Async call to OpenAI Chat Completions API. Falls back to the Responses API for models not served by chat completions, mirroring :meth

Return (sandbox, should_delete_after_use).

Return (sandbox, should_kill_after_use).

Internal async implementation of AgentCard fetching.

Re-run the executor flow asynchronously using feedback messages.

Async cross-process lock that acquires/releases in an executor.

Process pending async tasks and return their outputs.

Update lineage fields after a successful checkpoint write. Sets ``_checkpoint_id`` and ``_parent_id`` so the next write records the c

Check if tool has reached its usage limit. Args: tool: The tool to check tool_name: The name of the tool (used for er

Clear fired OR listeners for cyclic flows.

Collapse unfinished display-only plan steps once the task succeeds.

Compute SHA-256 hash of file content. Uses streaming for FilePath sources to avoid loading large files into memory.

Snapshot the cumulative usage counters backing this agent's LLM. The counters live on the LLM instance (or the agent's token process

Resolve the default checkpoint location. When the caller passes the default directory path, check whether a SQLite database exists at the con

Block until all pending background memory saves have completed. Covers the crew memory, per-agent memories, and the manager agent's m

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

Tighten a dedicated config directory to 0o700. Skips directories shared with other users or content (the system temp dir and the current work

Register checkpoint handlers on the event bus once, lazily.

Create payload indexes for efficient filtering.

Executes all listeners and routers triggered by a method completion. This internal method manages the execution flow by: 1. First exe

Executes a flow's start method and its triggered listeners. This internal method handles the execution of methods marked with @start

Executes tasks sequentially and returns the final output. Args: tasks (List[Task]): List of tasks to execute manager

ChatCompletion / ChatCompletionChunk share the choices-shape; delegate to the shared extractor.

Azure ``ChatCompletions`` / ``StreamingChatCompletionsUpdate`` share the choices-shape; delegate to the shared extractor.

Extract and format thinking block from content block. Args: content_block: Content block from Anthropic response Returns

Fetch columns for *table_name* using a parameterised query. The table name is bound via SQLAlchemy's ``:param`` syntax to prevent SQL

Parse an OpenAI-shaped ``GET /models`` response.

Filter methods by attribute presence, preserving exact callable types. Args: methods: Dictionary of methods to filter. attribute:

Find JSON objects/arrays in text and pretty-print them.

Format messages for Gemini API. Gemini has specific requirements: - System messages are separate system_instruction - Content

Format messages according to provider requirements. Args: messages: List of message dictionaries with 'role' and 'content' keys.

Format prompt template with input values. Args: prompt: Template string. inputs: Values to substitute. Retur

Convert CrewAI tools to Converse API format following AWS specification.

Generate, display, and optionally edit release notes. Args: version: Version being released. tag_name: Tag name for the release.

Safely gets the agent's backstory, providing a default if not available. Returns: The agent's backstory or a default value.

Parse the click requirement directly from a pyproject.toml file.

Get the filesystem path if file source is FilePath. Args: file: The file input to check. Returns: Path if source is FilePath

Get inference configuration following AWS Converse API specification.

Generate a search query for the knowledge base based on the task description.

Get max replans from planning config or default to 3.

Look up a dotted attribute path inside ``pyproject_path``. The file must declare ``crewai`` in ``[project].dependencies`` for the lookup to s

Get the reasoning effort level from the agent's planning config. Returns: The reasoning effort level: "low", "medium", or "high".

Get A2AServerConfig from an agent's a2a configuration. Args: agent: The Agent instance to check. Returns: A2AServerConfig if

Build a getattr side_effect that resolves the patched tool name to ``tool_cls`` while delegating every other lookup (e.g. the ``is_deprecated_

True when either shard already holds persisted records.

Return True when the project has a Poetry lockfile.

Return True when the project has a uv lockfile.

Heuristically reject embedding/audio/image/etc. models by their id.

Detect if we're running in a test environment.

Return True if *instance* validates against *schema*.

Return a cross-process lock context manager, or nullcontext if no lock name.

Create a TraceBatchManager with a pre-set trace_batch_id (simulating first-time user).

Create a FirstTimeTraceHandler wired to a TraceBatchManager.

Create a cache key from file hash and provider.

Build a message list that will produce exactly *n* chunks. Each message has 400 chars (~100 tokens). With max_tokens=100 returned by

Observe a completed step via LLM or a lightweight heuristic.

Open the deployment show page in the user's browser when possible.

Open an existing shard or create a new one at *path*.

Poll a GitHub PR until it is merged. Exit if closed without merging. Args: branch_name: Branch name to look up the PR. label: Hum

Return the start-time state for a completion/failure event's task. Tasks can run async/overlapping, so the event's task identity is m

Prepare Git for deploy while preserving remote deploy when possible.

Register a handler for the given event type. Args: event_type: The event class to listen for handler: The handler fun

Record that this model must use the Responses API.