Returns the final embedding cached response and a boolean indicating if all elements in the list have a cache hit For embedding resp

Reap ALL zombie child processes via waitpid(-1, WNOHANG). Returns a set of reaped PIDs. As PID 1 in Docker (or any process that spaw

Parse collected SSE bytes into an Anthropic Messages response dict. Processes SSE events in order: - message_start -> enve

Record exception information on the span using OTEL standard methods. This extracts error information from StandardLoggingPayload an

Relay an elicitation request to the downstream MCP client. Uses the ServerSession's elicit_form() or elicit_url() methods to send the eli

Resolve a category file path that may be relative. Paths in policy templates (e.g. category_file) are often stored as relati

Pick a stable per-conversation key for owner-cache attribution. Order: 1. Honor a client-supplied session id (`litellm_session_id` on eithe

Return a dict mimicking ResponsesAPIResponse with a message output.

When users call `litellm.anthropic.messages.*` with a non-Anthropic model and `thinking={"type": "enabled", ...}`, LiteLLM converts t

Reject path traversal and URL-encode each path segment.

Score available models by MCP priority weights and return the best. Scoring strategy (per the MCP spec, priorities are 0-1 floats): * **cost

Sets Langfuse specific metadata attributes onto the OTEL span. All keys supported by the vanilla Langfuse integration are mapped to

Determine if Prometheus metrics should be skipped for invalid API key requests. This is a centralized validation method that extract

Pull the credential_b64 value passed to the most recent upsert call.

Default reconsider-guard stubs: pretend bot closed + no cooldown. The new safety guards (`was_closed_by_agent_shin`, `seconds_since_l

Copy standard_logging_guardrail_information from request_data into logging_obj. The @log_guardrail_information decorator writes guardrail info to

Transform a Responses API input_file item to a Chat Completion file item Args: item: Dictionary containing input_file ty

For team-scoped DB rows, replace `model_name` with the public name in `model_info.team_public_model_name` before returning. The DB column and

Auto-inject anthropic-beta headers based on features used. Handles: - context_management: adds 'context-management-2025-06-2

Helper function to update a single user. Used by both user_update and bulk_user_update endpoints. Returns the updated user data or raise

Verify the caller is authorized to manage the given team. Access is granted if: - Caller is a proxy admin, OR - Caller is an org adm

Stand-in for a pytest TestReport delivered to the xdist controller. Only the attributes ``aggregate_report_outcome`` reads (``nodeid``, ``whe

Escape a string for safe embedding in a double-quoted YAML scalar.

Async completion with MCP integration. This function handles MCP tool integration following the same pattern as aresponses_api_with_mcp.

Build a Cohere ``chatHistory`` list from an OpenAI-format message array. - All messages except the *last user message* are included. The caller p

Add two numbers

Add fallback headers to the response Args: response: The response to add the headers to attempted_fallbacks: The number of f

Add a new member to a team - add team id to user table - add team member w/ budget to team member table Returns created/existing us

Add metadata to openai optional parameters, excluding hidden params. OpenAI 'metadata' only supports string values. Args: param

Add the vertex credentials for the given project-id, location

Asynchronously calls the `image_edit` function with the given arguments and keyword arguments. Parameters: - `args` (tuple): Positional

Run edits in order; return a single ``PolyfillResult``. The dispatcher is async so async editors (``compact_20260112``) can ``await`` the con

Apply a guardrail to text input and return the processed result. This endpoint allows testing guardrails by applying them to custom text inp

Apply Ovalix guardrail to the given inputs (request or response text). Used by the unified guardrail flow and the /apply_guardrail A

Apply OpenAI moderation guardrail using the unified guardrail interface. This method is called by the UnifiedLLMGuardrails system fo

Filters MCP tools the end user cannot access based on their object_permission.mcp_servers / mcp_access_groups settings.

Write redacted messages back to whichever field(s) the caller used. Mask/anonymize paths take a synthesised messages list (from :func:`build_

Async: Send a streaming message to an A2A agent. If litellm_params contains custom_llm_provider, routes through the completion bridge.

Resolve a managed vector store id and enforce ownership if it exists. Unknown ids are treated as provider-native ids and are not rejected he

Returns the set of model names where every backing deployment is currently marked unhealthy by background health checks (and the heal

Health check for SQS by sending a small test message to the configured queue.

Allow user to modify outgoing streaming data -> per chunk Covers: 1. /chat/completions

Register a Lua script with Redis asynchronously. Works with both standalone Redis and Redis Cluster. The returned callable n

Build masked debug response headers. Parameters ---------- inbound_headers : dict Raw headers received f

Follows the exact same API spec as `OpenAI's Chat API https://platform.openai.com/docs/api-reference/chat` ```bash curl -X POST http://

Check if the batch JOB has been tracked. - get all status="validating" and file_purpose="batch" jobs - check if batch is now

Split text into chunks using RecursiveCharacterTextSplitter. Args: text: Text from OCR (if used) file_conten

Cleanup after each test

Main entry point for deleting expired UI session keys. Uses PodLockManager to ensure only one pod runs cleanup in multi-pod deploymen

Clear router caches and reload models.

Helper function to clear the Prometheus registry

Construct a DATABASE_URL from individual environment variables. Returns: Optional[str]: The constructed DATABASE_URL or None if requi

Count tokens in text using litellm.token_counter. Args: text: Text to count tokens for Returns: Tok

Create an A2A client for the given agent URL. This resolves the agent card and returns a ready-to-use A2A client. The client can be reus

-------------------------------------------------- NOTE: Copy of `litellm/db_scripts/create_views.py`. ----------------------------------

Helper function to create a mock HTTP response.

Create a user according to SCIM v2 protocol

Decode provider and model_id from encoded character_id.

Helper that deletes the list of tokens from the database - check if user is proxy admin - check if user is team admin and key is a team

Normalize batch file IDs to managed unified IDs before DB persistence.