Replay the crew execution from a specific task.

Import a module, optionally returning a specific attribute. Args: name: The module name to import. purpose: Description of what r

Execute the crew and capture the result.

Validate a video file against constraints. Args: file: The video file to validate. constraints: Video constraints to validate aga

Create a simple analyst crew for file analysis.

Create a batch slice from prepared documents. Args: prepared: PreparedDocuments containing ids, texts, and metadatas. start_index

Emit stream chunk event. Args: chunk: The text content of the chunk. from_task: The task that initiated the call.

LiteLLM responses/chunks share the choices-shape with OpenAI/Azure; delegate to the shared extractor.

Find the CheckpointConfig for an event source. Walks known relationships: Task -> Agent -> Crew. Flow and Agent carry their own checkpoint fi

Format messages with role labels for summarization. Skips system messages. Handles None content, tool_calls, and multimodal content blocks.

Apply the executor's stop words to the LLM for the duration of one call. Uses :func:`crewai.llms.base_llm.call_stop_override` so the LLM's stop

Save task result to unified memory (memory or crew._memory).

Show version update message if a newer version is available. Only displays when verbose mode is enabled and not running in CI/CD.

Validate and parse structured output. Args: response: Raw response string response_format: Optional Pydantic model fo

Build a comprehensive field description including constraints. Embeds format, enum, pattern, min/max, and example constraints into the descri

Convert a result to a Pydantic model or JSON. Args: result: The result to convert. Usually a JSON string, but a Pydantic inst

Create an agent executor for the agent. Returns: An instance of the CrewAgentExecutor class.

Enforce the requests per minute (RPM) limit if applicable. Args: request_within_rpm_limit: Function to enforce RPM limit.

Execute an action with the given arguments.

Emit a final ``FlowFinishedEvent`` and finalize the trace batch. Pairs with ``flow.defer_trace_finalization = True`` (or ``Conversati

Freeze the current Edge into ``docs/v<version>/`` and update docs.json. Args: version: Release version as ``"X.Y.Z"`` (no leading ``v``).

Build an expression with the standard Flow runtime context.

Parse the ``application[/action][@connection_uuid]`` syntax. Raises: ValueError: If the selector does not follow the supported sy

Convert MCP server references/configs to CrewAI tools. Delegates to :class:`~crewai.mcp.tool_resolver.MCPToolResolver`.

Increment the tools errors counter.

Executes the Crew's workflow for each input and aggregates results. Args: inputs: List of input dictionaries, one per execution.

Retrieve relevant memories. ``shallow`` embeds the query directly and runs a single vector search. ``deep`` (default) uses the Recall

Register a global after_llm_call hook. Global hooks are added to all executors automatically. This is a convenience function for registering

Reset specific or all memories for the crew. Args: command_type: Type of memory to reset. Valid options: 'memory'

GET a URL through :func:`safe_get`, refusing bodies over *max_bytes*. The body is streamed and abandoned as soon as it crosses the limit, so an

Search for similar documents using a query. Keyword Args: collection_name: Name of the collection to search in. query

Set available input files. Args: files: Dictionary mapping file names to file inputs.

Get the git status in porcelain format.

Generate a standard prompt for task execution. Returns: A dictionary containing the constructed prompt(s).

Save the documents to the storage asynchronously. This method should be called after the chunks and embeddings are generated. Raises

Build a minimal docs/ tree with one previous snapshot + Edge + docs.json. The shape mirrors the real repo: an Edge directory with per-locale fold

Turn a raw import-time exception into a user-actionable finding.

Run one telemetry span method against a mocked tracer. The singleton is reset first: it caches the provider built on the very first construct

Extract token usage and response metadata from Anthropic response.

Get or create a code interpreter for the specified thread. Args: thread_id: Thread ID for the code interpreter session R

Return ``True`` if *fn* cannot survive a serialize/deserialize round-trip. Built-in functions, plain module-level functions, and classes produce

Read the user-data JSON file, returning ``{}`` on missing/corrupt.

Prepare documents for ChromaDB by extracting IDs, texts, and metadata. Args: documents: List of BaseRecord documents to prepare. Ret

Resolve an ``MCPServerConfig`` into tools. Returns ``(tools, clients)`` where *clients* is always empty for native tools (clients are

Run an async coroutine from sync context without blocking event loop.

Set status for an explicit plan step reported by the planner.

Promote a skill to INSTRUCTIONS disclosure level. Idempotent: returns the skill unchanged if already at or above INSTRUCTIONS. Args:

Add a tool result to the cache. Declared failures are never stored: replaying one would make a transient error permanent for the rest

Asynchronously emit an event to registered async handlers. Only processes async handlers. Use in async contexts. Args: s

Return the bytes of a single-sheet XLSX holding *rows*.

Call Anthropic messages API. Args: messages: Input messages for the chat completion tools: List of tool/function defi

Clean up Stagehand resources.

Return the configured declarative flow source for flow projects.

Create a new collection in Qdrant. Keyword Args: collection_name: Name of the collection to create. Must be unique. v

Extract output from code interpreter response stream. Args: response: Response from code interpreter execution Returns: Extr

Restore a Crew from a checkpoint, ready to resume via kickoff(). Args: config: Checkpoint configuration with ``restore_from`` set

Normalize a provider's raw usage dict into a ``UsageMetrics``. Accepts the full set of key aliases CrewAI providers emit: ``prompt_to

Get all registered global after_tool_call hooks. Returns: List of registered after hooks

Get the current parent event ID from the stack.

Return ``[tool.crewai].project_id`` if the project has one. Read-only and safe to call from library code: it never creates or modifies anythi

Initialize a new trace batch (thread-safe)

Return ``(yanked, reason)`` for the currently installed version.

Check if a response from the LLM is a list of tool calls. Supports OpenAI, Anthropic, Bedrock, and Gemini formats. Args: response: T

Check if tracing is enabled in current execution context. Returns: True if tracing is enabled in context, False otherwise. Return

Return ``message`` with the cache-breakpoint flag set. Returns a new dict so callers can safely pass literal dicts.

Mark a todo as completed by step number.

Push an event ID and type onto the scope stack.

Register a global hook for an interception point.

Save the data to the specified file using pickle. Args: data: The data to be saved to the file.

Convert a dotted path string to the callable it references. If *value* is already callable it is returned as-is, with a warning if it cannot

Validate an audio file against constraints. Args: file: The audio file to validate. constraints: Audio constraints to validate ag

Resolve a declared or live LLM the way a crew agent's ``llm`` resolves. ``create_llm`` is the shared helper the crew/agent declaration layer

Common payload refinement for all tools.

Helper to create a skill directory with SKILL.md.

Emit one fake ``LLMCallCompletedEvent`` with ``current_flow_id`` pinned to ``flow_id``. Runs in a freshly-copied context so the value the bus

Format messages for OpenAI API.

Format messages for Converse API following AWS documentation. Note: Returns dict[str, Any] instead of LLMMessage because Bedrock uses

Get or create the Gemini client.