tool_name)
         try:
-            tool_input = self._validate_tool_input(self.action.tool_input)
-            arguments = ast.literal_eval(tool_input)
+            arguments = self._validate_tool_input(self.action.tool_input)
+
         except Exception:
             if raise_error:
                 raise
             else:
-                return ToolUsageErrorException(  # type: ignore # Incompatible return value type (got "ToolUsageErrorException", expected "ToolCalling | InstructorToolCalling")
+                return ToolUsageErrorException(
                     f'{self._i18n.errors("tool_arguments_error")}'
                 )
 
@@ -353,14 +374,14 @@ class ToolUsage:
             if raise_error:
                 raise
             else:
-                return ToolUsageErrorException(  # type: ignore # Incompatible return value type (got "ToolUsageErrorException", expected "ToolCalling | InstructorToolCalling")
+                return ToolUsageErrorException(
                     f'{self._i18n.errors("tool_arguments_error")}'
                 )
 
         return ToolCalling(
             tool_name=tool.name,
             arguments=arguments,
-            log=tool_string,  # type: ignore
+            log=tool_string,
         )
 
     def _tool_calling(
@@ -386,56 +407,28 @@ class ToolUsage:
                 )
             return self._tool_calling(tool_string)
 
-    def _validate_tool_input(self, tool_input: str) -> str:
+    def _validate_tool_input(self, tool_input: str) -> Dict[str, Any]:
         try:
-            ast.literal_eval(tool_input)
-            return tool_input
-        except Exception:
-            # Clean and ensure the string is properly enclosed in braces
-            tool_input = tool_input.strip()
-            if not tool_input.startswith("{"):
-                tool_input = "{" + tool_input
-            if not tool_input.endswith("}"):
-                tool_input += "}"
+            # Replace Python literals with JSON equivalents
+            replacements = {
+                r"'": '"',
+                r"None": "null",
+                r"True": "true",
+                r"False": "false",
+            }
+            for pattern, replacement in replacements.items():
+                tool_input = re.sub(pattern, replacement, tool_input)
 
-            # Manually split the input into key-value pairs
-            entries = tool_input.strip("{} ").split(",")
-            formatted_entries = []
+            arguments = json.loads(tool_input)
+        except json.JSONDecodeError:
+            # Attempt to repair JSON string
+            repaired_input = repair_json(tool_input)
+            try:
+                arguments = json.loads(repaired_input)
+            except json.JSONDecodeError as e:
+                raise Exception(f"Invalid tool input JSON: {e}")
 
-            for entry in entries:
-                if ":" not in entry:
-                    continue  # Skip malformed entries
-                key, value = entry.split(":", 1)
-
-                # Remove extraneous white spaces and quotes, replace single quotes
-                key = key.strip().strip('"').replace("'", '"')
-                value = value.strip()
-
-                # Handle replacement of single quotes at the start and end of the value string
-                if value.startswith("'") and value.endswith("'"):
-                    value = value[1:-1]  # Remove single quotes
-                    value = (
-                        '"' + value.replace('"', '\\"') + '"'
-                    )  # Re-encapsulate with double quotes
-                elif value.isdigit():  # Check if value is a digit, hence integer
-                    value = value
-                elif value.lower() in [
-                    "true",
-                    "false",
-                    "null",
-                ]:  # Check for boolean and null values
-                    value = value.lower()
-                else:
-                    # Assume the value is a string and needs quotes
-                    value = '"' + value.replace('"', '\\"') + '"'
-
-                # Rebuild the entry with proper quoting
-                formatted_entry = f'"{key}": {value}'
-                formatted_entries.append(formatted_entry)
-
-            # Reconstruct the JSON string
-            new_json_string = "{" + ", ".join(formatted_entries) + "}"
-            return new_json_string
+        return arguments
 
     def on_tool_error(self, tool: Any, tool_calling: ToolCalling, e: Exception) -> None:
         event_data = self._prepare_event_data(tool, tool_calling)

src/crewai/tools/tool_usage_events.py

@@ -1,6 +1,7 @@
-from typing import Any, Dict
-from pydantic import BaseModel
 from datetime import datetime
+from typing import Any, Dict
+
+from pydantic import BaseModel
 
 
 class ToolUsageEvent(BaseModel):

src/crewai/translations/en.json

@@ -9,34 +9,42 @@
     "task": "\nCurrent Task: {input}\n\nBegin! This is VERY important to you, use the tools available and give your best Final Answer, your job depends on it!\n\nThought:",
     "memory": "\n\n# Useful context: \n{memory}",
     "role_playing": "You are {role}. {backstory}\nYour personal goal is: {goal}",
-    "tools": "\nYou ONLY have access to the following tools, and should NEVER make up tools that are not listed here:\n\n{tools}\n\nUse the following format:\n\nThought: you should always think about what to do\nAction: the action to take, only one name of [{tool_names}], just the name, exactly as it's written.\nAction Input: the input to the action, just a simple python dictionary, enclosed in curly braces, using \" to wrap keys and values.\nObservation: the result of the action\n\nOnce all necessary information is gathered:\n\nThought: I now know the final answer\nFinal Answer: the final answer to the original input question\n",
-    "no_tools": "\nTo give my best complete final answer to the task use the exact following format:\n\nThought: I now can give a great answer\nFinal Answer: Your final answer must be the great and the most complete as possible, it must be outcome described.\n\nI MUST use these formats, my job depends on it!",
-    "format": "I MUST either use a tool (use one at time) OR give my best final answer not both at the same time. To Use the following format:\n\nThought: you should always think about what to do\nAction: the action to take, should be one of [{tool_names}]\nAction Input: the input to the action, dictionary enclosed in curly braces\nObservation: the result of the action\n... (this Thought/Action/Action Input/Result can repeat N times)\nThought: I now can give a great answer\nFinal Answer: Your final answer must be the great and the most complete as possible, it must be outcome described\n\n",
-    "final_answer_format": "If you don't need to use any more tools, you must give your best complete final answer, make sure it satisfy the expect criteria, use the EXACT format below:\n\nThought: I now can give a great answer\nFinal Answer: my best complete final answer to the task.\n\n",
-    "format_without_tools": "\nSorry, I didn't use the right format. I MUST either use a tool (among the available ones), OR give my best final answer.\nI just remembered the expected format I must follow:\n\nQuestion: the input question you must answer\nThought: you should always think about what to do\nAction: the action to take, should be one of [{tool_names}]\nAction Input: the input to the action\nObservation: the result of the action\n... (this Thought/Action/Action Input/Result can repeat N times)\nThought: I now can give a great answer\nFinal Answer: Your final answer must be the great and the most complete as possible, it must be outcome described\n\n",
+    "tools": "\nYou ONLY have access to the following tools, and should NEVER make up tools that are not listed here:\n\n{tools}\n\nIMPORTANT: Use the following format in your response:\n\n```\nThought: you should always think about what to do\nAction: the action to take, only one name of [{tool_names}], just the name, exactly as it's written.\nAction Input: the input to the action, just a simple JSON object, enclosed in curly braces, using \" to wrap keys and values.\nObservation: the result of the action\n```\n\nOnce all necessary information is gathered, return the following format:\n\n```\nThought: I now know the final answer\nFinal Answer: the final answer to the original input question\n```",
+    "no_tools": "\nTo give my best complete final answer to the task respond using the exact following format:\n\nThought: I now can give a great answer\nFinal Answer: Your final answer must be the great and the most complete as possible, it must be outcome described.\n\nI MUST use these formats, my job depends on it!",
+    "format": "I MUST either use a tool (use one at time) OR give my best final answer not both at the same time. When responding, I must use the following format:\n\n```\nThought: you should always think about what to do\nAction: the action to take, should be one of [{tool_names}]\nAction Input: the input to the action, dictionary enclosed in curly braces\nObservation: the result of the action\n```\nThis Thought/Action/Action Input/Result can repeat N times. Once I know the final answer, I must return the following format:\n\n```\nThought: I now can give a great answer\nFinal Answer: Your final answer must be the great and the most complete as possible, it must be outcome described\n\n```",
+    "final_answer_format": "If you don't need to use any more tools, you must give your best complete final answer, make sure it satisfies the expected criteria, use the EXACT format below:\n\n```\nThought: I now can give a great answer\nFinal Answer: my best complete final answer to the task.\n\n```",
+    "format_without_tools": "\nSorry, I didn't use the right format. I MUST either use a tool (among the available ones), OR give my best final answer.\nHere is the expected format I must follow:\n\n```\nQuestion: the input question you must answer\nThought: you should always think about what to do\nAction: the action to take, should be one of [{tool_names}]\nAction Input: the input to the action\nObservation: the result of the action\n```\n This Thought/Action/Action Input/Result process can repeat N times. Once I know the final answer, I must return the following format:\n\n```\nThought: I now can give a great answer\nFinal Answer: Your final answer must be the great and the most complete as possible, it must be outcome described\n\n```",
     "task_with_context": "{task}\n\nThis is the context you're working with:\n{context}",
     "expected_output": "\nThis is the expect criteria for your final answer: {expected_output}\nyou MUST return the actual complete content as the final answer, not a summary.",
     "human_feedback": "You got human feedback on your work, re-evaluate it and give a new Final Answer when ready.\n {human_feedback}",
     "getting_input": "This is the agent's final answer: {final_answer}\n\n",
     "summarizer_system_message": "You are a helpful assistant that summarizes text.",
-    "sumamrize_instruction": "Summarize the following text, make sure to include all the important information: {group}",
+    "summarize_instruction": "Summarize the following text, make sure to include all the important information: {group}",
     "summary": "This is a summary of our conversation so far:\n{merged_summary}",
     "manager_request": "Your best answer to your coworker asking you this, accounting for the context shared.",
     "formatted_task_instructions": "Ensure your final answer contains only the content in the following format: {output_format}\n\nEnsure the final output does not include any code block markers like ```json or ```python.",
-    "human_feedback_classification": "Determine if the following feedback indicates that the user is satisfied or if further changes are needed. Respond with 'True' if further changes are needed, or 'False' if the user is satisfied. **Important** Do not include any additional commentary outside of your 'True' or 'False' response.\n\nFeedback: \"{feedback}\""
+    "human_feedback_classification": "Determine if the following feedback indicates that the user is satisfied or if further changes are needed. Respond with 'True' if further changes are needed, or 'False' if the user is satisfied. **Important** Do not include any additional commentary outside of your 'True' or 'False' response.\n\nFeedback: \"{feedback}\"",
+    "conversation_history_instruction": "You are a member of a crew collaborating to achieve a common goal. Your task is a specific action that contributes to this larger objective. For additional context, please review the conversation history between you and the user that led to the initiation of this crew. Use any relevant information or feedback from the conversation to inform your task execution and ensure your response aligns with both the immediate task and the crew's overall goals."
   },
   "errors": {
-    "force_final_answer_error": "You can't keep going, this was the best you could do.\n {formatted_answer.text}",
+    "force_final_answer_error": "You can't keep going, here is the best final answer you generated:\n\n {formatted_answer}",
     "force_final_answer": "Now it's time you MUST give your absolute best final answer. You'll ignore all previous instructions, stop using any tools, and just return your absolute BEST Final answer.",
-    "agent_tool_unexsiting_coworker": "\nError executing tool. coworker mentioned not found, it must be one of the following options:\n{coworkers}\n",
+    "agent_tool_unexisting_coworker": "\nError executing tool. coworker mentioned not found, it must be one of the following options:\n{coworkers}\n",
     "task_repeated_usage": "I tried reusing the same input, I must stop using this action input. I'll try something else instead.\n\n",
     "tool_usage_error": "I encountered an error: {error}",
     "tool_arguments_error": "Error: the Action Input is not a valid key, value dictionary.",
     "wrong_tool_name": "You tried to use the tool {tool}, but it doesn't exist. You must use one of the following tools, use one at time: {tools}.",
-    "tool_usage_exception": "I encountered an error while trying to use the tool. This was the error: {error}.\n Tool {tool} accepts these inputs: {tool_inputs}"
+    "tool_usage_exception": "I encountered an error while trying to use the tool. This was the error: {error}.\n Tool {tool} accepts these inputs: {tool_inputs}",
+    "agent_tool_execution_error": "Error executing task with agent '{agent_role}'. Error: {error}",
+    "validation_error": "### Previous attempt failed validation: {guardrail_result_error}\n\n\n### Previous result:\n{task_output}\n\n\nTry again, making sure to address the validation error."
   },
   "tools": {
     "delegate_work": "Delegate a specific task to one of the following coworkers: {coworkers}\nThe input to this tool should be the coworker, the task you want them to do, and ALL necessary context to execute the task, they know nothing about the task, so share absolute everything you know, don't reference things but instead explain them.",
-    "ask_question": "Ask a specific question to one of the following coworkers: {coworkers}\nThe input to this tool should be the coworker, the question you have for them, and ALL necessary context to ask the question properly, they know nothing about the question, so share absolute everything you know, don't reference things but instead explain them."
+    "ask_question": "Ask a specific question to one of the following coworkers: {coworkers}\nThe input to this tool should be the coworker, the question you have for them, and ALL necessary context to ask the question properly, they know nothing about the question, so share absolute everything you know, don't reference things but instead explain them.",
+    "add_image": {
+      "name": "Add image to content",
+      "description": "See image to understand it's content, you can optionally ask a question about the image",
+      "default_action": "Please provide a detailed description of this image, including all visual elements, context, and any notable details you can observe."
+    }
   }
 }

src/crewai/types/crew_chat.py

@@ -0,0 +1,40 @@
+from typing import List
+
+from pydantic import BaseModel, Field
+
+
+class ChatInputField(BaseModel):
+    """
+    Represents a single required input for the crew, with a name and short description.
+    Example:
+        {
+            "name": "topic",
+            "description": "The topic to focus on for the conversation"
+        }
+    """
+
+    name: str = Field(..., description="The name of the input field")
+    description: str = Field(..., description="A short description of the input field")
+
+
+class ChatInputs(BaseModel):
+    """
+    Holds a high-level crew_description plus a list of ChatInputFields.
+    Example:
+        {
+            "crew_name": "topic-based-qa",
+            "crew_description": "Use this crew for topic-based Q&A",
+            "inputs": [
+                {"name": "topic", "description": "The topic to focus on"},
+                {"name": "username", "description": "Name of the user"},
+            ]
+        }
+    """
+
+    crew_name: str = Field(..., description="The name of the crew")
+    crew_description: str = Field(
+        ..., description="A description of the crew's purpose"
+    )
+    inputs: List[ChatInputField] = Field(
+        default_factory=list, description="A list of input fields for the crew"
+    )

src/crewai/utilities/crew_json_encoder.py

@@ -1,15 +1,20 @@
-from datetime import datetime, date
+"""JSON encoder for handling CrewAI specific types."""
+
 import json
-from uuid import UUID
-from pydantic import BaseModel
+from datetime import date, datetime
 from decimal import Decimal
+from enum import Enum
+from uuid import UUID
+
+from pydantic import BaseModel
 
 
 class CrewJSONEncoder(json.JSONEncoder):
+    """Custom JSON encoder for CrewAI objects and special types."""
     def default(self, obj):
         if isinstance(obj, BaseModel):
             return self._handle_pydantic_model(obj)
-        elif isinstance(obj, UUID) or isinstance(obj, Decimal):
+        elif isinstance(obj, UUID) or isinstance(obj, Decimal) or isinstance(obj, Enum):
             return str(obj)
 
         elif isinstance(obj, datetime) or isinstance(obj, date):

src/crewai/utilities/crew_pydantic_output_parser.py

@@ -1,13 +1,15 @@
 import json
-import regex
 from typing import Any, Type
 
-from crewai.agents.parser import OutputParserException
+import regex
 from pydantic import BaseModel, ValidationError
 
+from crewai.agents.parser import OutputParserException
+
+"""Parser for converting text outputs into Pydantic models."""
 
 class CrewPydanticOutputParser:
-    """Parses the text into pydantic models"""
+    """Parses text outputs into specified Pydantic models."""
 
     pydantic_object: Type[BaseModel]
 

src/crewai/utilities/embedding_configurator.py

@@ -1,6 +1,7 @@
 import os
 from typing import Any, Dict, cast
-from chromadb import EmbeddingFunction, Documents, Embeddings
+
+from chromadb import Documents, EmbeddingFunction, Embeddings
 from chromadb.api.types import validate_embedding_function