Build an in-memory tar archive via `build(tf)` and return it for reading.

Track token usage internally in the LLM instance. Args: usage_data: Token usage data from the API response

Extract timestamp from checkpoint ID or filename.

Use trained data for the agent task prompt to improve output.

Call the LLM asynchronously and return the response. Args: llm: The LLM instance to call. messages: The messages to send to the L

Get an existing collection or create it if it doesn't exist asynchronously. Keyword Args: collection_name: Name of the collection

Cancel an A2A task. Publishes a cancel event that the cancellable decorator listens for. Args: context: The A2A request context cont

Split a PDF into chunks of maximum page count. Yields chunks one at a time to minimize memory usage. Args: file: The PDF file to chu

Clear all registered global after_tool_call hooks. Returns: Number of hooks that were cleared Example: >>> register_after_to

Close the underlying stream.

Creates the Content Crew

Build the default flow persistence backend. Returns the result of the registered factory if one is set, otherwise a built-in :class:`~crewai.

Evaluate this value as a full CEL expression.

Execute a single native tool call with full lifecycle management. Handles: arg parsing, tool lookup, max-usage check, cache read/write, befor

Execute crew with trigger payload.

Create a gRPC channel for the given URL.

Return a path label that does not expose absolute directory prefixes.

Format text as a content block for the LLM. Default implementation uses OpenAI/Anthropic format. Subclasses should override for provi

Get A2A agent configs and response model. Args: a2a_config: A2A configuration (any type). Returns: Tuple of client A2A confi

Get the dimensions of an image file. Args: file: The image file to measure. Returns: Tuple of (width, height) in pixels, or

Return the latest non-yanked PyPI version of CrewAI, or ``None`` on failure.

Call the LLM and return the response, handling any invalid responses. Args: llm: The LLM instance to call. messages: The messages

Get all pending todos. Returns: List of pending TodoItem objects.

Get the current human input provider. Returns: The current provider, or a new SyncHumanInputProvider if none set.

Handle the completion flow as shown in your diagram.

Handles the case when the maximum number of iterations is exceeded. Performs one more LLM call to get the final answer. Args: formatted_a

Handle partial JSON in a result string and convert to Pydantic model or dict. Args: result: The result string to process. model:

Check if the maximum number of iterations has been reached. Args: iterations: Current number of iterations. max_iterations: Maxim

Return True if the user has explicitly declined trace collection.

Return True when the enterprise non-interactive mode is enabled.

Execute the agent asynchronously with the given messages. This is the async version of the kickoff method that uses native async exec

List immediate child scopes under path. Args: path: Scope path to list children of. If None and root_scope is set,

Load prompts from a JSON file. Returns: The I18N instance with loaded prompts. Raises: Exception: If the pro

Convert a list of file sources to a named dictionary of FileInputs. Args: input_files: List of file source inputs or File objects. R

Process a single file according to constraints and its mode mode. Args: file: The file to process. Returns:

Render the tool name, description, and args in plain text. search: This tool is used for search, args: {"query": {"type": "string"}} calculat

Resize an image to fit within the specified dimensions. Args: file: The image file to resize. max_width: Maximum width in pixels.

Restore the event scope stack from a checkpoint.

Push the latest recorded ``task_started`` scope for a task. Args: task_id: The task identifier to look up in the active event record.

Safely extract and validate tool information. Combines extraction, validation, and logging for robust tool conversion. Args: tool: T

Set a CLI configuration parameter.

Set the ID of the last emitted event. Args: event_id: The event_id to set as the last emitted event.

Setup event listeners - delegates to specific handlers. Args: crewai_event_bus: The event bus to register listeners on.

Start the MCP server and initialize the tools.

Run the LLM call and stream scoped public ``StreamFrame`` events.

Iterate over frames, optionally filtered by channel.

Retrieve a tool prompt by key. Args: tool: The key of the tool prompt to retrieve. Returns: The tool prompt

Update a record by upserting with the same point ID.

Update workspace dependency versions in pyproject.toml. Args: file_path: Path to pyproject.toml file. new_version: New version st

Show the installed version of crewai.

Add message to state conversation history. Args: text: Message content. role: Message role (default: assistant).

Create a file only if it doesn't already exist.

Build a realistic multi-turn conversation for summarization tests.

Convert a config dict from crewai-oauth into an MCPServerConfig.

Calls the LLM with function calling to get a plan. Args: prompt: The prompt to send to the LLM. plan_type: The type o

Convert contents to dict format.

Helper to write JSON data to a temporary file and return its path.

A minimal crew whose agent goal references each ``{name}`` placeholder.

Delete a secure file (no-op if absent).

Execute a single table write with retry on commit conflicts. Caller must already hold ``store_lock(self._lock_name)``.

Build a rankable catalog entry. ``sort`` is a comparable tuple ``(created, date_int, version_tuple)`` so a real vendor timestamp wins, then a

Estimate token count using a conservative cross-provider heuristic. Args: text: The text to estimate tokens for. Returns: Es

Fetch all template repos from the GitHub org.

Derives the custom_llm_provider from the model string. - For example, if the model is "openrouter/deepseek/deepseek-chat", returns "o

Get or create the encryption key.

Handle the events for the LLM call. Args: response (str): The response from the LLM call. call_type (str): The type o

GET ``url`` and return parsed JSON, with a short timeout and TLS verify.

Derive a readable label from a raw model id. Best-effort only — vendor display names and the curated label map take precedence. Drops embedde

Inject the current date into the task description if inject_date is enabled.

Check if the error message indicates a context length limit error. Args: error_message: The error message to check. Retu

Build a simple chain: e0 → e1 → e2 → ... with previous_event_id.

Create a TraceBatchManager with a pre-set trace_batch_id.

Mark a trace batch as failed, routing to the correct endpoint.

Normalize Ollama base URL to ensure it ends with /v1. Ollama uses OLLAMA_HOST which may not include the /v1 suffix, but the OpenAI-compatible

Return a Snowflake Cortex REST OpenAI-compatible base URL.

Convert raw input to appropriate source type.

Format a class or instance as the canonical ``module:qualname`` ref.