fix: Fix type-checker error in crew_agent_executor.py

Co-Authored-By: Joe Moura <joao@crewai.com>

README.md

@@ -1,10 +1,18 @@
 <div align="center">
 
-![Logo of CrewAI, two people rowing on a boat](./docs/crewai_logo.png)
+![Logo of CrewAI](./docs/crewai_logo.png)
 
 # **CrewAI**
 
-🤖 **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.
+**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.
+
+**CrewAI Enterprise**
+Want to plan, build (+ no code), deploy, monitor and interare your agents: [CrewAI Enterprise](https://www.crewai.com/enterprise). Designed for complex, real-world applications, our enterprise solution offers:
+
+- **Seamless Integrations**
+- **Scalable & Secure Deployment**
+- **Actionable Insights**
+- **24/7 Support**
 
 <h3>
 
@@ -392,7 +400,7 @@ class AdvancedAnalysisFlow(Flow[MarketState]):
             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",
@@ -403,7 +411,7 @@ class AdvancedAnalysisFlow(Flow[MarketState]):
             expected_output="Corroborating evidence and potential contradictions",
             agent=researcher
         )
-        
+
         # Demonstrate crew autonomy
         analysis_crew = Crew(
             agents=[analyst, researcher],

docs/concepts/knowledge.mdx

@@ -91,7 +91,7 @@ result = crew.kickoff(inputs={"question": "What city does John live in and how o
 ```
 
 
-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. 
+Here's another example with the `CrewDoclingSource`. The CrewDoclingSource is actually quite versatile and can handle multiple file formats including MD, PDF, DOCX, HTML, and more. 
 
 <Note>
   You need to install `docling` for the following example to work: `uv add docling`
@@ -152,10 +152,10 @@ 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
+from crewai.knowledge.source.text_file_knowledge_source import TextFileKnowledgeSource
 
 # Create a text file knowledge source
-text_source = CrewDoclingSource(
+text_source = TextFileKnowledgeSource(
     file_paths=["document.txt", "another.txt"]
 )
 

docs/concepts/llms.mdx

@@ -463,26 +463,32 @@ Learn how to get the most out of your LLM configuration:
 
   <Accordion title="Google">
     ```python Code
-    # Option 1. Gemini accessed with an API key.
+    # 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.
+    # Option 2: Vertex AI IAM credentials for Gemini, Anthropic, and Model Garden.
     # https://cloud.google.com/vertex-ai/generative-ai/docs/overview
     ```
 
-    ## GET CREDENTIALS 
+    Get credentials:
+    ```python Code
+    import json
+
     file_path = 'path/to/vertex_ai_service_account.json'
 
     # Load the JSON file
     with open(file_path, 'r') as file:
         vertex_credentials = json.load(file)
 
-    # Convert to JSON string
+    # Convert the credentials to a JSON string
     vertex_credentials_json = json.dumps(vertex_credentials)
+    ```
 
     Example usage:
     ```python Code
+    from crewai import LLM
+
     llm = LLM(
         model="gemini/gemini-1.5-pro-latest",
         temperature=0.7,

docs/concepts/memory.mdx

@@ -58,41 +58,107 @@ my_crew = Crew(
 ### Example: Use Custom Memory Instances e.g FAISS as the VectorDB
 
 ```python Code
-from crewai import Crew, Agent, Task, Process
+from crewai import Crew, Process
+from crewai.memory import LongTermMemory, ShortTermMemory, EntityMemory
+from crewai.memory.storage import LTMSQLiteStorage, RAGStorage
+from typing import List, Optional
 
 # Assemble your crew with memory capabilities
-my_crew = Crew(
-    agents=[...],
-    tasks=[...],
-    process="Process.sequential",
-    memory=True,
-    long_term_memory=EnhanceLongTermMemory(
+my_crew: Crew = Crew(
+    agents = [...],
+    tasks = [...],
+    process = Process.sequential,
+    memory = True,
+    # Long-term memory for persistent storage across sessions
+    long_term_memory = LongTermMemory(
         storage=LTMSQLiteStorage(
-            db_path="/my_data_dir/my_crew1/long_term_memory_storage.db"
+            db_path="/my_crew1/long_term_memory_storage.db"
         )
     ),
-    short_term_memory=EnhanceShortTermMemory(
-        storage=CustomRAGStorage(
-            crew_name="my_crew",
-            storage_type="short_term",
-            data_dir="//my_data_dir",
-            model=embedder["model"],
-            dimension=embedder["dimension"],
+    # Short-term memory for current context using RAG
+    short_term_memory = ShortTermMemory(
+        storage = RAGStorage(
+                embedder_config={
+                    "provider": "openai",
+                    "config": {
+                        "model": 'text-embedding-3-small'
+                    }
+                },
+                type="short_term",
+                path="/my_crew1/"
+            )
         ),
     ),
-    entity_memory=EnhanceEntityMemory(
-        storage=CustomRAGStorage(
-            crew_name="my_crew",
-            storage_type="entities",
-            data_dir="//my_data_dir",
-            model=embedder["model"],
-            dimension=embedder["dimension"],
-        ),
+    # Entity memory for tracking key information about entities
+    entity_memory = EntityMemory(
+        storage=RAGStorage(
+            embedder_config={
+                "provider": "openai",
+                "config": {
+                    "model": 'text-embedding-3-small'
+                }
+            },
+            type="short_term",
+            path="/my_crew1/"
+        )
     ),
     verbose=True,
 )
 ```
 
+## Security Considerations
+
+When configuring memory storage:
+- Use environment variables for storage paths (e.g., `CREWAI_STORAGE_DIR`)
+- Never hardcode sensitive information like database credentials
+- Consider access permissions for storage directories
+- Use relative paths when possible to maintain portability
+
+Example using environment variables:
+```python
+import os
+from crewai import Crew
+from crewai.memory import LongTermMemory
+from crewai.memory.storage import LTMSQLiteStorage
+
+# Configure storage path using environment variable
+storage_path = os.getenv("CREWAI_STORAGE_DIR", "./storage")
+crew = Crew(
+    memory=True,
+    long_term_memory=LongTermMemory(
+        storage=LTMSQLiteStorage(
+            db_path="{storage_path}/memory.db".format(storage_path=storage_path)
+        )
+    )
+)
+```
+
+## Configuration Examples
+
+### Basic Memory Configuration
+```python
+from crewai import Crew
+from crewai.memory import LongTermMemory
+
+# Simple memory configuration
+crew = Crew(memory=True)  # Uses default storage locations
+```
+
+### Custom Storage Configuration
+```python
+from crewai import Crew
+from crewai.memory import LongTermMemory
+from crewai.memory.storage import LTMSQLiteStorage
+
+# Configure custom storage paths
+crew = Crew(
+    memory=True,
+    long_term_memory=LongTermMemory(
+        storage=LTMSQLiteStorage(db_path="./memory.db")
+    )
+)
+```
+
 ## Integrating Mem0 for Enhanced User Memory
 
 [Mem0](https://mem0.ai/) is a self-improving memory layer for LLM applications, enabling personalized AI experiences. 
@@ -216,6 +282,19 @@ my_crew = Crew(
 
 ### Using Google AI embeddings
 
+#### Prerequisites
+Before using Google AI embeddings, ensure you have:
+- Access to the Gemini API
+- The necessary API keys and permissions
+
+You will need to update your *pyproject.toml* dependencies:
+```YAML
+dependencies = [
+    "google-generativeai>=0.8.4", #main version in January/2025 - crewai v.0.100.0 and crewai-tools 0.33.0
+    "crewai[tools]>=0.100.0,<1.0.0"
+]
+```
+
 ```python Code
 from crewai import Crew, Agent, Task, Process
 
@@ -368,6 +447,65 @@ my_crew = Crew(
 )
 ```
 
+### Using Amazon Bedrock embeddings
+
+```python Code
+# Note: Ensure you have installed `boto3` for Bedrock embeddings to work.
+
+import os
+import boto3
+from crewai import Crew, Agent, Task, Process
+
+boto3_session = boto3.Session(
+    region_name=os.environ.get("AWS_REGION_NAME"),
+    aws_access_key_id=os.environ.get("AWS_ACCESS_KEY_ID"),
+    aws_secret_access_key=os.environ.get("AWS_SECRET_ACCESS_KEY")
+)
+
+my_crew = Crew(
+    agents=[...],
+    tasks=[...],
+    process=Process.sequential,
+    memory=True,
+    embedder={
+    "provider": "bedrock",
+        "config":{
+            "session": boto3_session,
+            "model": "amazon.titan-embed-text-v2:0",
+            "vector_dimension": 1024
+        }
+    }
+    verbose=True
+)
+```
+
+### Adding Custom Embedding Function
+
+```python Code
+from crewai import Crew, Agent, Task, Process
+from chromadb import Documents, EmbeddingFunction, Embeddings
+
+# Create a custom embedding function
+class CustomEmbedder(EmbeddingFunction):
+    def __call__(self, input: Documents) -> Embeddings:
+        # generate embeddings
+        return [1, 2, 3] # this is a dummy embedding
+
+my_crew = Crew(
+    agents=[...],
+    tasks=[...],
+    process=Process.sequential,
+    memory=True,
+    verbose=True,
+    embedder={
+        "provider": "custom",
+        "config": {
+            "embedder": CustomEmbedder()
+        }
+    }
+)
+```
+
 ### Resetting Memory
 
 ```shell

docs/concepts/tasks.mdx

@@ -268,7 +268,7 @@ analysis_task = Task(
 
 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.
+feedback to agents when their output doesn't meet specific criteria.
 
 ### Using Task Guardrails
 

docs/tools/filewritetool.mdx

@@ -8,9 +8,9 @@ icon: file-pen
 
 ## Description
 
-The `FileWriterTool` is a component of the crewai_tools package, designed to simplify the process of writing content to files. 
+The `FileWriterTool` is a component of the crewai_tools package, designed to simplify the process of writing content to files with cross-platform compatibility (Windows, Linux, macOS). 
 It is particularly useful in scenarios such as generating reports, saving logs, creating configuration files, and more. 
-This tool supports creating new directories if they don't exist, making it easier to organize your output.
+This tool handles path differences across operating systems, supports UTF-8 encoding, and automatically creates directories if they don't exist, making it easier to organize your output reliably across different platforms.
 
 ## Installation
 
@@ -43,6 +43,8 @@ print(result)
 
 ## Conclusion
 
-By integrating the `FileWriterTool` into your crews, the agents can execute the process of writing content to files and creating directories. 
-This tool is essential for tasks that require saving output data, creating structured file systems, and more. By adhering to the setup and usage guidelines provided, 
-incorporating this tool into projects is straightforward and efficient.
+By integrating the `FileWriterTool` into your crews, the agents can reliably write content to files across different operating systems. 
+This tool is essential for tasks that require saving output data, creating structured file systems, and handling cross-platform file operations. 
+It's particularly recommended for Windows users who may encounter file writing issues with standard Python file operations.
+
+By adhering to the setup and usage guidelines provided, incorporating this tool into projects is straightforward and ensures consistent file writing behavior across all platforms.

src/crewai/agent.py

@@ -16,7 +16,6 @@ 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
@@ -146,7 +145,7 @@ class Agent(BaseAgent):
     def _set_knowledge(self):
         try:
             if self.knowledge_sources:
-                full_pattern = re.compile(r'[^a-zA-Z0-9\-_\r\n]|(\.\.)')
+                full_pattern = re.compile(r"[^a-zA-Z0-9\-_\r\n]|(\.\.)")
                 knowledge_agent_name = f"{re.sub(full_pattern, '_', self.role)}"
                 if isinstance(self.knowledge_sources, list) and all(
                     isinstance(k, BaseKnowledgeSource) for k in self.knowledge_sources

src/crewai/agents/crew_agent_executor.py

@@ -366,23 +366,41 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
             tool_result = tool_calling.message
             return ToolResult(result=tool_result, result_as_answer=False)
         else:
-            if tool_calling.tool_name.casefold().strip() in [
-                name.casefold().strip() for name in self.tool_name_to_tool_map
-            ] or tool_calling.tool_name.casefold().replace("_", " ") in [
-                name.casefold().strip() for name in self.tool_name_to_tool_map
-            ]:
+            valid_names = [name.casefold().strip() for name in self.tool_name_to_tool_map]
+            tool_name = tool_calling.tool_name.casefold().strip()
+            tool_name_alt = tool_calling.tool_name.casefold().replace("_", " ")
+            
+            if tool_name in valid_names or tool_name_alt in valid_names:
                 tool_result = tool_usage.use(tool_calling, agent_action.text)
+                tool_result = self._clean_tool_result(tool_result)
                 tool = self.tool_name_to_tool_map.get(tool_calling.tool_name)
-                if tool:
-                    return ToolResult(
-                        result=tool_result, result_as_answer=tool.result_as_answer
-                    )
-            else:
-                tool_result = self._i18n.errors("wrong_tool_name").format(
-                    tool=tool_calling.tool_name,
-                    tools=", ".join([tool.name.casefold() for tool in self.tools]),
+                return ToolResult(
+                    result=tool_result,
+                    result_as_answer=tool.result_as_answer if tool else False
                 )
-        return ToolResult(result=tool_result, result_as_answer=False)
+            
+            tool_result = self._i18n.errors("wrong_tool_name").format(
+                tool=tool_calling.tool_name,
+                tools=", ".join([tool.name.casefold() for tool in self.tools]),
+            )
+            return ToolResult(result=tool_result, result_as_answer=False)
+
+    def _clean_tool_result(self, tool_result: Any) -> Any:
+        """Clean tool result by removing trailing backticks.
+        
+        This is particularly important in hierarchical mode where tool outputs
+        might contain markdown formatting that needs to be cleaned up.
+        
+        Args:
+            tool_result: The result from a tool execution, can be any type
+            
+        Returns:
+            The cleaned result with any trailing backticks removed if it's a string,
+            otherwise returns the original result unchanged
+        """
+        if isinstance(tool_result, str):
+            return tool_result.rstrip('`').rstrip('```')
+        return tool_result
 
     def _summarize_messages(self) -> None:
         messages_groups = []

src/crewai/cli/reset_memories_command.py

@@ -3,11 +3,6 @@ import subprocess
 import click
 
 from crewai.cli.utils import get_crew
-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
 
 
 def reset_memories_command(

src/crewai/crew.py

@@ -1,7 +1,6 @@
 import asyncio
 import json
 import re
-import sys
 import uuid
 import warnings
 from concurrent.futures import Future
@@ -381,6 +380,22 @@ class Crew(BaseModel):
 
         return self
 
+    @model_validator(mode="after")
+    def validate_must_have_non_conditional_task(self) -> "Crew":
+        """Ensure that a crew has at least one non-conditional task."""
+        if not self.tasks:
+            return self
+        non_conditional_count = sum(
+            1 for task in self.tasks if not isinstance(task, ConditionalTask)
+        )
+        if non_conditional_count == 0:
+            raise PydanticCustomError(
+                "only_conditional_tasks",
+                "Crew must include at least one non-conditional task",
+                {},
+            )
+        return self
+
     @model_validator(mode="after")
     def validate_first_task(self) -> "Crew":
         """Ensure the first task is not a ConditionalTask."""
@@ -441,6 +456,7 @@ class Crew(BaseModel):
         return self
 
 
+
     @property
     def key(self) -> str:
         source = [agent.key for agent in self.agents] + [
@@ -743,6 +759,7 @@ class Crew(BaseModel):
                     task, task_outputs, futures, task_index, was_replayed
                 )
                 if skipped_task_output:
+                    task_outputs.append(skipped_task_output)
                     continue
 
             if task.async_execution:
@@ -766,7 +783,7 @@ class Crew(BaseModel):
                     context=context,
                     tools=tools_for_task,
                 )
-                task_outputs = [task_output]
+                task_outputs.append(task_output)
                 self._process_task_result(task, task_output)
                 self._store_execution_log(task, task_output, task_index, was_replayed)
 
@@ -787,7 +804,7 @@ class Crew(BaseModel):
             task_outputs = self._process_async_tasks(futures, was_replayed)
             futures.clear()
 
-        previous_output = task_outputs[task_index - 1] if task_outputs else None
+        previous_output = task_outputs[-1] if task_outputs else None
         if previous_output is not None and not task.should_execute(previous_output):
             self._logger.log(
                 "debug",
@@ -909,11 +926,15 @@ class Crew(BaseModel):
             )
 
     def _create_crew_output(self, task_outputs: List[TaskOutput]) -> CrewOutput:
-        if len(task_outputs) != 1:
-            raise ValueError(
-                "Something went wrong. Kickoff should return only one task output."
-            )
-        final_task_output = task_outputs[0]
+        if not task_outputs:
+            raise ValueError("No task outputs available to create crew output.")
+            
+        # Filter out empty outputs and get the last valid one as the main output
+        valid_outputs = [t for t in task_outputs if t.raw]
+        if not valid_outputs:
+            raise ValueError("No valid task outputs available to create crew output.")
+        final_task_output = valid_outputs[-1]
+            
         final_string_output = final_task_output.raw
         self._finish_execution(final_string_output)
         token_usage = self.calculate_usage_metrics()
@@ -922,7 +943,7 @@ class Crew(BaseModel):
             raw=final_task_output.raw,
             pydantic=final_task_output.pydantic,
             json_dict=final_task_output.json_dict,
-            tasks_output=[task.output for task in self.tasks if task.output],
+            tasks_output=task_outputs,
             token_usage=token_usage,
         )
 

src/crewai/knowledge/source/excel_knowledge_source.py

@@ -1,28 +1,138 @@
 from pathlib import Path
-from typing import Dict, List
+from typing import Dict, Iterator, List, Optional, Union
+from urllib.parse import urlparse
 
-from crewai.knowledge.source.base_file_knowledge_source import BaseFileKnowledgeSource
+from pydantic import Field, field_validator
+
+from crewai.knowledge.source.base_knowledge_source import BaseKnowledgeSource
+from crewai.utilities.constants import KNOWLEDGE_DIRECTORY
+from crewai.utilities.logger import Logger
 
 
-class ExcelKnowledgeSource(BaseFileKnowledgeSource):
+class ExcelKnowledgeSource(BaseKnowledgeSource):
     """A knowledge source that stores and queries Excel file content using embeddings."""
 
-    def load_content(self) -> Dict[Path, str]:
-        """Load and preprocess Excel file content."""
-        pd = self._import_dependencies()
+    # override content to be a dict of file paths to sheet names to csv content
 
+    _logger: Logger = Logger(verbose=True)
+
+    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"
+    )
+    chunks: List[str] = Field(default_factory=list)
+    content: Dict[Path, Dict[str, str]] = Field(default_factory=dict)
+    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 _process_file_paths(self) -> List[Path]:
+        """Convert file_path to a list of Path objects."""
+
+        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 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 path_list]
+
+    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 it's inside the knowledge directory, use the relative path.",
+                    color="red",
+                )
+                raise FileNotFoundError(f"File not found: {path}")
+            if not path.is_file():
+                self._logger.log(
+                    "error",
+                    f"Path is not a file: {path}",
+                    color="red",
+                )
+
+    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._process_file_paths()
+        self.validate_content()
+        self.content = self._load_content()
+
+    def _load_content(self) -> Dict[Path, Dict[str, str]]:
+        """Load and preprocess Excel file content from multiple sheets.
+
+        Each sheet's content is converted to CSV format and stored.
+
+        Returns:
+            Dict[Path, Dict[str, str]]: A mapping of file paths to their respective sheet contents.
+
+        Raises:
+            ImportError: If required dependencies are missing.
+            FileNotFoundError: If the specified Excel file cannot be opened.
+        """
+        pd = self._import_dependencies()
         content_dict = {}
         for file_path in self.safe_file_paths:
             file_path = self.convert_to_path(file_path)
-            df = pd.read_excel(file_path)
-            content = df.to_csv(index=False)
-            content_dict[file_path] = content
+            with pd.ExcelFile(file_path) as xl:
+                sheet_dict = {
+                    str(sheet_name): str(
+                        pd.read_excel(xl, sheet_name).to_csv(index=False)
+                    )
+                    for sheet_name in xl.sheet_names
+                }
+            content_dict[file_path] = sheet_dict
         return content_dict
 
+    def convert_to_path(self, path: Union[Path, str]) -> Path:
+        """Convert a path to a Path object."""
+        return Path(KNOWLEDGE_DIRECTORY + "/" + path) if isinstance(path, str) else path
+
     def _import_dependencies(self):
         """Dynamically import dependencies."""
         try:
-            import openpyxl  # noqa
             import pandas as pd
 
             return pd
@@ -38,10 +148,14 @@ class ExcelKnowledgeSource(BaseFileKnowledgeSource):
         and save the embeddings.
         """
         # Convert dictionary values to a single string if content is a dictionary
-        if isinstance(self.content, dict):
-            content_str = "\n".join(str(value) for value in self.content.values())
-        else:
-            content_str = str(self.content)
+        # Updated to account for .xlsx workbooks with multiple tabs/sheets
+        content_str = ""
+        for value in self.content.values():
+            if isinstance(value, dict):
+                for sheet_value in value.values():
+                    content_str += str(sheet_value) + "\n"
+            else:
+                content_str += str(value) + "\n"
 
         new_chunks = self._chunk_text(content_str)
         self.chunks.extend(new_chunks)

src/crewai/llm.py

@@ -164,6 +164,7 @@ class LLM:
         self.context_window_size = 0
         self.reasoning_effort = reasoning_effort
         self.additional_params = kwargs
+        self.is_anthropic = self._is_anthropic_model(model)
 
         litellm.drop_params = True
 
@@ -178,42 +179,62 @@ class LLM:
         self.set_callbacks(callbacks)
         self.set_env_callbacks()
 
+    def _is_anthropic_model(self, model: str) -> bool:
+        """Determine if the model is from Anthropic provider.
+        
+        Args:
+            model: The model identifier string.
+            
+        Returns:
+            bool: True if the model is from Anthropic, False otherwise.
+        """
+        ANTHROPIC_PREFIXES = ('anthropic/', 'claude-', 'claude/')
+        return any(prefix in model.lower() for prefix in ANTHROPIC_PREFIXES)
+
     def call(
         self,
         messages: Union[str, List[Dict[str, str]]],
         tools: Optional[List[dict]] = None,
         callbacks: Optional[List[Any]] = None,
         available_functions: Optional[Dict[str, Any]] = None,
-    ) -> str:
-        """
-        High-level llm call method that:
-          1) Accepts either a string or a list of messages
-          2) Converts string input to the required message format
-          3) Calls litellm.completion
-          4) Handles function/tool calls if any
-          5) Returns the final text response or tool result
-
-        Parameters:
-        - messages (Union[str, List[Dict[str, str]]]): The input messages for the LLM.
-          - If a string is provided, it will be converted into a message list with a single entry.
-          - If a list of dictionaries is provided, each dictionary should have 'role' and 'content' keys.
-        - tools (Optional[List[dict]]): A list of tool schemas for function calling.
-        - callbacks (Optional[List[Any]]): A list of callback functions to be executed.
-        - available_functions (Optional[Dict[str, Any]]): A dictionary mapping function names to actual Python functions.
-
+    ) -> Union[str, Any]:
+        """High-level LLM call method.
+        
+        Args:
+            messages: Input messages for the LLM.
+                     Can be a string or list of message dictionaries.
+                     If string, it will be converted to a single user message.
+                     If list, each dict must have 'role' and 'content' keys.
+            tools: Optional list of tool schemas for function calling.
+                  Each tool should define its name, description, and parameters.
+            callbacks: Optional list of callback functions to be executed
+                      during and after the LLM call.
+            available_functions: Optional dict mapping function names to callables
+                               that can be invoked by the LLM.
+        
         Returns:
-        - str: The final text response from the LLM or the result of a tool function call.
-
+            Union[str, Any]: Either a text response from the LLM (str) or
+                           the result of a tool function call (Any).
+        
+        Raises:
+            TypeError: If messages format is invalid
+            ValueError: If response format is not supported
+            LLMContextLengthExceededException: If input exceeds model's context limit
+        
         Examples:
-        ---------
-        # Example 1: Using a string input
-        response = llm.call("Return the name of a random city in the world.")
-        print(response)
-
-        # Example 2: Using a list of messages
-        messages = [{"role": "user", "content": "What is the capital of France?"}]
-        response = llm.call(messages)
-        print(response)
+            # Example 1: Simple string input
+            >>> response = llm.call("Return the name of a random city.")
+            >>> print(response)
+            "Paris"
+            
+            # Example 2: Message list with system and user messages
+            >>> messages = [
+            ...     {"role": "system", "content": "You are a geography expert"},
+            ...     {"role": "user", "content": "What is France's capital?"}
+            ... ]
+            >>> response = llm.call(messages)
+            >>> print(response)
+            "The capital of France is Paris."
         """
         # Validate parameters before proceeding with the call.
         self._validate_call_params()
@@ -233,10 +254,13 @@ class LLM:
                 self.set_callbacks(callbacks)
 
             try:
-                # --- 1) Prepare the parameters for the completion call
+                # --- 1) Format messages according to provider requirements
+                formatted_messages = self._format_messages_for_provider(messages)
+
+                # --- 2) Prepare the parameters for the completion call
                 params = {
                     "model": self.model,
-                    "messages": messages,
+                    "messages": formatted_messages,
                     "timeout": self.timeout,
                     "temperature": self.temperature,
                     "top_p": self.top_p,
@@ -324,6 +348,38 @@ class LLM:
                     logging.error(f"LiteLLM call failed: {str(e)}")
                 raise
 
+    def _format_messages_for_provider(self, messages: List[Dict[str, str]]) -> List[Dict[str, str]]:
+        """Format messages according to provider requirements.
+        
+        Args:
+            messages: List of message dictionaries with 'role' and 'content' keys.
+                     Can be empty or None.
+        
+        Returns:
+            List of formatted messages according to provider requirements.
+            For Anthropic models, ensures first message has 'user' role.
+        
+        Raises:
+            TypeError: If messages is None or contains invalid message format.
+        """
+        if messages is None:
+            raise TypeError("Messages cannot be None")
+            
+        # Validate message format first
+        for msg in messages:
+            if not isinstance(msg, dict) or "role" not in msg or "content" not in msg:
+                raise TypeError("Invalid message format. Each message must be a dict with 'role' and 'content' keys")
+            
+        if not self.is_anthropic:
+            return messages
+            
+        # Anthropic requires messages to start with 'user' role
+        if not messages or messages[0]["role"] == "system":
+            # If first message is system or empty, add a placeholder user message
+            return [{"role": "user", "content": "."}, *messages]
+                
+        return messages
+
     def _get_custom_llm_provider(self) -> str:
         """
         Derives the custom_llm_provider from the model string.

src/crewai/memory/entity/entity_memory.py

@@ -1,4 +1,4 @@
-from typing import Any, Optional
+from typing import Optional
 
 from pydantic import PrivateAttr
 

src/crewai/memory/memory.py

@@ -1,15 +1,15 @@
-from typing import Any, Dict, List, Optional, Union
+from typing import Any, Dict, List, Optional
 
 from pydantic import BaseModel
 
-from crewai.memory.storage.rag_storage import RAGStorage
-
 
 class Memory(BaseModel):
     """
     Base class for memory, now supporting agent tags and generic metadata.
     """
 
+    embedder_config: Optional[Dict[str, Any]] = None
+
     storage: Any
 
     def __init__(self, storage: Any, **data: Any):

src/crewai/memory/storage/base_rag_storage.py

@@ -13,7 +13,7 @@ class BaseRAGStorage(ABC):
         self,
         type: str,
         allow_reset: bool = True,
-        embedder_config: Optional[Any] = None,
+        embedder_config: Optional[Dict[str, Any]] = None,
         crew: Any = None,
     ):
         self.type = type

src/crewai/task.py

@@ -674,19 +674,32 @@ class Task(BaseModel):
             return OutputFormat.PYDANTIC
         return OutputFormat.RAW
 
-    def _save_file(self, result: Any) -> None:
+    def _save_file(self, result: Union[Dict, str, Any]) -> None:
         """Save task output to a file.
 
+        Note:
+            For cross-platform file writing, especially on Windows, consider using FileWriterTool
+            from the crewai_tools package:
+                pip install 'crewai[tools]'
+                from crewai_tools import FileWriterTool
+
         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
+            RuntimeError: If there is an error writing to the file. For cross-platform
+                compatibility, especially on Windows, use FileWriterTool from crewai_tools
+                package.
         """
         if self.output_file is None:
             raise ValueError("output_file is not set.")
 
+        FILEWRITER_RECOMMENDATION = (
+            "For cross-platform file writing, especially on Windows, "
+            "use FileWriterTool from crewai_tools package."
+        )
+
         try:
             resolved_path = Path(self.output_file).expanduser().resolve()
             directory = resolved_path.parent
@@ -702,7 +715,12 @@ class Task(BaseModel):
                 else:
                     file.write(str(result))
         except (OSError, IOError) as e:
-            raise RuntimeError(f"Failed to save output file: {e}")
+            raise RuntimeError(
+                "\n".join([
+                    f"Failed to save output file: {e}",
+                    FILEWRITER_RECOMMENDATION
+                ])
+            )
         return None
 
     def __repr__(self):

src/crewai/utilities/embedding_configurator.py

@@ -1,5 +1,5 @@
 import os
-from typing import Any, Dict, cast
+from typing import Any, Dict, Optional, cast
 
 from chromadb import Documents, EmbeddingFunction, Embeddings
 from chromadb.api.types import validate_embedding_function
@@ -18,11 +18,12 @@ class EmbeddingConfigurator:
             "bedrock": self._configure_bedrock,
             "huggingface": self._configure_huggingface,
             "watson": self._configure_watson,
+            "custom": self._configure_custom,
         }
 
     def configure_embedder(
         self,
-        embedder_config: Dict[str, Any] | None = None,
+        embedder_config: Optional[Dict[str, Any]] = None,
     ) -> EmbeddingFunction:
         """Configures and returns an embedding function based on the provided config."""
         if embedder_config is None:
@@ -30,20 +31,19 @@ class EmbeddingConfigurator:
 
         provider = embedder_config.get("provider")
         config = embedder_config.get("config", {})
-        model_name = config.get("model")
-
-        if isinstance(provider, EmbeddingFunction):
-            try:
-                validate_embedding_function(provider)
-                return provider
-            except Exception as e:
-                raise ValueError(f"Invalid custom embedding function: {str(e)}")
+        model_name = config.get("model") if provider != "custom" else None
 
         if provider not in self.embedding_functions:
             raise Exception(
                 f"Unsupported embedding provider: {provider}, supported providers: {list(self.embedding_functions.keys())}"
             )
-        return self.embedding_functions[provider](config, model_name)
+
+        embedding_function = self.embedding_functions[provider]
+        return (
+            embedding_function(config)
+            if provider == "custom"
+            else embedding_function(config, model_name)
+        )
 
     @staticmethod
     def _create_default_embedding_function():
@@ -64,6 +64,13 @@ class EmbeddingConfigurator:
         return OpenAIEmbeddingFunction(
             api_key=config.get("api_key") or os.getenv("OPENAI_API_KEY"),
             model_name=model_name,
+            api_base=config.get("api_base", None),
+            api_type=config.get("api_type", None),
+            api_version=config.get("api_version", None),
+            default_headers=config.get("default_headers", None),
+            dimensions=config.get("dimensions", None),
+            deployment_id=config.get("deployment_id", None),
+            organization_id=config.get("organization_id", None),
         )
 
     @staticmethod
@@ -78,6 +85,10 @@ class EmbeddingConfigurator:
             api_type=config.get("api_type", "azure"),
             api_version=config.get("api_version"),
             model_name=model_name,
+            default_headers=config.get("default_headers"),
+            dimensions=config.get("dimensions"),
+            deployment_id=config.get("deployment_id"),
+            organization_id=config.get("organization_id"),
         )
 
     @staticmethod
@@ -100,6 +111,8 @@ class EmbeddingConfigurator:
         return GoogleVertexEmbeddingFunction(
             model_name=model_name,
             api_key=config.get("api_key"),
+            project_id=config.get("project_id"),
+            region=config.get("region"),
         )
 
     @staticmethod
@@ -111,6 +124,7 @@ class EmbeddingConfigurator:
         return GoogleGenerativeAiEmbeddingFunction(
             model_name=model_name,
             api_key=config.get("api_key"),
+            task_type=config.get("task_type"),
         )
 
     @staticmethod
@@ -195,3 +209,28 @@ class EmbeddingConfigurator:
                     raise e
 
         return WatsonEmbeddingFunction()
+
+    @staticmethod
+    def _configure_custom(config):
+        custom_embedder = config.get("embedder")
+        if isinstance(custom_embedder, EmbeddingFunction):
+            try:
+                validate_embedding_function(custom_embedder)
+                return custom_embedder
+            except Exception as e:
+                raise ValueError(f"Invalid custom embedding function: {str(e)}")
+        elif callable(custom_embedder):
+            try:
+                instance = custom_embedder()
+                if isinstance(instance, EmbeddingFunction):
+                    validate_embedding_function(instance)
+                    return instance
+                raise ValueError(
+                    "Custom embedder does not create an EmbeddingFunction instance"
+                )
+            except Exception as e:
+                raise ValueError(f"Error instantiating custom embedder: {str(e)}")
+        else:
+            raise ValueError(
+                "Custom embedder must be an instance of `EmbeddingFunction` or a callable that creates one"
+            )

tests/cassettes/test_hierarchical_tool_output_formatting[ - ].yaml

@@ -0,0 +1,111 @@
+interactions:
+- request:
+    body: '{"messages": [{"role": "system", "content": "You are Crew Manager. You
+      are a seasoned manager with a knack for getting the best out of your team.\nYou
+      are also known for your ability to delegate work to the right people, and to
+      ask the right questions to get the best out of your team.\nEven though you don''t
+      perform tasks by yourself, you have a lot of experience in the field, which
+      allows you to properly evaluate the work of your team members.\nYour personal
+      goal is: Manage the team to complete the task in the best way possible.\nYou
+      ONLY have access to the following tools, and should NEVER make up tools that
+      are not listed here:\n\nTool Name: Delegate work to coworker\nTool Arguments:
+      {''task'': {''description'': ''The task to delegate'', ''type'': ''str''}, ''context'':
+      {''description'': ''The context for the task'', ''type'': ''str''}, ''coworker'':
+      {''description'': ''The role/name of the coworker to delegate to'', ''type'':
+      ''str''}}\nTool Description: Delegate a specific task to one of the following
+      coworkers: Researcher\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.\nTool Name: Ask question to coworker\nTool
+      Arguments: {''question'': {''description'': ''The question to ask'', ''type'':
+      ''str''}, ''context'': {''description'': ''The context for the question'', ''type'':
+      ''str''}, ''coworker'': {''description'': ''The role/name of the coworker to
+      ask'', ''type'': ''str''}}\nTool Description: Ask a specific question to one
+      of the following coworkers: Researcher\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.\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 [Delegate work
+      to coworker, Ask question to coworker], 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```"}, {"role": "user", "content": "\nCurrent Task: Test task
+      using test_tool\n\nThis is the expected criteria for your final answer: Test
+      output\nyou MUST return the actual complete content as the final answer, not
+      a summary.\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:"}], "model":
+      "gpt-4o", "stop": ["\nObservation:"]}'
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate
+      connection:
+      - keep-alive
+      content-length:
+      - '2915'
+      content-type:
+      - application/json
+      cookie:
+      - __cf_bm=0nI_3cs9LKHlz0_cVifl9lVXZWfNQrzhLRZzoaDCVJs-1739358791-1.0.1.1-GJZzVGs1fHQe9gucVHUCoDlI3mwg3JyXCBIIChx_OsB0jgqbzt1s2et96vUgpdjONt9C2tB8OQ6fk70k3vhh0w;
+        _cfuvid=TvEPcNq35qaIgBFFzOP7g7NQf632eeMrvRDJyr7UTro-1739358791985-0.0.1.1-604800000
+      host:
+      - api.openai.com
+      user-agent:
+      - OpenAI/Python 1.61.0
+      x-stainless-arch:
+      - x64
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - Linux
+      x-stainless-package-version:
+      - 1.61.0
+      x-stainless-raw-response:
+      - 'true'
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.12.7
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    content: "{\n    \"error\": {\n        \"message\": \"Incorrect API key provided:
+      sk-fake-**********2345. You can find your API key at https://platform.openai.com/account/api-keys.\",\n
+      \       \"type\": \"invalid_request_error\",\n        \"param\": null,\n        \"code\":
+      \"invalid_api_key\"\n    }\n}\n"
+    headers:
+      CF-RAY:
+      - 910c25fe3c8ca373-SEA
+      Connection:
+      - keep-alive
+      Content-Length:
+      - '272'
+      Content-Type:
+      - application/json; charset=utf-8
+      Date:
+      - Wed, 12 Feb 2025 11:13:16 GMT
+      Server:
+      - cloudflare
+      X-Content-Type-Options:
+      - nosniff
+      alt-svc:
+      - h3=":443"; ma=86400
+      cf-cache-status:
+      - DYNAMIC
+      strict-transport-security:
+      - max-age=31536000; includeSubDomains; preload
+      vary:
+      - Origin
+      x-request-id:
+      - req_a5651cefb4c7ae9483499a8e538f579d
+    http_version: HTTP/1.1
+    status_code: 401
+version: 1

tests/cassettes/test_hierarchical_tool_output_formatting[None-None].yaml

@@ -0,0 +1,111 @@
+interactions:
+- request:
+    body: '{"messages": [{"role": "system", "content": "You are Crew Manager. You
+      are a seasoned manager with a knack for getting the best out of your team.\nYou
+      are also known for your ability to delegate work to the right people, and to
+      ask the right questions to get the best out of your team.\nEven though you don''t
+      perform tasks by yourself, you have a lot of experience in the field, which
+      allows you to properly evaluate the work of your team members.\nYour personal
+      goal is: Manage the team to complete the task in the best way possible.\nYou
+      ONLY have access to the following tools, and should NEVER make up tools that
+      are not listed here:\n\nTool Name: Delegate work to coworker\nTool Arguments:
+      {''task'': {''description'': ''The task to delegate'', ''type'': ''str''}, ''context'':
+      {''description'': ''The context for the task'', ''type'': ''str''}, ''coworker'':
+      {''description'': ''The role/name of the coworker to delegate to'', ''type'':
+      ''str''}}\nTool Description: Delegate a specific task to one of the following
+      coworkers: Researcher\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.\nTool Name: Ask question to coworker\nTool
+      Arguments: {''question'': {''description'': ''The question to ask'', ''type'':
+      ''str''}, ''context'': {''description'': ''The context for the question'', ''type'':
+      ''str''}, ''coworker'': {''description'': ''The role/name of the coworker to
+      ask'', ''type'': ''str''}}\nTool Description: Ask a specific question to one
+      of the following coworkers: Researcher\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.\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 [Delegate work
+      to coworker, Ask question to coworker], 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```"}, {"role": "user", "content": "\nCurrent Task: Test task
+      using test_tool\n\nThis is the expected criteria for your final answer: Test
+      output\nyou MUST return the actual complete content as the final answer, not
+      a summary.\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:"}], "model":
+      "gpt-4o", "stop": ["\nObservation:"]}'
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate
+      connection:
+      - keep-alive
+      content-length:
+      - '2915'
+      content-type:
+      - application/json
+      cookie:
+      - __cf_bm=0nI_3cs9LKHlz0_cVifl9lVXZWfNQrzhLRZzoaDCVJs-1739358791-1.0.1.1-GJZzVGs1fHQe9gucVHUCoDlI3mwg3JyXCBIIChx_OsB0jgqbzt1s2et96vUgpdjONt9C2tB8OQ6fk70k3vhh0w;
+        _cfuvid=TvEPcNq35qaIgBFFzOP7g7NQf632eeMrvRDJyr7UTro-1739358791985-0.0.1.1-604800000
+      host:
+      - api.openai.com
+      user-agent:
+      - OpenAI/Python 1.61.0
+      x-stainless-arch:
+      - x64
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - Linux
+      x-stainless-package-version:
+      - 1.61.0
+      x-stainless-raw-response:
+      - 'true'
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.12.7
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    content: "{\n    \"error\": {\n        \"message\": \"Incorrect API key provided:
+      sk-fake-**********2345. You can find your API key at https://platform.openai.com/account/api-keys.\",\n
+      \       \"type\": \"invalid_request_error\",\n        \"param\": null,\n        \"code\":
+      \"invalid_api_key\"\n    }\n}\n"
+    headers:
+      CF-Cache-Status:
+      - DYNAMIC
+      CF-RAY:
+      - 910c25fa2a76a373-SEA
+      Connection:
+      - keep-alive
+      Content-Length:
+      - '272'
+      Content-Type:
+      - application/json; charset=utf-8
+      Date:
+      - Wed, 12 Feb 2025 11:13:15 GMT
+      Server:
+      - cloudflare
+      X-Content-Type-Options:
+      - nosniff
+      alt-svc:
+      - h3=":443"; ma=86400
+      strict-transport-security:
+      - max-age=31536000; includeSubDomains; preload
+      vary:
+      - Origin
+      x-request-id:
+      - req_3ad40f35d4ad6d42f9bbb7fd8ace9a68
+    http_version: HTTP/1.1
+    status_code: 401
+version: 1

tests/cassettes/test_hierarchical_tool_output_formatting[malformed`result```test-malformed`result```test].yaml

@@ -0,0 +1,111 @@
+interactions:
+- request:
+    body: '{"messages": [{"role": "system", "content": "You are Crew Manager. You
+      are a seasoned manager with a knack for getting the best out of your team.\nYou
+      are also known for your ability to delegate work to the right people, and to
+      ask the right questions to get the best out of your team.\nEven though you don''t
+      perform tasks by yourself, you have a lot of experience in the field, which
+      allows you to properly evaluate the work of your team members.\nYour personal
+      goal is: Manage the team to complete the task in the best way possible.\nYou
+      ONLY have access to the following tools, and should NEVER make up tools that
+      are not listed here:\n\nTool Name: Delegate work to coworker\nTool Arguments:
+      {''task'': {''description'': ''The task to delegate'', ''type'': ''str''}, ''context'':
+      {''description'': ''The context for the task'', ''type'': ''str''}, ''coworker'':
+      {''description'': ''The role/name of the coworker to delegate to'', ''type'':
+      ''str''}}\nTool Description: Delegate a specific task to one of the following
+      coworkers: Researcher\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.\nTool Name: Ask question to coworker\nTool
+      Arguments: {''question'': {''description'': ''The question to ask'', ''type'':
+      ''str''}, ''context'': {''description'': ''The context for the question'', ''type'':
+      ''str''}, ''coworker'': {''description'': ''The role/name of the coworker to
+      ask'', ''type'': ''str''}}\nTool Description: Ask a specific question to one
+      of the following coworkers: Researcher\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.\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 [Delegate work
+      to coworker, Ask question to coworker], 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```"}, {"role": "user", "content": "\nCurrent Task: Test task
+      using test_tool\n\nThis is the expected criteria for your final answer: Test
+      output\nyou MUST return the actual complete content as the final answer, not
+      a summary.\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:"}], "model":
+      "gpt-4o", "stop": ["\nObservation:"]}'
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate
+      connection:
+      - keep-alive
+      content-length:
+      - '2915'
+      content-type:
+      - application/json
+      cookie:
+      - __cf_bm=0nI_3cs9LKHlz0_cVifl9lVXZWfNQrzhLRZzoaDCVJs-1739358791-1.0.1.1-GJZzVGs1fHQe9gucVHUCoDlI3mwg3JyXCBIIChx_OsB0jgqbzt1s2et96vUgpdjONt9C2tB8OQ6fk70k3vhh0w;
+        _cfuvid=TvEPcNq35qaIgBFFzOP7g7NQf632eeMrvRDJyr7UTro-1739358791985-0.0.1.1-604800000
+      host:
+      - api.openai.com
+      user-agent:
+      - OpenAI/Python 1.61.0
+      x-stainless-arch:
+      - x64
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - Linux
+      x-stainless-package-version:
+      - 1.61.0
+      x-stainless-raw-response:
+      - 'true'
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.12.7
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    content: "{\n    \"error\": {\n        \"message\": \"Incorrect API key provided:
+      sk-fake-**********2345. You can find your API key at https://platform.openai.com/account/api-keys.\",\n
+      \       \"type\": \"invalid_request_error\",\n        \"param\": null,\n        \"code\":
+      \"invalid_api_key\"\n    }\n}\n"
+    headers:
+      CF-Cache-Status:
+      - DYNAMIC
+      CF-RAY:
+      - 910c26028e7ba373-SEA
+      Connection:
+      - keep-alive
+      Content-Length:
+      - '272'
+      Content-Type:
+      - application/json; charset=utf-8
+      Date:
+      - Wed, 12 Feb 2025 11:13:17 GMT
+      Server:
+      - cloudflare
+      X-Content-Type-Options:
+      - nosniff
+      alt-svc:
+      - h3=":443"; ma=86400
+      strict-transport-security:
+      - max-age=31536000; includeSubDomains; preload
+      vary:
+      - Origin
+      x-request-id:
+      - req_6f0f7d9e36bbb3ffea4d672b30b700b9
+    http_version: HTTP/1.1
+    status_code: 401
+version: 1

tests/cassettes/test_hierarchical_tool_output_formatting[test ```result```-test ```result].yaml

@@ -0,0 +1,111 @@
+interactions:
+- request:
+    body: '{"messages": [{"role": "system", "content": "You are Crew Manager. You
+      are a seasoned manager with a knack for getting the best out of your team.\nYou
+      are also known for your ability to delegate work to the right people, and to
+      ask the right questions to get the best out of your team.\nEven though you don''t
+      perform tasks by yourself, you have a lot of experience in the field, which
+      allows you to properly evaluate the work of your team members.\nYour personal
+      goal is: Manage the team to complete the task in the best way possible.\nYou
+      ONLY have access to the following tools, and should NEVER make up tools that
+      are not listed here:\n\nTool Name: Delegate work to coworker\nTool Arguments:
+      {''task'': {''description'': ''The task to delegate'', ''type'': ''str''}, ''context'':
+      {''description'': ''The context for the task'', ''type'': ''str''}, ''coworker'':
+      {''description'': ''The role/name of the coworker to delegate to'', ''type'':
+      ''str''}}\nTool Description: Delegate a specific task to one of the following
+      coworkers: Researcher\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.\nTool Name: Ask question to coworker\nTool
+      Arguments: {''question'': {''description'': ''The question to ask'', ''type'':
+      ''str''}, ''context'': {''description'': ''The context for the question'', ''type'':
+      ''str''}, ''coworker'': {''description'': ''The role/name of the coworker to
+      ask'', ''type'': ''str''}}\nTool Description: Ask a specific question to one
+      of the following coworkers: Researcher\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.\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 [Delegate work
+      to coworker, Ask question to coworker], 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```"}, {"role": "user", "content": "\nCurrent Task: Test task
+      using test_tool\n\nThis is the expected criteria for your final answer: Test
+      output\nyou MUST return the actual complete content as the final answer, not
+      a summary.\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:"}], "model":
+      "gpt-4o", "stop": ["\nObservation:"]}'
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate
+      connection:
+      - keep-alive
+      content-length:
+      - '2915'
+      content-type:
+      - application/json
+      cookie:
+      - __cf_bm=0nI_3cs9LKHlz0_cVifl9lVXZWfNQrzhLRZzoaDCVJs-1739358791-1.0.1.1-GJZzVGs1fHQe9gucVHUCoDlI3mwg3JyXCBIIChx_OsB0jgqbzt1s2et96vUgpdjONt9C2tB8OQ6fk70k3vhh0w;
+        _cfuvid=TvEPcNq35qaIgBFFzOP7g7NQf632eeMrvRDJyr7UTro-1739358791985-0.0.1.1-604800000
+      host:
+      - api.openai.com
+      user-agent:
+      - OpenAI/Python 1.61.0
+      x-stainless-arch:
+      - x64
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - Linux
+      x-stainless-package-version:
+      - 1.61.0
+      x-stainless-raw-response:
+      - 'true'
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.12.7
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    content: "{\n    \"error\": {\n        \"message\": \"Incorrect API key provided:
+      sk-fake-**********2345. You can find your API key at https://platform.openai.com/account/api-keys.\",\n
+      \       \"type\": \"invalid_request_error\",\n        \"param\": null,\n        \"code\":
+      \"invalid_api_key\"\n    }\n}\n"
+    headers:
+      CF-RAY:
+      - 910c25f5e85ba373-SEA
+      Connection:
+      - keep-alive
+      Content-Length:
+      - '272'
+      Content-Type:
+      - application/json; charset=utf-8
+      Date:
+      - Wed, 12 Feb 2025 11:13:15 GMT
+      Server:
+      - cloudflare
+      X-Content-Type-Options:
+      - nosniff
+      alt-svc:
+      - h3=":443"; ma=86400
+      cf-cache-status:
+      - DYNAMIC
+      strict-transport-security:
+      - max-age=31536000; includeSubDomains; preload
+      vary:
+      - Origin
+      x-request-id:
+      - req_49a0c35d5999eb9f6d71fcea9fa3c68f
+    http_version: HTTP/1.1
+    status_code: 401
+version: 1

tests/cassettes/test_hierarchical_tool_output_formatting[test result-test result].yaml

@@ -0,0 +1,111 @@
+interactions:
+- request:
+    body: '{"messages": [{"role": "system", "content": "You are Crew Manager. You
+      are a seasoned manager with a knack for getting the best out of your team.\nYou
+      are also known for your ability to delegate work to the right people, and to
+      ask the right questions to get the best out of your team.\nEven though you don''t
+      perform tasks by yourself, you have a lot of experience in the field, which
+      allows you to properly evaluate the work of your team members.\nYour personal
+      goal is: Manage the team to complete the task in the best way possible.\nYou
+      ONLY have access to the following tools, and should NEVER make up tools that
+      are not listed here:\n\nTool Name: Delegate work to coworker\nTool Arguments:
+      {''task'': {''description'': ''The task to delegate'', ''type'': ''str''}, ''context'':
+      {''description'': ''The context for the task'', ''type'': ''str''}, ''coworker'':
+      {''description'': ''The role/name of the coworker to delegate to'', ''type'':
+      ''str''}}\nTool Description: Delegate a specific task to one of the following
+      coworkers: Researcher\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.\nTool Name: Ask question to coworker\nTool
+      Arguments: {''question'': {''description'': ''The question to ask'', ''type'':
+      ''str''}, ''context'': {''description'': ''The context for the question'', ''type'':
+      ''str''}, ''coworker'': {''description'': ''The role/name of the coworker to
+      ask'', ''type'': ''str''}}\nTool Description: Ask a specific question to one
+      of the following coworkers: Researcher\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.\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 [Delegate work
+      to coworker, Ask question to coworker], 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```"}, {"role": "user", "content": "\nCurrent Task: Test task
+      using test_tool\n\nThis is the expected criteria for your final answer: Test
+      output\nyou MUST return the actual complete content as the final answer, not
+      a summary.\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:"}], "model":
+      "gpt-4o", "stop": ["\nObservation:"]}'
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate
+      connection:
+      - keep-alive
+      content-length:
+      - '2915'
+      content-type:
+      - application/json
+      cookie:
+      - __cf_bm=0nI_3cs9LKHlz0_cVifl9lVXZWfNQrzhLRZzoaDCVJs-1739358791-1.0.1.1-GJZzVGs1fHQe9gucVHUCoDlI3mwg3JyXCBIIChx_OsB0jgqbzt1s2et96vUgpdjONt9C2tB8OQ6fk70k3vhh0w;
+        _cfuvid=TvEPcNq35qaIgBFFzOP7g7NQf632eeMrvRDJyr7UTro-1739358791985-0.0.1.1-604800000
+      host:
+      - api.openai.com
+      user-agent:
+      - OpenAI/Python 1.61.0
+      x-stainless-arch:
+      - x64
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - Linux
+      x-stainless-package-version:
+      - 1.61.0
+      x-stainless-raw-response:
+      - 'true'
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.12.7
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    content: "{\n    \"error\": {\n        \"message\": \"Incorrect API key provided:
+      sk-fake-**********2345. You can find your API key at https://platform.openai.com/account/api-keys.\",\n
+      \       \"type\": \"invalid_request_error\",\n        \"param\": null,\n        \"code\":
+      \"invalid_api_key\"\n    }\n}\n"
+    headers:
+      CF-RAY:
+      - 910c25f1be28a373-SEA
+      Connection:
+      - keep-alive
+      Content-Length:
+      - '272'
+      Content-Type:
+      - application/json; charset=utf-8
+      Date:
+      - Wed, 12 Feb 2025 11:13:14 GMT
+      Server:
+      - cloudflare
+      X-Content-Type-Options:
+      - nosniff
+      alt-svc:
+      - h3=":443"; ma=86400
+      cf-cache-status:
+      - DYNAMIC
+      strict-transport-security:
+      - max-age=31536000; includeSubDomains; preload
+      vary:
+      - Origin
+      x-request-id:
+      - req_573ccd417b499cf38012dec20261964d
+    http_version: HTTP/1.1
+    status_code: 401
+version: 1

tests/cassettes/test_hierarchical_tool_output_formatting[test result`-test result].yaml

@@ -0,0 +1,111 @@
+interactions:
+- request:
+    body: '{"messages": [{"role": "system", "content": "You are Crew Manager. You
+      are a seasoned manager with a knack for getting the best out of your team.\nYou
+      are also known for your ability to delegate work to the right people, and to
+      ask the right questions to get the best out of your team.\nEven though you don''t
+      perform tasks by yourself, you have a lot of experience in the field, which
+      allows you to properly evaluate the work of your team members.\nYour personal
+      goal is: Manage the team to complete the task in the best way possible.\nYou
+      ONLY have access to the following tools, and should NEVER make up tools that
+      are not listed here:\n\nTool Name: Delegate work to coworker\nTool Arguments:
+      {''task'': {''description'': ''The task to delegate'', ''type'': ''str''}, ''context'':
+      {''description'': ''The context for the task'', ''type'': ''str''}, ''coworker'':
+      {''description'': ''The role/name of the coworker to delegate to'', ''type'':
+      ''str''}}\nTool Description: Delegate a specific task to one of the following
+      coworkers: Researcher\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.\nTool Name: Ask question to coworker\nTool
+      Arguments: {''question'': {''description'': ''The question to ask'', ''type'':
+      ''str''}, ''context'': {''description'': ''The context for the question'', ''type'':
+      ''str''}, ''coworker'': {''description'': ''The role/name of the coworker to
+      ask'', ''type'': ''str''}}\nTool Description: Ask a specific question to one
+      of the following coworkers: Researcher\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.\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 [Delegate work
+      to coworker, Ask question to coworker], 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```"}, {"role": "user", "content": "\nCurrent Task: Test task
+      using test_tool\n\nThis is the expected criteria for your final answer: Test
+      output\nyou MUST return the actual complete content as the final answer, not
+      a summary.\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:"}], "model":
+      "gpt-4o", "stop": ["\nObservation:"]}'
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate
+      connection:
+      - keep-alive
+      content-length:
+      - '2915'
+      content-type:
+      - application/json
+      cookie:
+      - __cf_bm=0nI_3cs9LKHlz0_cVifl9lVXZWfNQrzhLRZzoaDCVJs-1739358791-1.0.1.1-GJZzVGs1fHQe9gucVHUCoDlI3mwg3JyXCBIIChx_OsB0jgqbzt1s2et96vUgpdjONt9C2tB8OQ6fk70k3vhh0w;
+        _cfuvid=TvEPcNq35qaIgBFFzOP7g7NQf632eeMrvRDJyr7UTro-1739358791985-0.0.1.1-604800000
+      host:
+      - api.openai.com
+      user-agent:
+      - OpenAI/Python 1.61.0
+      x-stainless-arch:
+      - x64
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - Linux
+      x-stainless-package-version:
+      - 1.61.0
+      x-stainless-raw-response:
+      - 'true'
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.12.7
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    content: "{\n    \"error\": {\n        \"message\": \"Incorrect API key provided:
+      sk-fake-**********2345. You can find your API key at https://platform.openai.com/account/api-keys.\",\n
+      \       \"type\": \"invalid_request_error\",\n        \"param\": null,\n        \"code\":
+      \"invalid_api_key\"\n    }\n}\n"
+    headers:
+      CF-RAY:
+      - 910c25e84886a373-SEA
+      Connection:
+      - keep-alive
+      Content-Length:
+      - '272'
+      Content-Type:
+      - application/json; charset=utf-8
+      Date:
+      - Wed, 12 Feb 2025 11:13:13 GMT
+      Server:
+      - cloudflare
+      X-Content-Type-Options:
+      - nosniff
+      alt-svc:
+      - h3=":443"; ma=86400
+      cf-cache-status:
+      - DYNAMIC
+      strict-transport-security:
+      - max-age=31536000; includeSubDomains; preload
+      vary:
+      - Origin
+      x-request-id:
+      - req_e48016407be31edd1f9ee2e1bb4d1b30
+    http_version: HTTP/1.1
+    status_code: 401
+version: 1

tests/cassettes/test_hierarchical_tool_output_formatting[test result```-test result].yaml

@@ -0,0 +1,114 @@
+interactions:
+- request:
+    body: '{"messages": [{"role": "system", "content": "You are Crew Manager. You
+      are a seasoned manager with a knack for getting the best out of your team.\nYou
+      are also known for your ability to delegate work to the right people, and to
+      ask the right questions to get the best out of your team.\nEven though you don''t
+      perform tasks by yourself, you have a lot of experience in the field, which
+      allows you to properly evaluate the work of your team members.\nYour personal
+      goal is: Manage the team to complete the task in the best way possible.\nYou
+      ONLY have access to the following tools, and should NEVER make up tools that
+      are not listed here:\n\nTool Name: Delegate work to coworker\nTool Arguments:
+      {''task'': {''description'': ''The task to delegate'', ''type'': ''str''}, ''context'':
+      {''description'': ''The context for the task'', ''type'': ''str''}, ''coworker'':
+      {''description'': ''The role/name of the coworker to delegate to'', ''type'':
+      ''str''}}\nTool Description: Delegate a specific task to one of the following
+      coworkers: Researcher\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.\nTool Name: Ask question to coworker\nTool
+      Arguments: {''question'': {''description'': ''The question to ask'', ''type'':
+      ''str''}, ''context'': {''description'': ''The context for the question'', ''type'':
+      ''str''}, ''coworker'': {''description'': ''The role/name of the coworker to
+      ask'', ''type'': ''str''}}\nTool Description: Ask a specific question to one
+      of the following coworkers: Researcher\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.\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 [Delegate work
+      to coworker, Ask question to coworker], 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```"}, {"role": "user", "content": "\nCurrent Task: Test task
+      using test_tool\n\nThis is the expected criteria for your final answer: Test
+      output\nyou MUST return the actual complete content as the final answer, not
+      a summary.\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:"}], "model":
+      "gpt-4o", "stop": ["\nObservation:"]}'
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate
+      connection:
+      - keep-alive
+      content-length:
+      - '2915'
+      content-type:
+      - application/json
+      host:
+      - api.openai.com
+      user-agent:
+      - OpenAI/Python 1.61.0
+      x-stainless-arch:
+      - x64
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - Linux
+      x-stainless-package-version:
+      - 1.61.0
+      x-stainless-raw-response:
+      - 'true'
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.12.7
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    content: "{\n    \"error\": {\n        \"message\": \"Incorrect API key provided:
+      sk-fake-**********2345. You can find your API key at https://platform.openai.com/account/api-keys.\",\n
+      \       \"type\": \"invalid_request_error\",\n        \"param\": null,\n        \"code\":
+      \"invalid_api_key\"\n    }\n}\n"
+    headers:
+      CF-RAY:
+      - 910c25e11d1da373-SEA
+      Connection:
+      - keep-alive
+      Content-Length:
+      - '272'
+      Content-Type:
+      - application/json; charset=utf-8
+      Date:
+      - Wed, 12 Feb 2025 11:13:11 GMT
+      Server:
+      - cloudflare
+      Set-Cookie:
+      - __cf_bm=0nI_3cs9LKHlz0_cVifl9lVXZWfNQrzhLRZzoaDCVJs-1739358791-1.0.1.1-GJZzVGs1fHQe9gucVHUCoDlI3mwg3JyXCBIIChx_OsB0jgqbzt1s2et96vUgpdjONt9C2tB8OQ6fk70k3vhh0w;
+        path=/; expires=Wed, 12-Feb-25 11:43:11 GMT; domain=.api.openai.com; HttpOnly;
+        Secure; SameSite=None
+      - _cfuvid=TvEPcNq35qaIgBFFzOP7g7NQf632eeMrvRDJyr7UTro-1739358791985-0.0.1.1-604800000;
+        path=/; domain=.api.openai.com; HttpOnly; Secure; SameSite=None
+      X-Content-Type-Options:
+      - nosniff
+      alt-svc:
+      - h3=":443"; ma=86400
+      cf-cache-status:
+      - DYNAMIC
+      strict-transport-security:
+      - max-age=31536000; includeSubDomains; preload
+      vary:
+      - Origin
+      x-request-id:
+      - req_e07e0bddd718b7d5978b6bd9f138fa40
+    http_version: HTTP/1.1
+    status_code: 401
+version: 1

tests/cassettes/test_hierarchical_tool_output_formatting[test result``````-test result].yaml

@@ -0,0 +1,111 @@
+interactions:
+- request:
+    body: '{"messages": [{"role": "system", "content": "You are Crew Manager. You
+      are a seasoned manager with a knack for getting the best out of your team.\nYou
+      are also known for your ability to delegate work to the right people, and to
+      ask the right questions to get the best out of your team.\nEven though you don''t
+      perform tasks by yourself, you have a lot of experience in the field, which
+      allows you to properly evaluate the work of your team members.\nYour personal
+      goal is: Manage the team to complete the task in the best way possible.\nYou
+      ONLY have access to the following tools, and should NEVER make up tools that
+      are not listed here:\n\nTool Name: Delegate work to coworker\nTool Arguments:
+      {''task'': {''description'': ''The task to delegate'', ''type'': ''str''}, ''context'':
+      {''description'': ''The context for the task'', ''type'': ''str''}, ''coworker'':
+      {''description'': ''The role/name of the coworker to delegate to'', ''type'':
+      ''str''}}\nTool Description: Delegate a specific task to one of the following
+      coworkers: Researcher\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.\nTool Name: Ask question to coworker\nTool
+      Arguments: {''question'': {''description'': ''The question to ask'', ''type'':
+      ''str''}, ''context'': {''description'': ''The context for the question'', ''type'':
+      ''str''}, ''coworker'': {''description'': ''The role/name of the coworker to
+      ask'', ''type'': ''str''}}\nTool Description: Ask a specific question to one
+      of the following coworkers: Researcher\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.\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 [Delegate work
+      to coworker, Ask question to coworker], 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```"}, {"role": "user", "content": "\nCurrent Task: Test task
+      using test_tool\n\nThis is the expected criteria for your final answer: Test
+      output\nyou MUST return the actual complete content as the final answer, not
+      a summary.\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:"}], "model":
+      "gpt-4o", "stop": ["\nObservation:"]}'
+    headers:
+      accept:
+      - application/json
+      accept-encoding:
+      - gzip, deflate
+      connection:
+      - keep-alive
+      content-length:
+      - '2915'
+      content-type:
+      - application/json
+      cookie:
+      - __cf_bm=0nI_3cs9LKHlz0_cVifl9lVXZWfNQrzhLRZzoaDCVJs-1739358791-1.0.1.1-GJZzVGs1fHQe9gucVHUCoDlI3mwg3JyXCBIIChx_OsB0jgqbzt1s2et96vUgpdjONt9C2tB8OQ6fk70k3vhh0w;
+        _cfuvid=TvEPcNq35qaIgBFFzOP7g7NQf632eeMrvRDJyr7UTro-1739358791985-0.0.1.1-604800000
+      host:
+      - api.openai.com
+      user-agent:
+      - OpenAI/Python 1.61.0
+      x-stainless-arch:
+      - x64
+      x-stainless-async:
+      - 'false'
+      x-stainless-lang:
+      - python
+      x-stainless-os:
+      - Linux
+      x-stainless-package-version:
+      - 1.61.0
+      x-stainless-raw-response:
+      - 'true'
+      x-stainless-retry-count:
+      - '0'
+      x-stainless-runtime:
+      - CPython
+      x-stainless-runtime-version:
+      - 3.12.7
+    method: POST
+    uri: https://api.openai.com/v1/chat/completions
+  response:
+    content: "{\n    \"error\": {\n        \"message\": \"Incorrect API key provided:
+      sk-fake-**********2345. You can find your API key at https://platform.openai.com/account/api-keys.\",\n
+      \       \"type\": \"invalid_request_error\",\n        \"param\": null,\n        \"code\":
+      \"invalid_api_key\"\n    }\n}\n"
+    headers:
+      CF-Cache-Status:
+      - DYNAMIC
+      CF-RAY:
+      - 910c25ec6a9fa373-SEA
+      Connection:
+      - keep-alive
+      Content-Length:
+      - '272'
+      Content-Type:
+      - application/json; charset=utf-8
+      Date:
+      - Wed, 12 Feb 2025 11:13:13 GMT
+      Server:
+      - cloudflare
+      X-Content-Type-Options:
+      - nosniff
+      alt-svc:
+      - h3=":443"; ma=86400
+      strict-transport-security:
+      - max-age=31536000; includeSubDomains; preload
+      vary:
+      - Origin
+      x-request-id:
+      - req_f9215c0efc514e3dfea73d6c5f49eb4e
+    http_version: HTTP/1.1
+    status_code: 401
+version: 1

tests/conftest.py

@@ -8,6 +8,11 @@ from dotenv import load_dotenv
 
 load_result = load_dotenv(override=True)
 
+@pytest.fixture(autouse=True)
+def mock_openai_key(monkeypatch):
+    """Mock OpenAI API key for VCR cassettes."""
+    monkeypatch.setenv("OPENAI_API_KEY", "sk-fake-test-key-12345")
+
 @pytest.fixture(autouse=True)
 def setup_test_environment():
     """Set up test environment with a temporary directory for SQLite storage."""

tests/crew_test.py

@@ -3,6 +3,7 @@
 import hashlib
 import json
 from concurrent.futures import Future
+from typing import Any
 from unittest import mock
 from unittest.mock import MagicMock, patch
 
@@ -14,7 +15,9 @@ from crewai.agent import Agent
 from crewai.agents.cache import CacheHandler
 from crewai.crew import Crew
 from crewai.crews.crew_output import CrewOutput
-from crewai.knowledge.source.string_knowledge_source import StringKnowledgeSource
+from crewai.knowledge.source.string_knowledge_source import (
+    StringKnowledgeSource,
+)
 from crewai.memory.contextual.contextual_memory import ContextualMemory
 from crewai.process import Process
 from crewai.project import crew
@@ -22,10 +25,13 @@ from crewai.task import Task
 from crewai.tasks.conditional_task import ConditionalTask
 from crewai.tasks.output_format import OutputFormat
 from crewai.tasks.task_output import TaskOutput
+from crewai.tools.base_tool import BaseTool
 from crewai.types.usage_metrics import UsageMetrics
 from crewai.utilities import Logger
 from crewai.utilities.rpm_controller import RPMController
-from crewai.utilities.task_output_storage_handler import TaskOutputStorageHandler
+from crewai.utilities.task_output_storage_handler import (
+    TaskOutputStorageHandler,
+)
 
 ceo = Agent(
     role="CEO",
@@ -49,6 +55,41 @@ writer = Agent(
 )
 
 
+def test_crew_with_only_conditional_tasks_raises_error():
+    """Test that creating a crew with only conditional tasks raises an error."""
+
+    def condition_func(task_output: TaskOutput) -> bool:
+        return True
+
+    conditional1 = ConditionalTask(
+        description="Conditional task 1",
+        expected_output="Output 1",
+        agent=researcher,
+        condition=condition_func,
+    )
+    conditional2 = ConditionalTask(
+        description="Conditional task 2",
+        expected_output="Output 2",
+        agent=researcher,
+        condition=condition_func,
+    )
+    conditional3 = ConditionalTask(
+        description="Conditional task 3",
+        expected_output="Output 3",
+        agent=researcher,
+        condition=condition_func,
+    )
+
+    with pytest.raises(
+        pydantic_core._pydantic_core.ValidationError,
+        match="Crew must include at least one non-conditional task",
+    ):
+        Crew(
+            agents=[researcher],
+            tasks=[conditional1, conditional2, conditional3],
+        )
+
+
 def test_crew_config_conditional_requirement():
     with pytest.raises(ValueError):
         Crew(process=Process.sequential)
@@ -280,6 +321,69 @@ def test_sync_task_execution():
         assert mock_execute_sync.call_count == len(tasks)
 
 
+@pytest.mark.vcr(
+    filter_headers=["authorization"],
+    record_mode="once",
+    decode_compressed_response=True,
+    ignore_localhost=True,
+    match_on=["method", "scheme", "host", "port", "path"]
+)
+@pytest.mark.parametrize(
+    "tool_output,expected",
+    [
+        ("test result```", "test result"),
+        ("test result`", "test result"),
+        ("test result``````", "test result"),
+        ("test result", "test result"),
+        ("test ```result```", "test ```result"),  # Only strip trailing backticks
+        (None, "None"),  # Test non-string input gets converted to string
+        (" ", " "),  # Test whitespace string
+        (
+            "malformed`result```test",
+            "malformed`result```test",
+        ),  # Test non-trailing backticks
+    ],
+)
+def test_hierarchical_tool_output_formatting(tool_output, expected):
+    """Test that tool outputs in hierarchical mode don't have extra backticks.
+
+    This test verifies that the tool output cleaning functionality correctly handles
+    various scenarios of backtick formatting, ensuring only trailing backticks are
+    removed while preserving any inline markdown formatting.
+    """
+
+    class TestTool(BaseTool):
+        name: str = "test_tool"
+        description: str = "A test tool"
+
+        def _run(self, *args: Any, **kwargs: Any) -> str:
+            return tool_output
+
+    task = Task(
+        description="Test task using test_tool",
+        expected_output="Test output",
+    )
+
+    crew = Crew(
+        agents=[researcher],
+        process=Process.hierarchical,
+        manager_llm="gpt-4o",
+        tasks=[task],
+        tools=[TestTool()],
+    )
+
+    with patch.object(
+        Task,
+        "execute_sync",
+        return_value=TaskOutput(
+            description="Test task", raw=expected, agent="researcher"
+        ),
+    ) as mock_execute_sync:
+        result = crew.kickoff()
+        assert mock_execute_sync.called
+        assert result.raw == expected
+
+
 @pytest.mark.vcr(filter_headers=["authorization"])
 def test_hierarchical_process():
     task = Task(
@@ -1919,6 +2023,7 @@ def test_task_callback_on_crew():
 
 def test_task_callback_both_on_task_and_crew():
     from unittest.mock import MagicMock, patch
+
     mock_callback_on_task = MagicMock()
     mock_callback_on_crew = MagicMock()
 
@@ -2060,6 +2165,210 @@ def test_tools_with_custom_caching():
             assert result.raw == "3"
 
 
+@pytest.mark.vcr(filter_headers=["authorization"])
+def test_conditional_task_uses_last_output():
+    """Test that conditional tasks use the last task output for condition evaluation."""
+    task1 = Task(
+        description="First task",
+        expected_output="First output",
+        agent=researcher,
+    )
+
+    def condition_fails(task_output: TaskOutput) -> bool:
+        # This condition will never be met
+        return "never matches" in task_output.raw.lower()
+
+    def condition_succeeds(task_output: TaskOutput) -> bool:
+        # This condition will match first task's output
+        return "first success" in task_output.raw.lower()
+
+    conditional_task1 = ConditionalTask(
+        description="Second task - conditional that fails condition",
+        expected_output="Second output",
+        agent=researcher,
+        condition=condition_fails,
+    )
+
+    conditional_task2 = ConditionalTask(
+        description="Third task - conditional that succeeds using first task output",
+        expected_output="Third output",
+        agent=writer,
+        condition=condition_succeeds,
+    )
+
+    crew = Crew(
+        agents=[researcher, writer],
+        tasks=[task1, conditional_task1, conditional_task2],
+    )
+
+    # Mock outputs for tasks
+    mock_first = TaskOutput(
+        description="First task output",
+        raw="First success output",  # Will be used by third task's condition
+        agent=researcher.role,
+    )
+    mock_third = TaskOutput(
+        description="Third task output",
+        raw="Third task executed",  # Output when condition succeeds using first task output
+        agent=writer.role,
+    )
+
+    # Set up mocks for task execution and conditional logic
+    with patch.object(ConditionalTask, "should_execute") as mock_should_execute:
+        # First conditional fails, second succeeds
+        mock_should_execute.side_effect = [False, True]
+        with patch.object(Task, "execute_sync") as mock_execute:
+            mock_execute.side_effect = [mock_first, mock_third]
+            result = crew.kickoff()
+
+            # Verify execution behavior
+            assert mock_execute.call_count == 2  # Only first and third tasks execute
+            assert mock_should_execute.call_count == 2  # Both conditionals checked
+
+            # Verify outputs collection:
+            # First executed task output, followed by an automatically generated (skipped) output, then the conditional execution
+            assert len(result.tasks_output) == 3
+            assert (
+                result.tasks_output[0].raw == "First success output"
+            )  # First task succeeded
+            assert (
+                result.tasks_output[1].raw == ""
+            )  # Second task skipped (condition failed)
+            assert (
+                result.tasks_output[2].raw == "Third task executed"
+            )  # Third task used first task's output
+
+
+@pytest.mark.vcr(filter_headers=["authorization"])
+def test_conditional_tasks_result_collection():
+    """Test that task outputs are properly collected based on execution status."""
+    task1 = Task(
+        description="Normal task that always executes",
+        expected_output="First output",
+        agent=researcher,
+    )
+
+    def condition_never_met(task_output: TaskOutput) -> bool:
+        return "never matches" in task_output.raw.lower()
+
+    def condition_always_met(task_output: TaskOutput) -> bool:
+        return "success" in task_output.raw.lower()
+
+    task2 = ConditionalTask(
+        description="Conditional task that never executes",
+        expected_output="Second output",
+        agent=researcher,
+        condition=condition_never_met,
+    )
+
+    task3 = ConditionalTask(
+        description="Conditional task that always executes",
+        expected_output="Third output",
+        agent=writer,
+        condition=condition_always_met,
+    )
+
+    crew = Crew(
+        agents=[researcher, writer],
+        tasks=[task1, task2, task3],
+    )
+
+    # Mock outputs for different execution paths
+    mock_success = TaskOutput(
+        description="Success output",
+        raw="Success output",  # Triggers third task's condition
+        agent=researcher.role,
+    )
+    mock_conditional = TaskOutput(
+        description="Conditional output",
+        raw="Conditional task executed",
+        agent=writer.role,
+    )
+
+    # Set up mocks for task execution and conditional logic
+    with patch.object(ConditionalTask, "should_execute") as mock_should_execute:
+        # First conditional fails, second succeeds
+        mock_should_execute.side_effect = [False, True]
+        with patch.object(Task, "execute_sync") as mock_execute:
+            mock_execute.side_effect = [mock_success, mock_conditional]
+            result = crew.kickoff()
+
+            # Verify execution behavior
+            assert mock_execute.call_count == 2  # Only first and third tasks execute
+            assert mock_should_execute.call_count == 2  # Both conditionals checked
+
+            # Verify task output collection:
+            # There should be three outputs: normal task, skipped conditional task (empty output),
+            # and the conditional task that executed.
+            assert len(result.tasks_output) == 3
+            assert (
+                result.tasks_output[0].raw == "Success output"
+            )  # Normal task executed
+            assert result.tasks_output[1].raw == ""  # Second task skipped
+            assert (
+                result.tasks_output[2].raw == "Conditional task executed"
+            )  # Third task executed
+
+            # Verify task output collection
+            assert len(result.tasks_output) == 3
+            assert (
+                result.tasks_output[0].raw == "Success output"
+            )  # Normal task executed
+            assert result.tasks_output[1].raw == ""  # Second task skipped
+            assert (
+                result.tasks_output[2].raw == "Conditional task executed"
+            )  # Third task executed
+
+
+@pytest.mark.vcr(filter_headers=["authorization"])
+def test_multiple_conditional_tasks():
+    """Test that having multiple conditional tasks in sequence works correctly."""
+    task1 = Task(
+        description="Initial research task",
+        expected_output="Research output",
+        agent=researcher,
+    )
+
+    def condition1(task_output: TaskOutput) -> bool:
+        return "success" in task_output.raw.lower()
+
+    def condition2(task_output: TaskOutput) -> bool:
+        return "proceed" in task_output.raw.lower()
+
+    task2 = ConditionalTask(
+        description="First conditional task",
+        expected_output="Conditional output 1",
+        agent=writer,
+        condition=condition1,
+    )
+
+    task3 = ConditionalTask(
+        description="Second conditional task",
+        expected_output="Conditional output 2",
+        agent=writer,
+        condition=condition2,
+    )
+
+    crew = Crew(
+        agents=[researcher, writer],
+        tasks=[task1, task2, task3],
+    )
+
+    # Mock different task outputs to test conditional logic
+    mock_success = TaskOutput(
+        description="Mock success",
+        raw="Success and proceed output",
+        agent=researcher.role,
+    )
+
+    # Set up mocks for task execution
+    with patch.object(Task, "execute_sync", return_value=mock_success) as mock_execute:
+        result = crew.kickoff()
+        # Verify all tasks were executed (no IndexError)
+        assert mock_execute.call_count == 3
+        assert len(result.tasks_output) == 3
+
+
 @pytest.mark.vcr(filter_headers=["authorization"])
 def test_using_contextual_memory():
     from unittest.mock import patch

tests/llm_test.py

@@ -286,6 +286,79 @@ def test_o3_mini_reasoning_effort_medium():
 
 
 @pytest.mark.vcr(filter_headers=["authorization"])
+@pytest.fixture
+def anthropic_llm():
+    """Fixture providing an Anthropic LLM instance."""
+    return LLM(model="anthropic/claude-3-sonnet")
+
+@pytest.fixture
+def system_message():
+    """Fixture providing a system message."""
+    return {"role": "system", "content": "test"}
+
+@pytest.fixture
+def user_message():
+    """Fixture providing a user message."""
+    return {"role": "user", "content": "test"}
+
+def test_anthropic_message_formatting_edge_cases(anthropic_llm):
+    """Test edge cases for Anthropic message formatting."""
+    # Test None messages
+    with pytest.raises(TypeError, match="Messages cannot be None"):
+        anthropic_llm._format_messages_for_provider(None)
+        
+    # Test empty message list
+    formatted = anthropic_llm._format_messages_for_provider([])
+    assert len(formatted) == 1
+    assert formatted[0]["role"] == "user"
+    assert formatted[0]["content"] == "."
+    
+    # Test invalid message format
+    with pytest.raises(TypeError, match="Invalid message format"):
+        anthropic_llm._format_messages_for_provider([{"invalid": "message"}])
+
+def test_anthropic_model_detection():
+    """Test Anthropic model detection with various formats."""
+    models = [
+        ("anthropic/claude-3", True),
+        ("claude-instant", True),
+        ("claude/v1", True),
+        ("gpt-4", False),
+        ("", False),
+        ("anthropomorphic", False),  # Should not match partial words
+    ]
+    
+    for model, expected in models:
+        llm = LLM(model=model)
+        assert llm.is_anthropic == expected, f"Failed for model: {model}"
+
+def test_anthropic_message_formatting(anthropic_llm, system_message, user_message):
+    """Test Anthropic message formatting with fixtures."""
+    # Test when first message is system
+    formatted = anthropic_llm._format_messages_for_provider([system_message])
+    assert len(formatted) == 2
+    assert formatted[0]["role"] == "user"
+    assert formatted[0]["content"] == "."
+    assert formatted[1] == system_message
+
+    # Test when first message is already user
+    formatted = anthropic_llm._format_messages_for_provider([user_message])
+    assert len(formatted) == 1
+    assert formatted[0] == user_message
+
+    # Test with empty message list
+    formatted = anthropic_llm._format_messages_for_provider([])
+    assert len(formatted) == 1
+    assert formatted[0]["role"] == "user"
+    assert formatted[0]["content"] == "."
+
+    # Test with non-Anthropic model (should not modify messages)
+    non_anthropic_llm = LLM(model="gpt-4")
+    formatted = non_anthropic_llm._format_messages_for_provider([system_message])
+    assert len(formatted) == 1
+    assert formatted[0] == system_message
+
+
 def test_deepseek_r1_with_open_router():
     if not os.getenv("OPEN_ROUTER_API_KEY"):
         pytest.skip("OPEN_ROUTER_API_KEY not set; skipping test.")