wip

* Clean up to match enterprise

* add additional params to LLM calls

* make sure additional params are getting passed to llm

* update docs

* drop print

docs/concepts/agents.mdx

@@ -43,7 +43,7 @@ Think of an agent as a specialized team member with specific skills, expertise,
 | **Max Retry Limit** _(optional)_        | `max_retry_limit`        | `int`                         | Maximum number of retries when an error occurs. Default is 2.                                                         |
 | **Respect Context Window** _(optional)_ | `respect_context_window` | `bool`                        | Keep messages under context window size by summarizing. Default is True.                                              |
 | **Code Execution Mode** _(optional)_    | `code_execution_mode`    | `Literal["safe", "unsafe"]`   | Mode for code execution: 'safe' (using Docker) or 'unsafe' (direct). Default is 'safe'.                               |
-| **Embedder Config** _(optional)_        | `embedder_config`        | `Optional[Dict[str, Any]]`    | Configuration for the embedder used by the agent.                                                                     |
+| **Embedder** _(optional)_               | `embedder`               | `Optional[Dict[str, Any]]`    | Configuration for the embedder used by the agent.                                                                     |
 | **Knowledge Sources** _(optional)_      | `knowledge_sources`      | `Optional[List[BaseKnowledgeSource]]` | Knowledge sources available to the agent.                                                                     |
 | **Use System Prompt** _(optional)_      | `use_system_prompt`      | `Optional[bool]`              | Whether to use system prompt (for o1 model support). Default is True.                                                 |
 
@@ -152,7 +152,7 @@ agent = Agent(
     use_system_prompt=True,  # Default: True
     tools=[SerperDevTool()],  # Optional: List of tools
     knowledge_sources=None,  # Optional: List of knowledge sources
-    embedder_config=None,  # Optional: Custom embedder configuration
+    embedder=None,  # Optional: Custom embedder configuration
     system_template=None,  # Optional: Custom system prompt template
     prompt_template=None,  # Optional: Custom prompt template
     response_template=None,  # Optional: Custom response template

docs/concepts/knowledge.mdx

@@ -324,6 +324,13 @@ agent = Agent(
     verbose=True,
     allow_delegation=False,
     llm=gemini_llm,
+    embedder={
+        "provider": "google",
+        "config": {
+            "model": "models/text-embedding-004",
+            "api_key": GEMINI_API_KEY,
+        }
+    }
 )
 
 task = Task(

docs/concepts/llms.mdx

@@ -465,11 +465,22 @@ Learn how to get the most out of your LLM configuration:
     # https://cloud.google.com/vertex-ai/generative-ai/docs/overview
     ```
 
+    ## GET CREDENTIALS 
+    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
+    vertex_credentials_json = json.dumps(vertex_credentials)
+
     Example usage:
     ```python Code
     llm = LLM(
         model="gemini/gemini-1.5-pro-latest",
-        temperature=0.7
+        temperature=0.7,
+        vertex_credentials=vertex_credentials_json
     )
     ```
   </Accordion>

docs/concepts/tasks.mdx

@@ -33,11 +33,12 @@ crew = Crew(
 | :------------------------------- | :---------------- | :---------------------------- | :------------------------------------------------------------------------------------------------------------------- |
 | **Description**                  | `description`     | `str`                         | A clear, concise statement of what the task entails.                                                                 |
 | **Expected Output**              | `expected_output` | `str`                         | A detailed description of what the task's completion looks like.                                                     |
-| **Name** _(optional)_           | `name`           | `Optional[str]`               | A name identifier for the task.                                                                                      |
-| **Agent** _(optional)_          | `agent`           | `Optional[BaseAgent]`         | The agent responsible for executing the task.                                                                        |
-| **Tools** _(optional)_          | `tools`           | `List[BaseTool]`             | The tools/resources the agent is limited to use for this task.                                                       |
+| **Name** _(optional)_            | `name`            | `Optional[str]`               | A name identifier for the task.                                                                                      |
+| **Agent** _(optional)_           | `agent`           | `Optional[BaseAgent]`         | The agent responsible for executing the task.                                                                        |
+| **Tools** _(optional)_           | `tools`           | `List[BaseTool]`              | The tools/resources the agent is limited to use for this task.                                                       |
 | **Context** _(optional)_         | `context`         | `Optional[List["Task"]]`      | Other tasks whose outputs will be used as context for this task.                                                     |
 | **Async Execution** _(optional)_ | `async_execution` | `Optional[bool]`              | Whether the task should be executed asynchronously. Defaults to False.                                               |
+| **Human Input** _(optional)_     | `human_input`     | `Optional[bool]`              | Whether the task should have a human review the final answer of the agent. Defaults to False.                        |
 | **Config** _(optional)_          | `config`          | `Optional[Dict[str, Any]]`    | Task-specific configuration parameters.                                                                              |
 | **Output File** _(optional)_     | `output_file`     | `Optional[str]`               | File path for storing the task output.                                                                               |
 | **Output JSON** _(optional)_     | `output_json`     | `Optional[Type[BaseModel]]`   | A Pydantic model to structure the JSON output.                                                                       |

pyproject.toml

@@ -12,7 +12,7 @@ dependencies = [
     "pydantic>=2.4.2",
     "openai>=1.13.3",
     "litellm==1.59.8",
-    "instructor>=1.3.3",
+    "instructor>=1.7.2",
     # Text Processing
     "pdfplumber>=0.11.4",
     "regex>=2024.9.11",

src/crewai/agent.py

@@ -61,6 +61,7 @@ class Agent(BaseAgent):
             tools: Tools at agents disposal
             step_callback: Callback to be executed after each step of the agent execution.
             knowledge_sources: Knowledge sources for the agent.
+            embedder: Embedder configuration for the agent.
     """
 
     _times_executed: int = PrivateAttr(default=0)
@@ -122,17 +123,10 @@ class Agent(BaseAgent):
         default="safe",
         description="Mode for code execution: 'safe' (using Docker) or 'unsafe' (direct execution).",
     )
-    embedder_config: Optional[Dict[str, Any]] = Field(
+    embedder: Optional[Dict[str, Any]] = Field(
         default=None,
         description="Embedder configuration for the agent.",
     )
-    knowledge_sources: Optional[List[BaseKnowledgeSource]] = Field(
-        default=None,
-        description="Knowledge sources for the agent.",
-    )
-    _knowledge: Optional[Knowledge] = PrivateAttr(
-        default=None,
-    )
 
     @model_validator(mode="after")
     def post_init_setup(self):
@@ -163,10 +157,11 @@ class Agent(BaseAgent):
                 if isinstance(self.knowledge_sources, list) and all(
                     isinstance(k, BaseKnowledgeSource) for k in self.knowledge_sources
                 ):
-                    self._knowledge = Knowledge(
+                    self.knowledge = Knowledge(
                         sources=self.knowledge_sources,
-                        embedder_config=self.embedder_config,
+                        embedder=self.embedder,
                         collection_name=knowledge_agent_name,
+                        storage=self.knowledge_storage or None,
                     )
         except (TypeError, ValueError) as e:
             raise ValueError(f"Invalid Knowledge Configuration: {str(e)}")
@@ -225,8 +220,8 @@ class Agent(BaseAgent):
             if memory.strip() != "":
                 task_prompt += self.i18n.slice("memory").format(memory=memory)
 
-        if self._knowledge:
-            agent_knowledge_snippets = self._knowledge.query([task.prompt()])
+        if self.knowledge:
+            agent_knowledge_snippets = self.knowledge.query([task.prompt()])
             if agent_knowledge_snippets:
                 agent_knowledge_context = extract_knowledge_context(
                     agent_knowledge_snippets

src/crewai/agents/agent_builder/base_agent.py

@@ -18,6 +18,8 @@ from pydantic_core import PydanticCustomError
 from crewai.agents.agent_builder.utilities.base_token_process import TokenProcess
 from crewai.agents.cache.cache_handler import CacheHandler
 from crewai.agents.tools_handler import ToolsHandler
+from crewai.knowledge.knowledge import Knowledge
+from crewai.knowledge.source.base_knowledge_source import BaseKnowledgeSource
 from crewai.tools import BaseTool
 from crewai.tools.base_tool import Tool
 from crewai.utilities import I18N, Logger, RPMController
@@ -48,6 +50,8 @@ class BaseAgent(ABC, BaseModel):
         cache_handler (InstanceOf[CacheHandler]): An instance of the CacheHandler class.
         tools_handler (InstanceOf[ToolsHandler]): An instance of the ToolsHandler class.
         max_tokens: Maximum number of tokens for the agent to generate in a response.
+        knowledge_sources: Knowledge sources for the agent.
+        knowledge_storage: Custom knowledge storage for the agent.
 
 
     Methods:
@@ -130,6 +134,17 @@ class BaseAgent(ABC, BaseModel):
     max_tokens: Optional[int] = Field(
         default=None, description="Maximum number of tokens for the agent's execution."
     )
+    knowledge: Optional[Knowledge] = Field(
+        default=None, description="Knowledge for the agent."
+    )
+    knowledge_sources: Optional[List[BaseKnowledgeSource]] = Field(
+        default=None,
+        description="Knowledge sources for the agent.",
+    )
+    knowledge_storage: Optional[Any] = Field(
+        default=None,
+        description="Custom knowledge storage for the agent.",
+    )
 
     @model_validator(mode="before")
     @classmethod
@@ -256,13 +271,44 @@ class BaseAgent(ABC, BaseModel):
             "tools_handler",
             "cache_handler",
             "llm",
+            "knowledge_sources",
+            "knowledge_storage",
+            "knowledge",
         }
 
-        # Copy llm and clear callbacks
+        # Copy llm
         existing_llm = shallow_copy(self.llm)
+        copied_knowledge = shallow_copy(self.knowledge)
+        copied_knowledge_storage = shallow_copy(self.knowledge_storage)
+        # Properly copy knowledge sources if they exist
+        existing_knowledge_sources = None
+        if self.knowledge_sources:
+            # Create a shared storage instance for all knowledge sources
+            shared_storage = (
+                self.knowledge_sources[0].storage if self.knowledge_sources else None
+            )
+
+            existing_knowledge_sources = []
+            for source in self.knowledge_sources:
+                copied_source = (
+                    source.model_copy()
+                    if hasattr(source, "model_copy")
+                    else shallow_copy(source)
+                )
+                # Ensure all copied sources use the same storage instance
+                copied_source.storage = shared_storage
+                existing_knowledge_sources.append(copied_source)
+
         copied_data = self.model_dump(exclude=exclude)
         copied_data = {k: v for k, v in copied_data.items() if v is not None}
-        copied_agent = type(self)(**copied_data, llm=existing_llm, tools=self.tools)
+        copied_agent = type(self)(
+            **copied_data,
+            llm=existing_llm,
+            tools=self.tools,
+            knowledge_sources=existing_knowledge_sources,
+            knowledge=copied_knowledge,
+            knowledge_storage=copied_knowledge_storage,
+        )
 
         return copied_agent
 

src/crewai/agents/agent_builder/base_agent_executor_mixin.py

@@ -95,18 +95,29 @@ class CrewAgentExecutorMixin:
                 pass
 
     def _ask_human_input(self, final_answer: str) -> str:
-        """Prompt human input for final decision making."""
+        """Prompt human input with mode-appropriate messaging."""
         self._printer.print(
             content=f"\033[1m\033[95m ## Final Result:\033[00m \033[92m{final_answer}\033[00m"
         )
 
-        self._printer.print(
-            content=(
+        # Training mode prompt (single iteration)
+        if self.crew and getattr(self.crew, "_train", False):
+            prompt = (
                 "\n\n=====\n"
-                "## Please provide feedback on the Final Result and the Agent's actions. "
-                "Respond with 'looks good' or a similar phrase when you're satisfied.\n"
+                "## TRAINING MODE: Provide feedback to improve the agent's performance.\n"
+                "This will be used to train better versions of the agent.\n"
+                "Please provide detailed feedback about the result quality and reasoning process.\n"
                 "=====\n"
-            ),
-            color="bold_yellow",
-        )
+            )
+        # Regular human-in-the-loop prompt (multiple iterations)
+        else:
+            prompt = (
+                "\n\n=====\n"
+                "## HUMAN FEEDBACK: Provide feedback on the Final Result and Agent's actions.\n"
+                "Respond with 'looks good' to accept or provide specific improvement requests.\n"
+                "You can provide multiple rounds of feedback until satisfied.\n"
+                "=====\n"
+            )
+
+        self._printer.print(content=prompt, color="bold_yellow")
         return input()

src/crewai/agents/crew_agent_executor.py

@@ -100,6 +100,12 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
 
         try:
             formatted_answer = self._invoke_loop()
+        except AssertionError:
+            self._printer.print(
+                content="Agent failed to reach a final answer. This is likely a bug - please report it.",
+                color="red",
+            )
+            raise
         except Exception as e:
             if e.__class__.__module__.startswith("litellm"):
                 # Do not retry on litellm errors
@@ -115,7 +121,7 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
         self._create_long_term_memory(formatted_answer)
         return {"output": formatted_answer.output}
 
-    def _invoke_loop(self):
+    def _invoke_loop(self) -> AgentFinish:
         """
         Main loop to invoke the agent's thought process until it reaches a conclusion
         or the maximum number of iterations is reached.
@@ -161,6 +167,11 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
             finally:
                 self.iterations += 1
 
+        # During the invoke loop, formatted_answer alternates between AgentAction
+        # (when the agent is using tools) and eventually becomes AgentFinish
+        # (when the agent reaches a final answer). This assertion confirms we've
+        # reached a final answer and helps type checking understand this transition.
+        assert isinstance(formatted_answer, AgentFinish)
         self._show_logs(formatted_answer)
         return formatted_answer
 
@@ -292,8 +303,11 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
             self._printer.print(
                 content=f"\033[1m\033[95m# Agent:\033[00m \033[1m\033[92m{agent_role}\033[00m"
             )
+            description = (
+                getattr(self.task, "description") if self.task else "Not Found"
+            )
             self._printer.print(
-                content=f"\033[95m## Task:\033[00m \033[92m{self.task.description}\033[00m"
+                content=f"\033[95m## Task:\033[00m \033[92m{description}\033[00m"
             )
 
     def _show_logs(self, formatted_answer: Union[AgentAction, AgentFinish]):
@@ -418,58 +432,50 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
             )
 
     def _handle_crew_training_output(
-        self, result: AgentFinish, human_feedback: str | None = None
+        self, result: AgentFinish, human_feedback: Optional[str] = None
     ) -> None:
-        """Function to handle the process of the training data."""
+        """Handle the process of saving training data."""
         agent_id = str(self.agent.id)  # type: ignore
+        train_iteration = (
+            getattr(self.crew, "_train_iteration", None) if self.crew else None
+        )
+
+        if train_iteration is None or not isinstance(train_iteration, int):
+            self._printer.print(
+                content="Invalid or missing train iteration. Cannot save training data.",
+                color="red",
+            )
+            return
 
-        # Load training data
         training_handler = CrewTrainingHandler(TRAINING_DATA_FILE)
-        training_data = training_handler.load()
+        training_data = training_handler.load() or {}
 
-        # Check if training data exists, human input is not requested, and self.crew is valid
-        if training_data and not self.ask_for_human_input:
-            if self.crew is not None and hasattr(self.crew, "_train_iteration"):
-                train_iteration = self.crew._train_iteration
-                if agent_id in training_data and isinstance(train_iteration, int):
-                    training_data[agent_id][train_iteration][
-                        "improved_output"
-                    ] = result.output
-                    training_handler.save(training_data)
-                else:
-                    self._printer.print(
-                        content="Invalid train iteration type or agent_id not in training data.",
-                        color="red",
-                    )
-            else:
-                self._printer.print(
-                    content="Crew is None or does not have _train_iteration attribute.",
-                    color="red",
-                )
+        # Initialize or retrieve agent's training data
+        agent_training_data = training_data.get(agent_id, {})
 
-        if self.ask_for_human_input and human_feedback is not None:
-            training_data = {
+        if human_feedback is not None:
+            # Save initial output and human feedback
+            agent_training_data[train_iteration] = {
                 "initial_output": result.output,
                 "human_feedback": human_feedback,
-                "agent": agent_id,
-                "agent_role": self.agent.role,  # type: ignore
             }
-            if self.crew is not None and hasattr(self.crew, "_train_iteration"):
-                train_iteration = self.crew._train_iteration
-                if isinstance(train_iteration, int):
-                    CrewTrainingHandler(TRAINING_DATA_FILE).append(
-                        train_iteration, agent_id, training_data
-                    )
-                else:
-                    self._printer.print(
-                        content="Invalid train iteration type. Expected int.",
-                        color="red",
-                    )
+        else:
+            # Save improved output
+            if train_iteration in agent_training_data:
+                agent_training_data[train_iteration]["improved_output"] = result.output
             else:
                 self._printer.print(
-                    content="Crew is None or does not have _train_iteration attribute.",
+                    content=(
+                        f"No existing training data for agent {agent_id} and iteration "
+                        f"{train_iteration}. Cannot save improved output."
+                    ),
                     color="red",
                 )
+                return
+
+        # Update the training data and save
+        training_data[agent_id] = agent_training_data
+        training_handler.save(training_data)
 
     def _format_prompt(self, prompt: str, inputs: Dict[str, str]) -> str:
         prompt = prompt.replace("{input}", inputs["input"])
@@ -485,82 +491,111 @@ class CrewAgentExecutor(CrewAgentExecutorMixin):
         return {"role": role, "content": prompt}
 
     def _handle_human_feedback(self, formatted_answer: AgentFinish) -> AgentFinish:
-        """
-        Handles the human feedback loop, allowing the user to provide feedback
-        on the agent's output and determining if additional iterations are needed.
+        """Handle human feedback with different flows for training vs regular use.
 
-        Parameters:
-            formatted_answer (AgentFinish): The initial output from the agent.
+        Args:
+            formatted_answer: The initial AgentFinish result to get feedback on
 
         Returns:
-            AgentFinish: The final output after incorporating human feedback.
+            AgentFinish: The final answer after processing feedback
         """
+        human_feedback = self._ask_human_input(formatted_answer.output)
+
+        if self._is_training_mode():
+            return self._handle_training_feedback(formatted_answer, human_feedback)
+
+        return self._handle_regular_feedback(formatted_answer, human_feedback)
+
+    def _is_training_mode(self) -> bool:
+        """Check if crew is in training mode."""
+        return bool(self.crew and self.crew._train)
+
+    def _handle_training_feedback(
+        self, initial_answer: AgentFinish, feedback: str
+    ) -> AgentFinish:
+        """Process feedback for training scenarios with single iteration."""
+        self._printer.print(
+            content="\nProcessing training feedback.\n",
+            color="yellow",
+        )
+        self._handle_crew_training_output(initial_answer, feedback)
+        self.messages.append(
+            self._format_msg(
+                self._i18n.slice("feedback_instructions").format(feedback=feedback)
+            )
+        )
+        improved_answer = self._invoke_loop()
+        self._handle_crew_training_output(improved_answer)
+        self.ask_for_human_input = False
+        return improved_answer
+
+    def _handle_regular_feedback(
+        self, current_answer: AgentFinish, initial_feedback: str
+    ) -> AgentFinish:
+        """Process feedback for regular use with potential multiple iterations."""
+        feedback = initial_feedback
+        answer = current_answer
+
         while self.ask_for_human_input:
-            human_feedback = self._ask_human_input(formatted_answer.output)
+            response = self._get_llm_feedback_response(feedback)
 
-            if self.crew and self.crew._train:
-                self._handle_crew_training_output(formatted_answer, human_feedback)
-
-            # Make an LLM call to verify if additional changes are requested based on human feedback
-            additional_changes_prompt = self._i18n.slice(
-                "human_feedback_classification"
-            ).format(feedback=human_feedback)
-
-            retry_count = 0
-            llm_call_successful = False
-            additional_changes_response = None
-
-            while retry_count < MAX_LLM_RETRY and not llm_call_successful:
-                try:
-                    additional_changes_response = (
-                        self.llm.call(
-                            [
-                                self._format_msg(
-                                    additional_changes_prompt, role="system"
-                                )
-                            ],
-                            callbacks=self.callbacks,
-                        )
-                        .strip()
-                        .lower()
-                    )
-                    llm_call_successful = True
-                except Exception as e:
-                    retry_count += 1
-
-                    self._printer.print(
-                        content=f"Error during LLM call to classify human feedback: {e}. Retrying... ({retry_count}/{MAX_LLM_RETRY})",
-                        color="red",
-                    )
-
-            if not llm_call_successful:
-                self._printer.print(
-                    content="Error processing feedback after multiple attempts.",
-                    color="red",
-                )
+            if not self._feedback_requires_changes(response):
                 self.ask_for_human_input = False
-                break
-
-            if additional_changes_response == "false":
-                self.ask_for_human_input = False
-            elif additional_changes_response == "true":
-                self.ask_for_human_input = True
-                # Add human feedback to messages
-                self.messages.append(self._format_msg(f"Feedback: {human_feedback}"))
-                # Invoke the loop again with updated messages
-                formatted_answer = self._invoke_loop()
-
-                if self.crew and self.crew._train:
-                    self._handle_crew_training_output(formatted_answer)
             else:
-                # Unexpected response
-                self._printer.print(
-                    content=f"Unexpected response from LLM: '{additional_changes_response}'. Assuming no additional changes requested.",
-                    color="red",
-                )
-                self.ask_for_human_input = False
+                answer = self._process_feedback_iteration(feedback)
+                feedback = self._ask_human_input(answer.output)
 
-        return formatted_answer
+        return answer
+
+    def _get_llm_feedback_response(self, feedback: str) -> Optional[str]:
+        """Get LLM classification of whether feedback requires changes."""
+        prompt = self._i18n.slice("human_feedback_classification").format(
+            feedback=feedback
+        )
+        message = self._format_msg(prompt, role="system")
+
+        for retry in range(MAX_LLM_RETRY):
+            try:
+                response = self.llm.call([message], callbacks=self.callbacks)
+                return response.strip().lower() if response else None
+            except Exception as error:
+                self._log_feedback_error(retry, error)
+
+        self._log_max_retries_exceeded()
+        return None
+
+    def _feedback_requires_changes(self, response: Optional[str]) -> bool:
+        """Determine if feedback response indicates need for changes."""
+        return response == "true" if response else False
+
+    def _process_feedback_iteration(self, feedback: str) -> AgentFinish:
+        """Process a single feedback iteration."""
+        self.messages.append(
+            self._format_msg(
+                self._i18n.slice("feedback_instructions").format(feedback=feedback)
+            )
+        )
+        return self._invoke_loop()
+
+    def _log_feedback_error(self, retry_count: int, error: Exception) -> None:
+        """Log feedback processing errors."""
+        self._printer.print(
+            content=(
+                f"Error processing feedback: {error}. "
+                f"Retrying... ({retry_count + 1}/{MAX_LLM_RETRY})"
+            ),
+            color="red",
+        )
+
+    def _log_max_retries_exceeded(self) -> None:
+        """Log when max retries for feedback processing are exceeded."""
+        self._printer.print(
+            content=(
+                f"Failed to process feedback after {MAX_LLM_RETRY} attempts. "
+                "Ending feedback loop."
+            ),
+            color="red",
+        )
 
     def _handle_max_iterations_exceeded(self, formatted_answer):
         """

src/crewai/crew.py

@@ -4,6 +4,7 @@ import re
 import uuid
 import warnings
 from concurrent.futures import Future
+from copy import copy as shallow_copy
 from hashlib import md5
 from typing import Any, Callable, Dict, List, Optional, Set, Tuple, Union
 
@@ -210,8 +211,9 @@ class Crew(BaseModel):
         default=None,
         description="LLM used to handle chatting with the crew.",
     )
-    _knowledge: Optional[Knowledge] = PrivateAttr(
+    knowledge: Optional[Knowledge] = Field(
         default=None,
+        description="Knowledge for the crew.",
     )
 
     @field_validator("id", mode="before")
@@ -289,7 +291,7 @@ class Crew(BaseModel):
                 if isinstance(self.knowledge_sources, list) and all(
                     isinstance(k, BaseKnowledgeSource) for k in self.knowledge_sources
                 ):
-                    self._knowledge = Knowledge(
+                    self.knowledge = Knowledge(
                         sources=self.knowledge_sources,
                         embedder_config=self.embedder,
                         collection_name="crew",
@@ -492,21 +494,26 @@ class Crew(BaseModel):
         train_crew = self.copy()
         train_crew._setup_for_training(filename)
 
-        for n_iteration in range(n_iterations):
-            train_crew._train_iteration = n_iteration
-            train_crew.kickoff(inputs=inputs)
+        try:
+            for n_iteration in range(n_iterations):
+                train_crew._train_iteration = n_iteration
+                train_crew.kickoff(inputs=inputs)
 
-        training_data = CrewTrainingHandler(TRAINING_DATA_FILE).load()
+            training_data = CrewTrainingHandler(TRAINING_DATA_FILE).load()
 
-        for agent in train_crew.agents:
-            if training_data.get(str(agent.id)):
-                result = TaskEvaluator(agent).evaluate_training_data(
-                    training_data=training_data, agent_id=str(agent.id)
-                )
-
-                CrewTrainingHandler(filename).save_trained_data(
-                    agent_id=str(agent.role), trained_data=result.model_dump()
-                )
+            for agent in train_crew.agents:
+                if training_data.get(str(agent.id)):
+                    result = TaskEvaluator(agent).evaluate_training_data(
+                        training_data=training_data, agent_id=str(agent.id)
+                    )
+                    CrewTrainingHandler(filename).save_trained_data(
+                        agent_id=str(agent.role), trained_data=result.model_dump()
+                    )
+        except Exception as e:
+            self._logger.log("error", f"Training failed: {e}", color="red")
+            CrewTrainingHandler(TRAINING_DATA_FILE).clear()
+            CrewTrainingHandler(filename).clear()
+            raise
 
     def kickoff(
         self,
@@ -991,8 +998,8 @@ class Crew(BaseModel):
         return result
 
     def query_knowledge(self, query: List[str]) -> Union[List[Dict[str, Any]], None]:
-        if self._knowledge:
-            return self._knowledge.query(query)
+        if self.knowledge:
+            return self.knowledge.query(query)
         return None
 
     def fetch_inputs(self) -> Set[str]:
@@ -1036,6 +1043,8 @@ class Crew(BaseModel):
             "_telemetry",
             "agents",
             "tasks",
+            "knowledge_sources",
+            "knowledge",
         }
 
         cloned_agents = [agent.copy() for agent in self.agents]
@@ -1043,6 +1052,9 @@ class Crew(BaseModel):
         task_mapping = {}
 
         cloned_tasks = []
+        existing_knowledge_sources = shallow_copy(self.knowledge_sources)
+        existing_knowledge = shallow_copy(self.knowledge)
+
         for task in self.tasks:
             cloned_task = task.copy(cloned_agents, task_mapping)
             cloned_tasks.append(cloned_task)
@@ -1062,7 +1074,13 @@ class Crew(BaseModel):
         copied_data.pop("agents", None)
         copied_data.pop("tasks", None)
 
-        copied_crew = Crew(**copied_data, agents=cloned_agents, tasks=cloned_tasks)
+        copied_crew = Crew(
+            **copied_data,
+            agents=cloned_agents,
+            tasks=cloned_tasks,
+            knowledge_sources=existing_knowledge_sources,
+            knowledge=existing_knowledge,
+        )
 
         return copied_crew
 

src/crewai/knowledge/knowledge.py

@@ -15,20 +15,20 @@ class Knowledge(BaseModel):
     Args:
         sources: List[BaseKnowledgeSource] = Field(default_factory=list)
         storage: Optional[KnowledgeStorage] = Field(default=None)
-        embedder_config: Optional[Dict[str, Any]] = None
+        embedder: Optional[Dict[str, Any]] = None
     """
 
     sources: List[BaseKnowledgeSource] = Field(default_factory=list)
     model_config = ConfigDict(arbitrary_types_allowed=True)
     storage: Optional[KnowledgeStorage] = Field(default=None)
-    embedder_config: Optional[Dict[str, Any]] = None
+    embedder: Optional[Dict[str, Any]] = None
     collection_name: Optional[str] = None
 
     def __init__(
         self,
         collection_name: str,
         sources: List[BaseKnowledgeSource],
-        embedder_config: Optional[Dict[str, Any]] = None,
+        embedder: Optional[Dict[str, Any]] = None,
         storage: Optional[KnowledgeStorage] = None,
         **data,
     ):
@@ -37,25 +37,23 @@ class Knowledge(BaseModel):
             self.storage = storage
         else:
             self.storage = KnowledgeStorage(
-                embedder_config=embedder_config, collection_name=collection_name
+                embedder=embedder, collection_name=collection_name
             )
         self.sources = sources
         self.storage.initialize_knowledge_storage()
-        for source in sources:
-            source.storage = self.storage
-            source.add()
+        self._add_sources()
 
     def query(self, query: List[str], limit: int = 3) -> List[Dict[str, Any]]:
         """
         Query across all knowledge sources to find the most relevant information.
         Returns the top_k most relevant chunks.
-        
+
         Raises:
             ValueError: If storage is not initialized.
         """
         if self.storage is None:
             raise ValueError("Storage is not initialized.")
-            
+
         results = self.storage.search(
             query,
             limit,
@@ -63,6 +61,9 @@ class Knowledge(BaseModel):
         return results
 
     def _add_sources(self):
-        for source in self.sources:
-            source.storage = self.storage
-            source.add()
+        try:
+            for source in self.sources:
+                source.storage = self.storage
+                source.add()
+        except Exception as e:
+            raise e

src/crewai/knowledge/source/base_file_knowledge_source.py

@@ -29,7 +29,13 @@ class BaseFileKnowledgeSource(BaseKnowledgeSource, ABC):
     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:
+        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
 

src/crewai/knowledge/storage/knowledge_storage.py

@@ -48,11 +48,11 @@ class KnowledgeStorage(BaseKnowledgeStorage):
 
     def __init__(
         self,
-        embedder_config: Optional[Dict[str, Any]] = None,
+        embedder: Optional[Dict[str, Any]] = None,
         collection_name: Optional[str] = None,
     ):
         self.collection_name = collection_name
-        self._set_embedder_config(embedder_config)
+        self._set_embedder_config(embedder)
 
     def search(
         self,
@@ -99,7 +99,7 @@ class KnowledgeStorage(BaseKnowledgeStorage):
             )
             if self.app:
                 self.collection = self.app.get_or_create_collection(
-                    name=collection_name, embedding_function=self.embedder_config
+                    name=collection_name, embedding_function=self.embedder
                 )
             else:
                 raise Exception("Vector Database Client not initialized")
@@ -187,17 +187,15 @@ class KnowledgeStorage(BaseKnowledgeStorage):
             api_key=os.getenv("OPENAI_API_KEY"), model_name="text-embedding-3-small"
         )
 
-    def _set_embedder_config(
-        self, embedder_config: Optional[Dict[str, Any]] = None
-    ) -> None:
+    def _set_embedder_config(self, embedder: Optional[Dict[str, Any]] = None) -> None:
         """Set the embedding configuration for the knowledge storage.
 
         Args:
             embedder_config (Optional[Dict[str, Any]]): Configuration dictionary for the embedder.
                 If None or empty, defaults to the default embedding function.
         """
-        self.embedder_config = (
-            EmbeddingConfigurator().configure_embedder(embedder_config)
-            if embedder_config
+        self.embedder = (
+            EmbeddingConfigurator().configure_embedder(embedder)
+            if embedder
             else self._create_default_embedding_function()
         )

src/crewai/llm.py

@@ -7,7 +7,10 @@ import warnings
 from contextlib import contextmanager
 from typing import Any, Dict, List, Optional, Union, cast
 
+import instructor
 from dotenv import load_dotenv
+from openai.types.chat import ChatCompletionMessageParam
+from pydantic import BaseModel
 
 with warnings.catch_warnings():
     warnings.simplefilter("ignore", UserWarning)
@@ -133,9 +136,11 @@ class LLM:
         logprobs: Optional[int] = None,
         top_logprobs: Optional[int] = None,
         base_url: Optional[str] = None,
+        api_base: Optional[str] = None,
         api_version: Optional[str] = None,
         api_key: Optional[str] = None,
         callbacks: List[Any] = [],
+        **kwargs,
     ):
         self.model = model
         self.timeout = timeout
@@ -152,10 +157,12 @@ class LLM:
         self.logprobs = logprobs
         self.top_logprobs = top_logprobs
         self.base_url = base_url
+        self.api_base = api_base
         self.api_version = api_version
         self.api_key = api_key
         self.callbacks = callbacks
         self.context_window_size = 0
+        self.additional_params = kwargs
 
         litellm.drop_params = True
 
@@ -178,34 +185,63 @@ class LLM:
         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
+        High-level LLM call method that handles:
+        1. Multiple input formats (string or message list)
+        2. Structured responses via Instructor integration
+        3. Tool/function calling with optional structured output
+        4. Callback integration
 
         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.
+            messages: Input prompt(s) as either:
+                - String (converted to single user message)
+                - List of message dicts with 'role' and 'content'
+
+            tools: List of tool schemas for function calling
+            callbacks: List of callback handlers
+            available_functions: Mapping of function names to callables
+            response_format: Pydantic model for structured responses
 
         Returns:
-        - str: The final text response from the LLM or the result of a tool function call.
+            str: Can be:
+            - Plain text response
+            - Structured response (if response_format provided)
+            - Tool function result (raw or structured)
+
+        Behavior:
+        - With response_format and no tools: Direct structured response
+        - With tools: Initial LLM call → Tool execution → Optional secondary structured call
+        - Without tools/response_format: Standard text completion
 
         Examples:
-        ---------
-        # Example 1: Using a string input
-        response = llm.call("Return the name of a random city in the world.")
-        print(response)
+            # Basic text completion
+            llm.call("Hello world")
 
-        # Example 2: Using a list of messages
-        messages = [{"role": "user", "content": "What is the capital of France?"}]
-        response = llm.call(messages)
-        print(response)
+            # Structured response without tools
+            class City(BaseModel):
+                name: str
+                population: int
+
+            response = llm.call(
+                "Name a major US city",
+                response_format=City
+            )
+            print(response.name)  # Structured access
+
+            # Tool usage with raw output
+            llm.call(
+                "What's 5 squared?",
+                tools=[math_tools],
+                available_functions={"square": square_number}
+            )
+
+            # Tool usage with structured output
+            response = llm.call(
+                "Analyze this data",
+                tools=[data_tools],
+                available_functions={"analyze": analyze_data},
+                response_format=AnalysisResult
+            )
+            print(response.metrics)  # Structured access
         """
         if isinstance(messages, str):
             messages = [{"role": "user", "content": messages}]
@@ -214,35 +250,54 @@ class LLM:
             if callbacks and len(callbacks) > 0:
                 self.set_callbacks(callbacks)
 
+            # Prepare the parameters for the completion call.
+            params = {
+                "model": self.model,
+                "messages": messages,
+                "timeout": self.timeout,
+                "temperature": self.temperature,
+                "top_p": self.top_p,
+                "n": self.n,
+                "stop": self.stop,
+                "max_tokens": self.max_tokens or self.max_completion_tokens,
+                "presence_penalty": self.presence_penalty,
+                "frequency_penalty": self.frequency_penalty,
+                "logit_bias": self.logit_bias,
+                "seed": self.seed,
+                "logprobs": self.logprobs,
+                "top_logprobs": self.top_logprobs,
+                "api_base": self.api_base,
+                "base_url": self.base_url,
+                "api_version": self.api_version,
+                "api_key": self.api_key,
+                "stream": False,
+                "tools": tools,
+                **self.additional_params,
+            }
+
+            # Remove any keys with None values.
+            params = {k: v for k, v in params.items() if v is not None}
+
+            # --- Direct structured response if no tools are provided.
+            if self.response_format is not None and (tools is None or len(tools) == 0):
+                print("Direct structured response")
+                try:
+                    # Cast messages to required type and remove model param
+                    params["messages"] = cast(
+                        List[ChatCompletionMessageParam], messages
+                    )
+                    params.pop("model", None)
+
+                    client = instructor.from_litellm(litellm.completion)
+                    response = client.chat.completions.create(**params)
+                    return response
+                except Exception as e:
+                    logging.error(f"LiteLLM call failed: {str(e)}")
+                    raise
+
+            # --- Standard flow with potential tool calls.
             try:
-                # --- 1) Prepare the parameters for the completion call
-                params = {
-                    "model": self.model,
-                    "messages": messages,
-                    "timeout": self.timeout,
-                    "temperature": self.temperature,
-                    "top_p": self.top_p,
-                    "n": self.n,
-                    "stop": self.stop,
-                    "max_tokens": self.max_tokens or self.max_completion_tokens,
-                    "presence_penalty": self.presence_penalty,
-                    "frequency_penalty": self.frequency_penalty,
-                    "logit_bias": self.logit_bias,
-                    "response_format": self.response_format,
-                    "seed": self.seed,
-                    "logprobs": self.logprobs,
-                    "top_logprobs": self.top_logprobs,
-                    "api_base": self.base_url,
-                    "api_version": self.api_version,
-                    "api_key": self.api_key,
-                    "stream": False,
-                    "tools": tools,
-                }
-
-                # Remove None values from params
-                params = {k: v for k, v in params.items() if v is not None}
-
-                # --- 2) Make the completion call
+                print("NOT DIRECT STRUCTURED RESPONSE")
                 response = litellm.completion(**params)
                 response_message = cast(Choices, cast(ModelResponse, response).choices)[
                     0
@@ -250,7 +305,6 @@ class LLM:
                 text_response = response_message.content or ""
                 tool_calls = getattr(response_message, "tool_calls", [])
 
-                # --- 3) Handle callbacks with usage info
                 if callbacks and len(callbacks) > 0:
                     for callback in callbacks:
                         if hasattr(callback, "log_success_event"):
@@ -263,11 +317,11 @@ class LLM:
                                     end_time=0,
                                 )
 
-                # --- 4) If no tool calls, return the text response
+                # If no tool call is requested or available_functions is not provided, return the text response.
                 if not tool_calls or not available_functions:
                     return text_response
 
-                # --- 5) Handle the tool call
+                # --- Handle tool calls.
                 tool_call = tool_calls[0]
                 function_name = tool_call.function.name
 
@@ -280,22 +334,40 @@ class LLM:
 
                     fn = available_functions[function_name]
                     try:
-                        # Call the actual tool function
                         result = fn(**function_args)
-                        return result
-
                     except Exception as e:
                         logging.error(
                             f"Error executing function '{function_name}': {e}"
                         )
                         return text_response
 
+                    # If a structured response is requested, perform a secondary call using the tool result.
+                    if self.response_format is not None:
+                        new_params = dict(params)
+                        # Cast tool result message to required type
+                        new_params["messages"] = cast(
+                            List[ChatCompletionMessageParam],
+                            [{"role": "user", "content": result}],
+                        )
+                        new_params.pop("model", None)
+                        if "tools" in new_params:
+                            del new_params["tools"]
+                        try:
+                            client = instructor.from_litellm(litellm.completion)
+                            final_response = client.chat.completions.create(
+                                **new_params, response_model=response_format
+                            )
+                            return final_response
+                        except Exception as e:
+                            logging.error(f"LiteLLM structured call failed: {e}")
+                            return result
+                    else:
+                        return result
                 else:
                     logging.warning(
                         f"Tool call requested unknown function '{function_name}'"
                     )
                     return text_response
-
             except Exception as e:
                 if not LLMContextLengthExceededException(
                     str(e)

src/crewai/task.py

@@ -431,7 +431,9 @@ class Task(BaseModel):
             content = (
                 json_output
                 if json_output
-                else pydantic_output.model_dump_json() if pydantic_output else result
+                else pydantic_output.model_dump_json()
+                if pydantic_output
+                else result
             )
             self._save_file(content)
 
@@ -452,7 +454,7 @@ class Task(BaseModel):
         return "\n".join(tasks_slices)
 
     def interpolate_inputs_and_add_conversation_history(
-        self, inputs: Dict[str, Union[str, int, float]]
+        self, inputs: Dict[str, Union[str, int, float, Dict[str, Any], List[Any]]]
     ) -> None:
         """Interpolate inputs into the task description, expected output, and output file path.
            Add conversation history if present.
@@ -524,7 +526,9 @@ class Task(BaseModel):
             )
 
     def interpolate_only(
-        self, input_string: Optional[str], inputs: Dict[str, Union[str, int, float]]
+        self,
+        input_string: Optional[str],
+        inputs: Dict[str, Union[str, int, float, Dict[str, Any], List[Any]]],
     ) -> str:
         """Interpolate placeholders (e.g., {key}) in a string while leaving JSON untouched.
 
@@ -532,17 +536,39 @@ class Task(BaseModel):
             input_string: The string containing template variables to interpolate.
                          Can be None or empty, in which case an empty string is returned.
             inputs: Dictionary mapping template variables to their values.
-                   Supported value types are strings, integers, and floats.
-                   If input_string is empty or has no placeholders, inputs can be empty.
+                   Supported value types are strings, integers, floats, and dicts/lists
+                   containing only these types and other nested dicts/lists.
 
         Returns:
             The interpolated string with all template variables replaced with their values.
             Empty string if input_string is None or empty.
 
         Raises:
-            ValueError: If a required template variable is missing from inputs.
-            KeyError: If a template variable is not found in the inputs dictionary.
+            ValueError: If a value contains unsupported types
         """
+
+        # Validation function for recursive type checking
+        def validate_type(value: Any) -> None:
+            if value is None:
+                return
+            if isinstance(value, (str, int, float, bool)):
+                return
+            if isinstance(value, (dict, list)):
+                for item in value.values() if isinstance(value, dict) else value:
+                    validate_type(item)
+                return
+            raise ValueError(
+                f"Unsupported type {type(value).__name__} in inputs. "
+                "Only str, int, float, bool, dict, and list are allowed."
+            )
+
+        # Validate all input values
+        for key, value in inputs.items():
+            try:
+                validate_type(value)
+            except ValueError as e:
+                raise ValueError(f"Invalid value for key '{key}': {str(e)}") from e
+
         if input_string is None or not input_string:
             return ""
         if "{" not in input_string and "}" not in input_string:
@@ -551,15 +577,7 @@ class Task(BaseModel):
             raise ValueError(
                 "Inputs dictionary cannot be empty when interpolating variables"
             )
-
         try:
-            # Validate input types
-            for key, value in inputs.items():
-                if not isinstance(value, (str, int, float)):
-                    raise ValueError(
-                        f"Value for key '{key}' must be a string, integer, or float, got {type(value).__name__}"
-                    )
-
             escaped_string = input_string.replace("{", "{{").replace("}", "}}")
 
             for key in inputs.keys():

src/crewai/translations/en.json

@@ -24,7 +24,8 @@
     "manager_request": "Your best answer to your coworker asking you this, accounting for the context shared.",
     "formatted_task_instructions": "Ensure your final answer contains only the content in the following format: {output_format}\n\nEnsure the final output does not include any code block markers like ```json or ```python.",
     "human_feedback_classification": "Determine if the following feedback indicates that the user is satisfied or if further changes are needed. Respond with 'True' if further changes are needed, or 'False' if the user is satisfied. **Important** Do not include any additional commentary outside of your 'True' or 'False' response.\n\nFeedback: \"{feedback}\"",
-    "conversation_history_instruction": "You are a member of a crew collaborating to achieve a common goal. Your task is a specific action that contributes to this larger objective. For additional context, please review the conversation history between you and the user that led to the initiation of this crew. Use any relevant information or feedback from the conversation to inform your task execution and ensure your response aligns with both the immediate task and the crew's overall goals."
+    "conversation_history_instruction": "You are a member of a crew collaborating to achieve a common goal. Your task is a specific action that contributes to this larger objective. For additional context, please review the conversation history between you and the user that led to the initiation of this crew. Use any relevant information or feedback from the conversation to inform your task execution and ensure your response aligns with both the immediate task and the crew's overall goals.",
+    "feedback_instructions": "User feedback: {feedback}\nInstructions: Use this feedback to enhance the next output iteration.\nNote: Do not respond or add commentary."
   },
   "errors": {
     "force_final_answer_error": "You can't keep going, here is the best final answer you generated:\n\n {formatted_answer}",

src/crewai/utilities/embedding_configurator.py

@@ -43,7 +43,6 @@ class EmbeddingConfigurator:
             raise Exception(
                 f"Unsupported embedding provider: {provider}, supported providers: {list(self.embedding_functions.keys())}"
             )
-
         return self.embedding_functions[provider](config, model_name)
 
     @staticmethod

src/crewai/utilities/evaluators/task_evaluator.py

@@ -92,13 +92,34 @@ class TaskEvaluator:
         """
 
         output_training_data = training_data[agent_id]
-
         final_aggregated_data = ""
-        for _, data in output_training_data.items():
+
+        for iteration, data in output_training_data.items():
+            improved_output = data.get("improved_output")
+            initial_output = data.get("initial_output")
+            human_feedback = data.get("human_feedback")
+
+            if not all([improved_output, initial_output, human_feedback]):
+                missing_fields = [
+                    field
+                    for field in ["improved_output", "initial_output", "human_feedback"]
+                    if not data.get(field)
+                ]
+                error_msg = (
+                    f"Critical training data error: Missing fields ({', '.join(missing_fields)}) "
+                    f"for agent {agent_id} in iteration {iteration}.\n"
+                    "This indicates a broken training process. "
+                    "Cannot proceed with evaluation.\n"
+                    "Please check your training implementation."
+                )
+                raise ValueError(error_msg)
+
             final_aggregated_data += (
-                f"Initial Output:\n{data.get('initial_output', '')}\n\n"
-                f"Human Feedback:\n{data.get('human_feedback', '')}\n\n"
-                f"Improved Output:\n{data.get('improved_output', '')}\n\n"
+                f"Iteration: {iteration}\n"
+                f"Initial Output:\n{initial_output}\n\n"
+                f"Human Feedback:\n{human_feedback}\n\n"
+                f"Improved Output:\n{improved_output}\n\n"
+                "------------------------------------------------\n\n"
             )
 
         evaluation_query = (

src/crewai/utilities/llm_utils.py

@@ -53,6 +53,7 @@ def create_llm(
         timeout: Optional[float] = getattr(llm_value, "timeout", None)
         api_key: Optional[str] = getattr(llm_value, "api_key", None)
         base_url: Optional[str] = getattr(llm_value, "base_url", None)
+        api_base: Optional[str] = getattr(llm_value, "api_base", None)
 
         created_llm = LLM(
             model=model,
@@ -62,6 +63,7 @@ def create_llm(
             timeout=timeout,
             api_key=api_key,
             base_url=base_url,
+            api_base=api_base,
         )
         return created_llm
     except Exception as e:
@@ -101,8 +103,18 @@ def _llm_via_environment_or_fallback() -> Optional[LLM]:
     callbacks: List[Any] = []
 
     # Optional base URL from env
-    api_base = os.environ.get("OPENAI_API_BASE") or os.environ.get("OPENAI_BASE_URL")
-    if api_base:
+    base_url = (
+        os.environ.get("BASE_URL")
+        or os.environ.get("OPENAI_API_BASE")
+        or os.environ.get("OPENAI_BASE_URL")
+    )
+
+    api_base = os.environ.get("API_BASE") or os.environ.get("AZURE_API_BASE")
+
+    # Synchronize base_url and api_base if one is populated and the other is not
+    if base_url and not api_base:
+        api_base = base_url
+    elif api_base and not base_url:
         base_url = api_base
 
     # Initialize llm_params dictionary
@@ -115,6 +127,7 @@ def _llm_via_environment_or_fallback() -> Optional[LLM]:
         "timeout": timeout,
         "api_key": api_key,
         "base_url": base_url,
+        "api_base": api_base,
         "api_version": api_version,
         "presence_penalty": presence_penalty,
         "frequency_penalty": frequency_penalty,

src/crewai/utilities/training_handler.py

@@ -1,3 +1,5 @@
+import os
+
 from crewai.utilities.file_handler import PickleHandler
 
 
@@ -29,3 +31,8 @@ class CrewTrainingHandler(PickleHandler):
             data[agent_id] = {train_iteration: new_data}
 
         self.save(data)
+
+    def clear(self) -> None:
+        """Clear the training data by removing the file or resetting its contents."""
+        if os.path.exists(self.file_path):
+            self.save({})

tests/agent_test.py

@@ -10,13 +10,14 @@ from crewai import Agent, Crew, Task
 from crewai.agents.cache import CacheHandler
 from crewai.agents.crew_agent_executor import CrewAgentExecutor
 from crewai.agents.parser import AgentAction, CrewAgentParser, OutputParserException
+from crewai.knowledge.source.base_knowledge_source import BaseKnowledgeSource
 from crewai.knowledge.source.string_knowledge_source import StringKnowledgeSource
 from crewai.llm import LLM
 from crewai.tools import tool
 from crewai.tools.tool_calling import InstructorToolCalling
 from crewai.tools.tool_usage import ToolUsage
 from crewai.tools.tool_usage_events import ToolUsageFinished
-from crewai.utilities import Printer, RPMController
+from crewai.utilities import RPMController
 from crewai.utilities.events import Emitter
 
 
@@ -1602,6 +1603,45 @@ def test_agent_with_knowledge_sources():
         assert "red" in result.raw.lower()
 
 
+@pytest.mark.vcr(filter_headers=["authorization"])
+def test_agent_with_knowledge_sources_works_with_copy():
+    content = "Brandon's favorite color is red and he likes Mexican food."
+    string_source = StringKnowledgeSource(content=content)
+
+    with patch(
+        "crewai.knowledge.source.base_knowledge_source.BaseKnowledgeSource",
+        autospec=True,
+    ) as MockKnowledgeSource:
+        mock_knowledge_source_instance = MockKnowledgeSource.return_value
+        mock_knowledge_source_instance.__class__ = BaseKnowledgeSource
+        mock_knowledge_source_instance.sources = [string_source]
+
+        agent = Agent(
+            role="Information Agent",
+            goal="Provide information based on knowledge sources",
+            backstory="You have access to specific knowledge sources.",
+            llm=LLM(model="gpt-4o-mini"),
+            knowledge_sources=[string_source],
+        )
+
+        with patch(
+            "crewai.knowledge.storage.knowledge_storage.KnowledgeStorage"
+        ) as MockKnowledgeStorage:
+            mock_knowledge_storage = MockKnowledgeStorage.return_value
+            agent.knowledge_storage = mock_knowledge_storage
+
+            agent_copy = agent.copy()
+
+            assert agent_copy.role == agent.role
+            assert agent_copy.goal == agent.goal
+            assert agent_copy.backstory == agent.backstory
+            assert agent_copy.knowledge_sources is not None
+            assert len(agent_copy.knowledge_sources) == 1
+            assert isinstance(agent_copy.knowledge_sources[0], StringKnowledgeSource)
+            assert agent_copy.knowledge_sources[0].content == content
+            assert isinstance(agent_copy.llm, LLM)
+
+
 @pytest.mark.vcr(filter_headers=["authorization"])
 def test_litellm_auth_error_handling():
     """Test that LiteLLM authentication errors are handled correctly and not retried."""

tests/crew_test.py

@@ -14,6 +14,7 @@ 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.memory.contextual.contextual_memory import ContextualMemory
 from crewai.process import Process
 from crewai.project import crew
@@ -555,12 +556,12 @@ def test_crew_with_delegating_agents_should_not_override_task_tools():
         _, kwargs = mock_execute_sync.call_args
         tools = kwargs["tools"]
 
-        assert any(
-            isinstance(tool, TestTool) for tool in tools
-        ), "TestTool should be present"
-        assert any(
-            "delegate" in tool.name.lower() for tool in tools
-        ), "Delegation tool should be present"
+        assert any(isinstance(tool, TestTool) for tool in tools), (
+            "TestTool should be present"
+        )
+        assert any("delegate" in tool.name.lower() for tool in tools), (
+            "Delegation tool should be present"
+        )
 
 
 @pytest.mark.vcr(filter_headers=["authorization"])
@@ -619,12 +620,12 @@ def test_crew_with_delegating_agents_should_not_override_agent_tools():
         _, kwargs = mock_execute_sync.call_args
         tools = kwargs["tools"]
 
-        assert any(
-            isinstance(tool, TestTool) for tool in new_ceo.tools
-        ), "TestTool should be present"
-        assert any(
-            "delegate" in tool.name.lower() for tool in tools
-        ), "Delegation tool should be present"
+        assert any(isinstance(tool, TestTool) for tool in new_ceo.tools), (
+            "TestTool should be present"
+        )
+        assert any("delegate" in tool.name.lower() for tool in tools), (
+            "Delegation tool should be present"
+        )
 
 
 @pytest.mark.vcr(filter_headers=["authorization"])
@@ -748,17 +749,17 @@ def test_task_tools_override_agent_tools_with_allow_delegation():
         used_tools = kwargs["tools"]
 
         # Confirm AnotherTestTool is present but TestTool is not
-        assert any(
-            isinstance(tool, AnotherTestTool) for tool in used_tools
-        ), "AnotherTestTool should be present"
-        assert not any(
-            isinstance(tool, TestTool) for tool in used_tools
-        ), "TestTool should not be present among used tools"
+        assert any(isinstance(tool, AnotherTestTool) for tool in used_tools), (
+            "AnotherTestTool should be present"
+        )
+        assert not any(isinstance(tool, TestTool) for tool in used_tools), (
+            "TestTool should not be present among used tools"
+        )
 
         # Confirm delegation tool(s) are present
-        assert any(
-            "delegate" in tool.name.lower() for tool in used_tools
-        ), "Delegation tool should be present"
+        assert any("delegate" in tool.name.lower() for tool in used_tools), (
+            "Delegation tool should be present"
+        )
 
     # Finally, make sure the agent's original tools remain unchanged
     assert len(researcher_with_delegation.tools) == 1
@@ -1466,7 +1467,6 @@ def test_dont_set_agents_step_callback_if_already_set():
 
 @pytest.mark.vcr(filter_headers=["authorization"])
 def test_crew_function_calling_llm():
-
     from crewai import LLM
     from crewai.tools import tool
 
@@ -1560,9 +1560,9 @@ def test_code_execution_flag_adds_code_tool_upon_kickoff():
 
         # Verify that exactly one tool was used and it was a CodeInterpreterTool
         assert len(used_tools) == 1, "Should have exactly one tool"
-        assert isinstance(
-            used_tools[0], CodeInterpreterTool
-        ), "Tool should be CodeInterpreterTool"
+        assert isinstance(used_tools[0], CodeInterpreterTool), (
+            "Tool should be CodeInterpreterTool"
+        )
 
 
 @pytest.mark.vcr(filter_headers=["authorization"])
@@ -3107,9 +3107,9 @@ def test_fetch_inputs():
     expected_placeholders = {"role_detail", "topic", "field"}
     actual_placeholders = crew.fetch_inputs()
 
-    assert (
-        actual_placeholders == expected_placeholders
-    ), f"Expected {expected_placeholders}, but got {actual_placeholders}"
+    assert actual_placeholders == expected_placeholders, (
+        f"Expected {expected_placeholders}, but got {actual_placeholders}"
+    )
 
 
 def test_task_tools_preserve_code_execution_tools():
@@ -3182,20 +3182,20 @@ def test_task_tools_preserve_code_execution_tools():
         used_tools = kwargs["tools"]
 
         # Verify all expected tools are present
-        assert any(
-            isinstance(tool, TestTool) for tool in used_tools
-        ), "Task's TestTool should be present"
-        assert any(
-            isinstance(tool, CodeInterpreterTool) for tool in used_tools
-        ), "CodeInterpreterTool should be present"
-        assert any(
-            "delegate" in tool.name.lower() for tool in used_tools
-        ), "Delegation tool should be present"
+        assert any(isinstance(tool, TestTool) for tool in used_tools), (
+            "Task's TestTool should be present"
+        )
+        assert any(isinstance(tool, CodeInterpreterTool) for tool in used_tools), (
+            "CodeInterpreterTool should be present"
+        )
+        assert any("delegate" in tool.name.lower() for tool in used_tools), (
+            "Delegation tool should be present"
+        )
 
         # Verify the total number of tools (TestTool + CodeInterpreter + 2 delegation tools)
-        assert (
-            len(used_tools) == 4
-        ), "Should have TestTool, CodeInterpreter, and 2 delegation tools"
+        assert len(used_tools) == 4, (
+            "Should have TestTool, CodeInterpreter, and 2 delegation tools"
+        )
 
 
 @pytest.mark.vcr(filter_headers=["authorization"])
@@ -3239,9 +3239,9 @@ def test_multimodal_flag_adds_multimodal_tools():
         used_tools = kwargs["tools"]
 
         # Check that the multimodal tool was added
-        assert any(
-            isinstance(tool, AddImageTool) for tool in used_tools
-        ), "AddImageTool should be present when agent is multimodal"
+        assert any(isinstance(tool, AddImageTool) for tool in used_tools), (
+            "AddImageTool should be present when agent is multimodal"
+        )
 
         # Verify we have exactly one tool (just the AddImageTool)
         assert len(used_tools) == 1, "Should only have the AddImageTool"
@@ -3467,9 +3467,9 @@ def test_crew_guardrail_feedback_in_context():
     assert len(execution_contexts) > 1, "Task should have been executed multiple times"
 
     # Verify that the second execution included the guardrail feedback
-    assert (
-        "Output must contain the keyword 'IMPORTANT'" in execution_contexts[1]
-    ), "Guardrail feedback should be included in retry context"
+    assert "Output must contain the keyword 'IMPORTANT'" in execution_contexts[1], (
+        "Guardrail feedback should be included in retry context"
+    )
 
     # Verify final output meets guardrail requirements
     assert "IMPORTANT" in result.raw, "Final output should contain required keyword"
@@ -3494,7 +3494,6 @@ def test_before_kickoff_callback():
 
         @before_kickoff
         def modify_inputs(self, inputs):
-
             self.inputs_modified = True
             inputs["modified"] = True
             return inputs
@@ -3596,3 +3595,21 @@ def test_before_kickoff_without_inputs():
     # Verify that the inputs were initialized and modified inside the before_kickoff method
     assert test_crew_instance.received_inputs is not None
     assert test_crew_instance.received_inputs.get("modified") is True
+
+
+@pytest.mark.vcr(filter_headers=["authorization"])
+def test_crew_with_knowledge_sources_works_with_copy():
+    content = "Brandon's favorite color is red and he likes Mexican food."
+    string_source = StringKnowledgeSource(content=content)
+
+    crew = Crew(
+        agents=[researcher, writer],
+        tasks=[Task(description="test", expected_output="test", agent=researcher)],
+        knowledge_sources=[string_source],
+    )
+
+    crew_copy = crew.copy()
+
+    assert crew_copy.knowledge_sources == crew.knowledge_sources
+    assert len(crew_copy.agents) == len(crew.agents)
+    assert len(crew_copy.tasks) == len(crew.tasks)

tests/llm_test.py

@@ -1,4 +1,5 @@
 from time import sleep
+from unittest.mock import MagicMock, patch
 
 import pytest
 
@@ -154,3 +155,50 @@ def test_llm_call_with_tool_and_message_list():
 
     assert isinstance(result, int)
     assert result == 25
+
+
+@pytest.mark.vcr(filter_headers=["authorization"])
+def test_llm_passes_additional_params():
+    llm = LLM(
+        model="gpt-4o-mini",
+        vertex_credentials="test_credentials",
+        vertex_project="test_project",
+    )
+
+    messages = [{"role": "user", "content": "Hello, world!"}]
+
+    with patch("litellm.completion") as mocked_completion:
+        # Create mocks for response structure
+        mock_message = MagicMock()
+        mock_message.content = "Test response"
+        mock_choice = MagicMock()
+        mock_choice.message = mock_message
+        mock_response = MagicMock()
+        mock_response.choices = [mock_choice]
+        mock_response.usage = {
+            "prompt_tokens": 5,
+            "completion_tokens": 5,
+            "total_tokens": 10,
+        }
+
+        # Set up the mocked completion to return the mock response
+        mocked_completion.return_value = mock_response
+
+        result = llm.call(messages)
+
+        # Assert that litellm.completion was called once
+        mocked_completion.assert_called_once()
+
+        # Retrieve the actual arguments with which litellm.completion was called
+        _, kwargs = mocked_completion.call_args
+
+        # Check that the additional_params were passed to litellm.completion
+        assert kwargs["vertex_credentials"] == "test_credentials"
+        assert kwargs["vertex_project"] == "test_project"
+
+        # Also verify that other expected parameters are present
+        assert kwargs["model"] == "gpt-4o-mini"
+        assert kwargs["messages"] == messages
+
+        # Check the result from llm.call
+        assert result == "Test response"

tests/task_test.py

@@ -779,6 +779,43 @@ def test_interpolate_only():
     assert result == no_placeholders
 
 
+def test_interpolate_only_with_dict_inside_expected_output():
+    """Test the interpolate_only method for various scenarios including JSON structure preservation."""
+    task = Task(
+        description="Unused in this test",
+        expected_output="Unused in this test: {questions}",
+    )
+
+    json_string = '{"questions": {"main_question": "What is the user\'s name?", "secondary_question": "What is the user\'s age?"}}'
+    result = task.interpolate_only(
+        input_string=json_string,
+        inputs={
+            "questions": {
+                "main_question": "What is the user's name?",
+                "secondary_question": "What is the user's age?",
+            }
+        },
+    )
+    assert '"main_question": "What is the user\'s name?"' in result
+    assert '"secondary_question": "What is the user\'s age?"' in result
+    assert result == json_string
+
+    normal_string = "Hello {name}, welcome to {place}!"
+    result = task.interpolate_only(
+        input_string=normal_string, inputs={"name": "John", "place": "CrewAI"}
+    )
+    assert result == "Hello John, welcome to CrewAI!"
+
+    result = task.interpolate_only(input_string="", inputs={"unused": "value"})
+    assert result == ""
+
+    no_placeholders = "Hello, this is a test"
+    result = task.interpolate_only(
+        input_string=no_placeholders, inputs={"unused": "value"}
+    )
+    assert result == no_placeholders
+
+
 def test_task_output_str_with_pydantic():
     from crewai.tasks.output_format import OutputFormat
 
@@ -966,3 +1003,283 @@ def test_task_execution_times():
     assert task.start_time is not None
     assert task.end_time is not None
     assert task.execution_duration == (task.end_time - task.start_time).total_seconds()
+
+
+def test_interpolate_with_list_of_strings():
+    task = Task(
+        description="Test list interpolation",
+        expected_output="List: {items}",
+    )
+
+    # Test simple list of strings
+    input_str = "Available items: {items}"
+    inputs = {"items": ["apple", "banana", "cherry"]}
+    result = task.interpolate_only(input_str, inputs)
+    assert result == f"Available items: {inputs['items']}"
+
+    # Test empty list
+    empty_list_input = {"items": []}
+    result = task.interpolate_only(input_str, empty_list_input)
+    assert result == "Available items: []"
+
+
+def test_interpolate_with_list_of_dicts():
+    task = Task(
+        description="Test list of dicts interpolation",
+        expected_output="People: {people}",
+    )
+
+    input_data = {
+        "people": [
+            {"name": "Alice", "age": 30, "skills": ["Python", "AI"]},
+            {"name": "Bob", "age": 25, "skills": ["Java", "Cloud"]},
+        ]
+    }
+    result = task.interpolate_only("{people}", input_data)
+
+    parsed_result = eval(result)
+    assert isinstance(parsed_result, list)
+    assert len(parsed_result) == 2
+    assert parsed_result[0]["name"] == "Alice"
+    assert parsed_result[0]["age"] == 30
+    assert parsed_result[0]["skills"] == ["Python", "AI"]
+    assert parsed_result[1]["name"] == "Bob"
+    assert parsed_result[1]["age"] == 25
+    assert parsed_result[1]["skills"] == ["Java", "Cloud"]
+
+
+def test_interpolate_with_nested_structures():
+    task = Task(
+        description="Test nested structures",
+        expected_output="Company: {company}",
+    )
+
+    input_data = {
+        "company": {
+            "name": "TechCorp",
+            "departments": [
+                {
+                    "name": "Engineering",
+                    "employees": 50,
+                    "tools": ["Git", "Docker", "Kubernetes"],
+                },
+                {"name": "Sales", "employees": 20, "regions": {"north": 5, "south": 3}},
+            ],
+        }
+    }
+    result = task.interpolate_only("{company}", input_data)
+    parsed = eval(result)
+
+    assert parsed["name"] == "TechCorp"
+    assert len(parsed["departments"]) == 2
+    assert parsed["departments"][0]["tools"] == ["Git", "Docker", "Kubernetes"]
+    assert parsed["departments"][1]["regions"]["north"] == 5
+
+
+def test_interpolate_with_special_characters():
+    task = Task(
+        description="Test special characters in dicts",
+        expected_output="Data: {special_data}",
+    )
+
+    input_data = {
+        "special_data": {
+            "quotes": """This has "double" and 'single' quotes""",
+            "unicode": "文字化けテスト",
+            "symbols": "!@#$%^&*()",
+            "empty": "",
+        }
+    }
+    result = task.interpolate_only("{special_data}", input_data)
+    parsed = eval(result)
+
+    assert parsed["quotes"] == """This has "double" and 'single' quotes"""
+    assert parsed["unicode"] == "文字化けテスト"
+    assert parsed["symbols"] == "!@#$%^&*()"
+    assert parsed["empty"] == ""
+
+
+def test_interpolate_mixed_types():
+    task = Task(
+        description="Test mixed type interpolation",
+        expected_output="Mixed: {data}",
+    )
+
+    input_data = {
+        "data": {
+            "name": "Test Dataset",
+            "samples": 1000,
+            "features": ["age", "income", "location"],
+            "metadata": {
+                "source": "public",
+                "validated": True,
+                "tags": ["demo", "test", "temp"],
+            },
+        }
+    }
+    result = task.interpolate_only("{data}", input_data)
+    parsed = eval(result)
+
+    assert parsed["name"] == "Test Dataset"
+    assert parsed["samples"] == 1000
+    assert parsed["metadata"]["tags"] == ["demo", "test", "temp"]
+
+
+def test_interpolate_complex_combination():
+    task = Task(
+        description="Test complex combination",
+        expected_output="Report: {report}",
+    )
+
+    input_data = {
+        "report": [
+            {
+                "month": "January",
+                "metrics": {"sales": 15000, "expenses": 8000, "profit": 7000},
+                "top_products": ["Product A", "Product B"],
+            },
+            {
+                "month": "February",
+                "metrics": {"sales": 18000, "expenses": 8500, "profit": 9500},
+                "top_products": ["Product C", "Product D"],
+            },
+        ]
+    }
+    result = task.interpolate_only("{report}", input_data)
+    parsed = eval(result)
+
+    assert len(parsed) == 2
+    assert parsed[0]["month"] == "January"
+    assert parsed[1]["metrics"]["profit"] == 9500
+    assert "Product D" in parsed[1]["top_products"]
+
+
+def test_interpolate_invalid_type_validation():
+    task = Task(
+        description="Test invalid type validation",
+        expected_output="Should never reach here",
+    )
+
+    # Test with invalid top-level type
+    with pytest.raises(ValueError) as excinfo:
+        task.interpolate_only("{data}", {"data": set()})  # type: ignore we are purposely testing this failure
+
+    assert "Unsupported type set" in str(excinfo.value)
+
+    # Test with invalid nested type
+    invalid_nested = {
+        "profile": {
+            "name": "John",
+            "age": 30,
+            "tags": {"a", "b", "c"},  # Set is invalid
+        }
+    }
+    with pytest.raises(ValueError) as excinfo:
+        task.interpolate_only("{data}", {"data": invalid_nested})
+    assert "Unsupported type set" in str(excinfo.value)
+
+
+def test_interpolate_custom_object_validation():
+    task = Task(
+        description="Test custom object rejection",
+        expected_output="Should never reach here",
+    )
+
+    class CustomObject:
+        def __init__(self, value):
+            self.value = value
+
+        def __str__(self):
+            return str(self.value)
+
+    # Test with custom object at top level
+    with pytest.raises(ValueError) as excinfo:
+        task.interpolate_only("{obj}", {"obj": CustomObject(5)})  # type: ignore we are purposely testing this failure
+    assert "Unsupported type CustomObject" in str(excinfo.value)
+
+    # Test with nested custom object in dictionary
+    with pytest.raises(ValueError) as excinfo:
+        task.interpolate_only(
+            "{data}", {"data": {"valid": 1, "invalid": CustomObject(5)}}
+        )
+    assert "Unsupported type CustomObject" in str(excinfo.value)
+
+    # Test with nested custom object in list
+    with pytest.raises(ValueError) as excinfo:
+        task.interpolate_only("{data}", {"data": [1, "valid", CustomObject(5)]})
+    assert "Unsupported type CustomObject" in str(excinfo.value)
+
+    # Test with deeply nested custom object
+    with pytest.raises(ValueError) as excinfo:
+        task.interpolate_only(
+            "{data}", {"data": {"level1": {"level2": [{"level3": CustomObject(5)}]}}}
+        )
+    assert "Unsupported type CustomObject" in str(excinfo.value)
+
+
+def test_interpolate_valid_complex_types():
+    task = Task(
+        description="Test valid complex types",
+        expected_output="Validation should pass",
+    )
+
+    # Valid complex structure
+    valid_data = {
+        "name": "Valid Dataset",
+        "stats": {
+            "count": 1000,
+            "distribution": [0.2, 0.3, 0.5],
+            "features": ["age", "income"],
+            "nested": {"deep": [1, 2, 3], "deeper": {"a": 1, "b": 2.5}},
+        },
+    }
+
+    # Should not raise any errors
+    result = task.interpolate_only("{data}", {"data": valid_data})
+    parsed = eval(result)
+    assert parsed["name"] == "Valid Dataset"
+    assert parsed["stats"]["nested"]["deeper"]["b"] == 2.5
+
+
+def test_interpolate_edge_cases():
+    task = Task(
+        description="Test edge cases",
+        expected_output="Edge case handling",
+    )
+
+    # Test empty dict and list
+    assert task.interpolate_only("{}", {"data": {}}) == "{}"
+    assert task.interpolate_only("[]", {"data": []}) == "[]"
+
+    # Test numeric types
+    assert task.interpolate_only("{num}", {"num": 42}) == "42"
+    assert task.interpolate_only("{num}", {"num": 3.14}) == "3.14"
+
+    # Test boolean values (valid JSON types)
+    assert task.interpolate_only("{flag}", {"flag": True}) == "True"
+    assert task.interpolate_only("{flag}", {"flag": False}) == "False"
+
+
+def test_interpolate_valid_types():
+    task = Task(
+        description="Test valid types including null and boolean",
+        expected_output="Should pass validation",
+    )
+
+    # Test with boolean and null values (valid JSON types)
+    valid_data = {
+        "name": "Test",
+        "active": True,
+        "deleted": False,
+        "optional": None,
+        "nested": {"flag": True, "empty": None},
+    }
+
+    result = task.interpolate_only("{data}", {"data": valid_data})
+    parsed = eval(result)
+
+    assert parsed["active"] is True
+    assert parsed["deleted"] is False
+    assert parsed["optional"] is None
+    assert parsed["nested"]["flag"] is True
+    assert parsed["nested"]["empty"] is None

tests/utilities/evaluators/test_task_evaluator.py

@@ -48,9 +48,9 @@ def test_evaluate_training_data(converter_mock):
             mock.call(
                 llm=original_agent.llm,
                 text="Assess the quality of the training data based on the llm output, human feedback , and llm "
-                "output improved result.\n\nInitial Output:\nInitial output 1\n\nHuman Feedback:\nHuman feedback "
-                "1\n\nImproved Output:\nImproved output 1\n\nInitial Output:\nInitial output 2\n\nHuman "
-                "Feedback:\nHuman feedback 2\n\nImproved Output:\nImproved output 2\n\nPlease provide:\n- Provide "
+                "output improved result.\n\nIteration: data1\nInitial Output:\nInitial output 1\n\nHuman Feedback:\nHuman feedback "
+                "1\n\nImproved Output:\nImproved output 1\n\n------------------------------------------------\n\nIteration: data2\nInitial Output:\nInitial output 2\n\nHuman "
+                "Feedback:\nHuman feedback 2\n\nImproved Output:\nImproved output 2\n\n------------------------------------------------\n\nPlease provide:\n- Provide "
                 "a list of clear, actionable instructions derived from the Human Feedbacks to enhance the Agent's "
                 "performance. Analyze the differences between Initial Outputs and Improved Outputs to generate specific "
                 "action items for future tasks. Ensure all key and specificpoints from the human feedback are "

uv.lock

@@ -736,7 +736,7 @@ requires-dist = [
     { name = "crewai-tools", marker = "extra == 'tools'", specifier = ">=0.32.1" },
     { name = "docling", marker = "extra == 'docling'", specifier = ">=2.12.0" },
     { name = "fastembed", marker = "extra == 'fastembed'", specifier = ">=0.4.1" },
-    { name = "instructor", specifier = ">=1.3.3" },
+    { name = "instructor", specifier = ">=1.7.2" },
     { name = "json-repair", specifier = ">=0.25.2" },
     { name = "json5", specifier = ">=0.10.0" },
     { name = "jsonref", specifier = ">=1.1.0" },
@@ -1961,7 +1961,7 @@ wheels = [
 
 [[package]]
 name = "instructor"
-version = "1.6.3"
+version = "1.7.2"
 source = { registry = "https://pypi.org/simple" }
 dependencies = [
     { name = "aiohttp" },
@@ -1971,13 +1971,14 @@ dependencies = [
     { name = "openai" },
     { name = "pydantic" },
     { name = "pydantic-core" },
+    { name = "requests" },
     { name = "rich" },
     { name = "tenacity" },
     { name = "typer" },
 ]
-sdist = { url = "https://files.pythonhosted.org/packages/b8/e6/21969fe0de9d278979872240b6af17510af8bd5020f6845891719c1d3eef/instructor-1.6.3.tar.gz", hash = "sha256:399cd90e30b5bc7cbd47acd7399c9c4e84926a96c20c8b5d00c5a04b41ed41ab", size = 56708 }
+sdist = { url = "https://files.pythonhosted.org/packages/63/ba/692739c76959191aa7e5f0fccda871b36548355f4a09c8733687e64e62b0/instructor-1.7.2.tar.gz", hash = "sha256:6c01b2b159766df24865dc81f7bf8457cbda88a3c0bbc810da3467d19b185ed2", size = 66200177 }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/10/98/c96bf0b1656173d06cd6c3a5adaf3ac429f86d5d696ae8e90e6eb15e89be/instructor-1.6.3-py3-none-any.whl", hash = "sha256:a8f973fea621c0188009b65a3429a526c24aeb249fc24100b605ea496e92d622", size = 69447 },
+    { url = "https://files.pythonhosted.org/packages/c5/82/fd319382c1a33d7021cf151007b4cbd5daddf09d9ca5fb670e476668f9fc/instructor-1.7.2-py3-none-any.whl", hash = "sha256:cb43d27f6d7631c31762b936b2fcb44d2a3f9d8a020430a0f4d3484604ffb95b", size = 71353 },
 ]
 
 [[package]]
@@ -2028,46 +2029,46 @@ wheels = [
 
 [[package]]
 name = "jiter"
-version = "0.5.0"
+version = "0.8.2"
 source = { registry = "https://pypi.org/simple" }
-sdist = { url = "https://files.pythonhosted.org/packages/d7/1a/aa64be757afc614484b370a4d9fc1747dc9237b37ce464f7f9d9ca2a3d38/jiter-0.5.0.tar.gz", hash = "sha256:1d916ba875bcab5c5f7d927df998c4cb694d27dceddf3392e58beaf10563368a", size = 158300 }
+sdist = { url = "https://files.pythonhosted.org/packages/f8/70/90bc7bd3932e651486861df5c8ffea4ca7c77d28e8532ddefe2abc561a53/jiter-0.8.2.tar.gz", hash = "sha256:cd73d3e740666d0e639f678adb176fad25c1bcbdae88d8d7b857e1783bb4212d", size = 163007 }
 wheels = [
-    { url = "https://files.pythonhosted.org/packages/af/09/f659fc67d6aaa82c56432c4a7cc8365fff763acbf1c8f24121076617f207/jiter-0.5.0-cp310-cp310-macosx_10_12_x86_64.whl", hash = "sha256:b599f4e89b3def9a94091e6ee52e1d7ad7bc33e238ebb9c4c63f211d74822c3f", size = 284126 },
-    { url = "https://files.pythonhosted.org/packages/07/2d/5bdaddfefc44f91af0f3340e75ef327950d790c9f86490757ac8b395c074/jiter-0.5.0-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:2a063f71c4b06225543dddadbe09d203dc0c95ba352d8b85f1221173480a71d5", size = 299265 },
-    { url = "https://files.pythonhosted.org/packages/74/bd/964485231deaec8caa6599f3f27c8787a54e9f9373ae80dcfbda2ad79c02/jiter-0.5.0-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:acc0d5b8b3dd12e91dd184b87273f864b363dfabc90ef29a1092d269f18c7e28", size = 332178 },
-    { url = "https://files.pythonhosted.org/packages/cf/4f/6353179174db10254549bbf2eb2c7ea102e59e0460ee374adb12071c274d/jiter-0.5.0-cp310-cp310-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:c22541f0b672f4d741382a97c65609332a783501551445ab2df137ada01e019e", size = 342533 },
-    { url = "https://files.pythonhosted.org/packages/76/6f/21576071b8b056ef743129b9dacf9da65e328b58766f3d1ea265e966f000/jiter-0.5.0-cp310-cp310-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:63314832e302cc10d8dfbda0333a384bf4bcfce80d65fe99b0f3c0da8945a91a", size = 363469 },
-    { url = "https://files.pythonhosted.org/packages/73/a1/9ef99a279c72a031dbe8a4085db41e3521ae01ab0058651d6ccc809a5e93/jiter-0.5.0-cp310-cp310-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:a25fbd8a5a58061e433d6fae6d5298777c0814a8bcefa1e5ecfff20c594bd749", size = 379078 },
-    { url = "https://files.pythonhosted.org/packages/41/6a/c038077509d67fe876c724bfe9ad15334593851a7def0d84518172bdd44a/jiter-0.5.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:503b2c27d87dfff5ab717a8200fbbcf4714516c9d85558048b1fc14d2de7d8dc", size = 318943 },
-    { url = "https://files.pythonhosted.org/packages/67/0d/d82673814eb38c208b7881581df596e680f8c2c003e2b80c25ca58975ee4/jiter-0.5.0-cp310-cp310-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:6d1f3d27cce923713933a844872d213d244e09b53ec99b7a7fdf73d543529d6d", size = 357394 },
-    { url = "https://files.pythonhosted.org/packages/56/9e/cbd8f6612346c38cc42e41e35cda19ce78f5b12e4106d1186e8e95ee839b/jiter-0.5.0-cp310-cp310-musllinux_1_1_aarch64.whl", hash = "sha256:c95980207b3998f2c3b3098f357994d3fd7661121f30669ca7cb945f09510a87", size = 511080 },
-    { url = "https://files.pythonhosted.org/packages/ff/33/135c0c33565b6d5c3010d047710837427dd24c9adbc9ca090f3f92df446e/jiter-0.5.0-cp310-cp310-musllinux_1_1_x86_64.whl", hash = "sha256:afa66939d834b0ce063f57d9895e8036ffc41c4bd90e4a99631e5f261d9b518e", size = 492827 },
-    { url = "https://files.pythonhosted.org/packages/68/c1/491a8ef682508edbaf2a32e41c1b1e34064078b369b0c2d141170999d1c9/jiter-0.5.0-cp310-none-win32.whl", hash = "sha256:f16ca8f10e62f25fd81d5310e852df6649af17824146ca74647a018424ddeccf", size = 195081 },
-    { url = "https://files.pythonhosted.org/packages/31/20/8cda4faa9571affea6130b150289522a22329778bdfa45a7aab4e7edff95/jiter-0.5.0-cp310-none-win_amd64.whl", hash = "sha256:b2950e4798e82dd9176935ef6a55cf6a448b5c71515a556da3f6b811a7844f1e", size = 190977 },
-    { url = "https://files.pythonhosted.org/packages/94/5f/3ac960ed598726aae46edea916e6df4df7ff6fe084bc60774b95cf3154e6/jiter-0.5.0-cp311-cp311-macosx_10_12_x86_64.whl", hash = "sha256:d4c8e1ed0ef31ad29cae5ea16b9e41529eb50a7fba70600008e9f8de6376d553", size = 284131 },
-    { url = "https://files.pythonhosted.org/packages/03/eb/2308fa5f5c14c97c4c7720fef9465f1fa0771826cddb4eec9866bdd88846/jiter-0.5.0-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:c6f16e21276074a12d8421692515b3fd6d2ea9c94fd0734c39a12960a20e85f3", size = 299310 },
-    { url = "https://files.pythonhosted.org/packages/3c/f6/dba34ca10b44715fa5302b8e8d2113f72eb00a9297ddf3fa0ae4fd22d1d1/jiter-0.5.0-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:5280e68e7740c8c128d3ae5ab63335ce6d1fb6603d3b809637b11713487af9e6", size = 332282 },
-    { url = "https://files.pythonhosted.org/packages/69/f7/64e0a7439790ec47f7681adb3871c9d9c45fff771102490bbee5e92c00b7/jiter-0.5.0-cp311-cp311-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:583c57fc30cc1fec360e66323aadd7fc3edeec01289bfafc35d3b9dcb29495e4", size = 342370 },
-    { url = "https://files.pythonhosted.org/packages/55/31/1efbfff2ae8e4d919144c53db19b828049ad0622a670be3bbea94a86282c/jiter-0.5.0-cp311-cp311-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:26351cc14507bdf466b5f99aba3df3143a59da75799bf64a53a3ad3155ecded9", size = 363591 },
-    { url = "https://files.pythonhosted.org/packages/30/c3/7ab2ca2276426a7398c6dfb651e38dbc81954c79a3bfbc36c514d8599499/jiter-0.5.0-cp311-cp311-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:4829df14d656b3fb87e50ae8b48253a8851c707da9f30d45aacab2aa2ba2d614", size = 378551 },
-    { url = "https://files.pythonhosted.org/packages/47/e7/5d88031cd743c62199b125181a591b1671df3ff2f6e102df85c58d8f7d31/jiter-0.5.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:a42a4bdcf7307b86cb863b2fb9bb55029b422d8f86276a50487982d99eed7c6e", size = 319152 },
-    { url = "https://files.pythonhosted.org/packages/4c/2d/09ea58e1adca9f0359f3d41ef44a1a18e59518d7c43a21f4ece9e72e28c0/jiter-0.5.0-cp311-cp311-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:04d461ad0aebf696f8da13c99bc1b3e06f66ecf6cfd56254cc402f6385231c06", size = 357377 },
-    { url = "https://files.pythonhosted.org/packages/7d/2f/83ff1058cb56fc3ff73e0d3c6440703ddc9cdb7f759b00cfbde8228fc435/jiter-0.5.0-cp311-cp311-musllinux_1_1_aarch64.whl", hash = "sha256:e6375923c5f19888c9226582a124b77b622f8fd0018b843c45eeb19d9701c403", size = 511091 },
-    { url = "https://files.pythonhosted.org/packages/ae/c9/4f85f97c9894382ab457382337aea0012711baaa17f2ed55c0ff25f3668a/jiter-0.5.0-cp311-cp311-musllinux_1_1_x86_64.whl", hash = "sha256:2cec323a853c24fd0472517113768c92ae0be8f8c384ef4441d3632da8baa646", size = 492948 },
-    { url = "https://files.pythonhosted.org/packages/4d/f2/2e987e0eb465e064c5f52c2f29c8d955452e3b316746e326269263bfb1b7/jiter-0.5.0-cp311-none-win32.whl", hash = "sha256:aa1db0967130b5cab63dfe4d6ff547c88b2a394c3410db64744d491df7f069bb", size = 195183 },
-    { url = "https://files.pythonhosted.org/packages/ab/59/05d1c3203c349b37c4dd28b02b9b4e5915a7bcbd9319173b4548a67d2e93/jiter-0.5.0-cp311-none-win_amd64.whl", hash = "sha256:aa9d2b85b2ed7dc7697597dcfaac66e63c1b3028652f751c81c65a9f220899ae", size = 191032 },
-    { url = "https://files.pythonhosted.org/packages/aa/bd/c3950e2c478161e131bed8cb67c36aed418190e2a961a1c981e69954e54b/jiter-0.5.0-cp312-cp312-macosx_10_12_x86_64.whl", hash = "sha256:9f664e7351604f91dcdd557603c57fc0d551bc65cc0a732fdacbf73ad335049a", size = 283511 },
-    { url = "https://files.pythonhosted.org/packages/80/1c/8ce58d8c37a589eeaaa5d07d131fd31043886f5e77ab50c00a66d869a361/jiter-0.5.0-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:044f2f1148b5248ad2c8c3afb43430dccf676c5a5834d2f5089a4e6c5bbd64df", size = 296974 },
-    { url = "https://files.pythonhosted.org/packages/4d/b8/6faeff9eed8952bed93a77ea1cffae7b946795b88eafd1a60e87a67b09e0/jiter-0.5.0-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:702e3520384c88b6e270c55c772d4bd6d7b150608dcc94dea87ceba1b6391248", size = 331897 },
-    { url = "https://files.pythonhosted.org/packages/4f/54/1d9a2209b46d39ce6f0cef3ad87c462f9c50312ab84585e6bd5541292b35/jiter-0.5.0-cp312-cp312-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:528d742dcde73fad9d63e8242c036ab4a84389a56e04efd854062b660f559544", size = 342962 },
-    { url = "https://files.pythonhosted.org/packages/2a/de/90360be7fc54b2b4c2dfe79eb4ed1f659fce9c96682e6a0be4bbe71371f7/jiter-0.5.0-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:8cf80e5fe6ab582c82f0c3331df27a7e1565e2dcf06265afd5173d809cdbf9ba", size = 363844 },
-    { url = "https://files.pythonhosted.org/packages/ba/ad/ef32b173191b7a53ea8a6757b80723cba321f8469834825e8c71c96bde17/jiter-0.5.0-cp312-cp312-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:44dfc9ddfb9b51a5626568ef4e55ada462b7328996294fe4d36de02fce42721f", size = 378709 },
-    { url = "https://files.pythonhosted.org/packages/07/de/353ce53743c0defbbbd652e89c106a97dbbac4eb42c95920b74b5056b93a/jiter-0.5.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:c451f7922992751a936b96c5f5b9bb9312243d9b754c34b33d0cb72c84669f4e", size = 319038 },
-    { url = "https://files.pythonhosted.org/packages/3f/92/42d47310bf9530b9dece9e2d7c6d51cf419af5586ededaf5e66622d160e2/jiter-0.5.0-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:308fce789a2f093dca1ff91ac391f11a9f99c35369117ad5a5c6c4903e1b3e3a", size = 357763 },
-    { url = "https://files.pythonhosted.org/packages/bd/8c/2bb76a9a84474d48fdd133d3445db8a4413da4e87c23879d917e000a9d87/jiter-0.5.0-cp312-cp312-musllinux_1_1_aarch64.whl", hash = "sha256:7f5ad4a7c6b0d90776fdefa294f662e8a86871e601309643de30bf94bb93a64e", size = 511031 },
-    { url = "https://files.pythonhosted.org/packages/33/4f/9f23d79c0795e0a8e56e7988e8785c2dcda27e0ed37977256d50c77c6a19/jiter-0.5.0-cp312-cp312-musllinux_1_1_x86_64.whl", hash = "sha256:ea189db75f8eca08807d02ae27929e890c7d47599ce3d0a6a5d41f2419ecf338", size = 493042 },
-    { url = "https://files.pythonhosted.org/packages/df/67/8a4f975aa834b8aecdb6b131422390173928fd47f42f269dcc32034ab432/jiter-0.5.0-cp312-none-win32.whl", hash = "sha256:e3bbe3910c724b877846186c25fe3c802e105a2c1fc2b57d6688b9f8772026e4", size = 195405 },
-    { url = "https://files.pythonhosted.org/packages/15/81/296b1e25c43db67848728cdab34ac3eb5c5cbb4955ceb3f51ae60d4a5e3d/jiter-0.5.0-cp312-none-win_amd64.whl", hash = "sha256:a586832f70c3f1481732919215f36d41c59ca080fa27a65cf23d9490e75b2ef5", size = 189720 },
+    { url = "https://files.pythonhosted.org/packages/f2/f3/8c11e0e87bd5934c414f9b1cfae3cbfd4a938d4669d57cb427e1c4d11a7f/jiter-0.8.2-cp310-cp310-macosx_10_12_x86_64.whl", hash = "sha256:ca8577f6a413abe29b079bc30f907894d7eb07a865c4df69475e868d73e71c7b", size = 303381 },
+    { url = "https://files.pythonhosted.org/packages/ea/28/4cd3f0bcbf40e946bc6a62a82c951afc386a25673d3d8d5ee461f1559bbe/jiter-0.8.2-cp310-cp310-macosx_11_0_arm64.whl", hash = "sha256:b25bd626bde7fb51534190c7e3cb97cee89ee76b76d7585580e22f34f5e3f393", size = 311718 },
+    { url = "https://files.pythonhosted.org/packages/0d/17/57acab00507e60bd954eaec0837d9d7b119b4117ff49b8a62f2b646f32ed/jiter-0.8.2-cp310-cp310-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:d5c826a221851a8dc028eb6d7d6429ba03184fa3c7e83ae01cd6d3bd1d4bd17d", size = 335465 },
+    { url = "https://files.pythonhosted.org/packages/74/b9/1a3ddd2bc95ae17c815b021521020f40c60b32137730126bada962ef32b4/jiter-0.8.2-cp310-cp310-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:d35c864c2dff13dfd79fb070fc4fc6235d7b9b359efe340e1261deb21b9fcb66", size = 355570 },
+    { url = "https://files.pythonhosted.org/packages/78/69/6d29e2296a934199a7d0dde673ecccf98c9c8db44caf0248b3f2b65483cb/jiter-0.8.2-cp310-cp310-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:f557c55bc2b7676e74d39d19bcb8775ca295c7a028246175d6a8b431e70835e5", size = 381383 },
+    { url = "https://files.pythonhosted.org/packages/22/d7/fbc4c3fb1bf65f9be22a32759b539f88e897aeb13fe84ab0266e4423487a/jiter-0.8.2-cp310-cp310-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:580ccf358539153db147e40751a0b41688a5ceb275e6f3e93d91c9467f42b2e3", size = 390454 },
+    { url = "https://files.pythonhosted.org/packages/4d/a0/3993cda2e267fe679b45d0bcc2cef0b4504b0aa810659cdae9737d6bace9/jiter-0.8.2-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:af102d3372e917cffce49b521e4c32c497515119dc7bd8a75665e90a718bbf08", size = 345039 },
+    { url = "https://files.pythonhosted.org/packages/b9/ef/69c18562b4c09ce88fab5df1dcaf643f6b1a8b970b65216e7221169b81c4/jiter-0.8.2-cp310-cp310-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:cadcc978f82397d515bb2683fc0d50103acff2a180552654bb92d6045dec2c49", size = 376200 },
+    { url = "https://files.pythonhosted.org/packages/4d/17/0b5a8de46a6ab4d836f70934036278b49b8530c292b29dde3483326d4555/jiter-0.8.2-cp310-cp310-musllinux_1_1_aarch64.whl", hash = "sha256:ba5bdf56969cad2019d4e8ffd3f879b5fdc792624129741d3d83fc832fef8c7d", size = 511158 },
+    { url = "https://files.pythonhosted.org/packages/6c/b2/c401a0a2554b36c9e6d6e4876b43790d75139cf3936f0222e675cbc23451/jiter-0.8.2-cp310-cp310-musllinux_1_1_x86_64.whl", hash = "sha256:3b94a33a241bee9e34b8481cdcaa3d5c2116f575e0226e421bed3f7a6ea71cff", size = 503956 },
+    { url = "https://files.pythonhosted.org/packages/d4/02/a0291ed7d72c0ac130f172354ee3cf0b2556b69584de391463a8ee534f40/jiter-0.8.2-cp310-cp310-win32.whl", hash = "sha256:6e5337bf454abddd91bd048ce0dca5134056fc99ca0205258766db35d0a2ea43", size = 202846 },
+    { url = "https://files.pythonhosted.org/packages/ad/20/8c988831ae4bf437e29f1671e198fc99ba8fe49f2895f23789acad1d1811/jiter-0.8.2-cp310-cp310-win_amd64.whl", hash = "sha256:4a9220497ca0cb1fe94e3f334f65b9b5102a0b8147646118f020d8ce1de70105", size = 204414 },
+    { url = "https://files.pythonhosted.org/packages/cb/b0/c1a7caa7f9dc5f1f6cfa08722867790fe2d3645d6e7170ca280e6e52d163/jiter-0.8.2-cp311-cp311-macosx_10_12_x86_64.whl", hash = "sha256:2dd61c5afc88a4fda7d8b2cf03ae5947c6ac7516d32b7a15bf4b49569a5c076b", size = 303666 },
+    { url = "https://files.pythonhosted.org/packages/f5/97/0468bc9eeae43079aaa5feb9267964e496bf13133d469cfdc135498f8dd0/jiter-0.8.2-cp311-cp311-macosx_11_0_arm64.whl", hash = "sha256:a6c710d657c8d1d2adbbb5c0b0c6bfcec28fd35bd6b5f016395f9ac43e878a15", size = 311934 },
+    { url = "https://files.pythonhosted.org/packages/e5/69/64058e18263d9a5f1e10f90c436853616d5f047d997c37c7b2df11b085ec/jiter-0.8.2-cp311-cp311-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:a9584de0cd306072635fe4b89742bf26feae858a0683b399ad0c2509011b9dc0", size = 335506 },
+    { url = "https://files.pythonhosted.org/packages/9d/14/b747f9a77b8c0542141d77ca1e2a7523e854754af2c339ac89a8b66527d6/jiter-0.8.2-cp311-cp311-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:5a90a923338531b7970abb063cfc087eebae6ef8ec8139762007188f6bc69a9f", size = 355849 },
+    { url = "https://files.pythonhosted.org/packages/53/e2/98a08161db7cc9d0e39bc385415890928ff09709034982f48eccfca40733/jiter-0.8.2-cp311-cp311-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:d21974d246ed0181558087cd9f76e84e8321091ebfb3a93d4c341479a736f099", size = 381700 },
+    { url = "https://files.pythonhosted.org/packages/7a/38/1674672954d35bce3b1c9af99d5849f9256ac8f5b672e020ac7821581206/jiter-0.8.2-cp311-cp311-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:32475a42b2ea7b344069dc1e81445cfc00b9d0e3ca837f0523072432332e9f74", size = 389710 },
+    { url = "https://files.pythonhosted.org/packages/f8/9b/92f9da9a9e107d019bcf883cd9125fa1690079f323f5a9d5c6986eeec3c0/jiter-0.8.2-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:8b9931fd36ee513c26b5bf08c940b0ac875de175341cbdd4fa3be109f0492586", size = 345553 },
+    { url = "https://files.pythonhosted.org/packages/44/a6/6d030003394e9659cd0d7136bbeabd82e869849ceccddc34d40abbbbb269/jiter-0.8.2-cp311-cp311-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:ce0820f4a3a59ddced7fce696d86a096d5cc48d32a4183483a17671a61edfddc", size = 376388 },
+    { url = "https://files.pythonhosted.org/packages/ad/8d/87b09e648e4aca5f9af89e3ab3cfb93db2d1e633b2f2931ede8dabd9b19a/jiter-0.8.2-cp311-cp311-musllinux_1_1_aarch64.whl", hash = "sha256:8ffc86ae5e3e6a93765d49d1ab47b6075a9c978a2b3b80f0f32628f39caa0c88", size = 511226 },
+    { url = "https://files.pythonhosted.org/packages/77/95/8008ebe4cdc82eac1c97864a8042ca7e383ed67e0ec17bfd03797045c727/jiter-0.8.2-cp311-cp311-musllinux_1_1_x86_64.whl", hash = "sha256:5127dc1abd809431172bc3fbe8168d6b90556a30bb10acd5ded41c3cfd6f43b6", size = 504134 },
+    { url = "https://files.pythonhosted.org/packages/26/0d/3056a74de13e8b2562e4d526de6dac2f65d91ace63a8234deb9284a1d24d/jiter-0.8.2-cp311-cp311-win32.whl", hash = "sha256:66227a2c7b575720c1871c8800d3a0122bb8ee94edb43a5685aa9aceb2782d44", size = 203103 },
+    { url = "https://files.pythonhosted.org/packages/4e/1e/7f96b798f356e531ffc0f53dd2f37185fac60fae4d6c612bbbd4639b90aa/jiter-0.8.2-cp311-cp311-win_amd64.whl", hash = "sha256:cde031d8413842a1e7501e9129b8e676e62a657f8ec8166e18a70d94d4682855", size = 206717 },
+    { url = "https://files.pythonhosted.org/packages/a1/17/c8747af8ea4e045f57d6cfd6fc180752cab9bc3de0e8a0c9ca4e8af333b1/jiter-0.8.2-cp312-cp312-macosx_10_12_x86_64.whl", hash = "sha256:e6ec2be506e7d6f9527dae9ff4b7f54e68ea44a0ef6b098256ddf895218a2f8f", size = 302027 },
+    { url = "https://files.pythonhosted.org/packages/3c/c1/6da849640cd35a41e91085723b76acc818d4b7d92b0b6e5111736ce1dd10/jiter-0.8.2-cp312-cp312-macosx_11_0_arm64.whl", hash = "sha256:76e324da7b5da060287c54f2fabd3db5f76468006c811831f051942bf68c9d44", size = 310326 },
+    { url = "https://files.pythonhosted.org/packages/06/99/a2bf660d8ccffee9ad7ed46b4f860d2108a148d0ea36043fd16f4dc37e94/jiter-0.8.2-cp312-cp312-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:180a8aea058f7535d1c84183c0362c710f4750bef66630c05f40c93c2b152a0f", size = 334242 },
+    { url = "https://files.pythonhosted.org/packages/a7/5f/cea1c17864828731f11427b9d1ab7f24764dbd9aaf4648a7f851164d2718/jiter-0.8.2-cp312-cp312-manylinux_2_17_armv7l.manylinux2014_armv7l.whl", hash = "sha256:025337859077b41548bdcbabe38698bcd93cfe10b06ff66617a48ff92c9aec60", size = 356654 },
+    { url = "https://files.pythonhosted.org/packages/e9/13/62774b7e5e7f5d5043efe1d0f94ead66e6d0f894ae010adb56b3f788de71/jiter-0.8.2-cp312-cp312-manylinux_2_17_ppc64le.manylinux2014_ppc64le.whl", hash = "sha256:ecff0dc14f409599bbcafa7e470c00b80f17abc14d1405d38ab02e4b42e55b57", size = 379967 },
+    { url = "https://files.pythonhosted.org/packages/ec/fb/096b34c553bb0bd3f2289d5013dcad6074948b8d55212aa13a10d44c5326/jiter-0.8.2-cp312-cp312-manylinux_2_17_s390x.manylinux2014_s390x.whl", hash = "sha256:ffd9fee7d0775ebaba131f7ca2e2d83839a62ad65e8e02fe2bd8fc975cedeb9e", size = 389252 },
+    { url = "https://files.pythonhosted.org/packages/17/61/beea645c0bf398ced8b199e377b61eb999d8e46e053bb285c91c3d3eaab0/jiter-0.8.2-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:14601dcac4889e0a1c75ccf6a0e4baf70dbc75041e51bcf8d0e9274519df6887", size = 345490 },
+    { url = "https://files.pythonhosted.org/packages/d5/df/834aa17ad5dcc3cf0118821da0a0cf1589ea7db9832589278553640366bc/jiter-0.8.2-cp312-cp312-manylinux_2_5_i686.manylinux1_i686.whl", hash = "sha256:92249669925bc1c54fcd2ec73f70f2c1d6a817928480ee1c65af5f6b81cdf12d", size = 376991 },
+    { url = "https://files.pythonhosted.org/packages/67/80/87d140399d382fb4ea5b3d56e7ecaa4efdca17cd7411ff904c1517855314/jiter-0.8.2-cp312-cp312-musllinux_1_1_aarch64.whl", hash = "sha256:e725edd0929fa79f8349ab4ec7f81c714df51dc4e991539a578e5018fa4a7152", size = 510822 },
+    { url = "https://files.pythonhosted.org/packages/5c/37/3394bb47bac1ad2cb0465601f86828a0518d07828a650722e55268cdb7e6/jiter-0.8.2-cp312-cp312-musllinux_1_1_x86_64.whl", hash = "sha256:bf55846c7b7a680eebaf9c3c48d630e1bf51bdf76c68a5f654b8524335b0ad29", size = 503730 },
+    { url = "https://files.pythonhosted.org/packages/f9/e2/253fc1fa59103bb4e3aa0665d6ceb1818df1cd7bf3eb492c4dad229b1cd4/jiter-0.8.2-cp312-cp312-win32.whl", hash = "sha256:7efe4853ecd3d6110301665a5178b9856be7e2a9485f49d91aa4d737ad2ae49e", size = 203375 },
+    { url = "https://files.pythonhosted.org/packages/41/69/6d4bbe66b3b3b4507e47aa1dd5d075919ad242b4b1115b3f80eecd443687/jiter-0.8.2-cp312-cp312-win_amd64.whl", hash = "sha256:83c0efd80b29695058d0fd2fa8a556490dbce9804eac3e281f373bbc99045f6c", size = 204740 },
 ]
 
 [[package]]