Classify a prompt by complexity. Args: prompt: The user's prompt/message. system_prompt: Optional system pro

Generate a mock MCP server record from config.yaml

Extract client IP from a FastAPI request for MCP access control. Security: Only trusts X-Forwarded-For if: 1. use_x_forwarde

Get the global PolicyRegistry singleton. Returns: The global PolicyRegistry instance

Build a minimal kwargs dict representative of a litellm callback invocation.

Return a UUID4 using the selected backend.

Transform the image generation request to Gemini format For Gemini 2.5 Flash Image Preview, use the standard Gemini format with resp

Accept ``redirect_uri`` when it is (a) same-origin with the proxy's own request origin, (b) loopback, (c) listed in the ``MCP_TRUSTED_REDIRECT

Returns -> Deployment or None Raise Exception -> if model found in invalid format Optimized with O(1) index lookup instead

- Check if user id in proxy User Table - if valid, return LiteLLM_UserTable object with defined limits - if not, then raise an error

Helper to checks if provided route is an OpenAI route Returns: - True: if route is an OpenAI route - False:

A short unique token per call/run, so concurrent runs and the shared response cache never collide on prompts, tags, or customer ids.

Async get cache implementation. Used for embedding calls in async wrapper

Return a urlsafe-base64 managed ID string (trailing ``=`` stripped).

Returns the provider config for a given provider. Uses O(1) dictionary lookup for fast provider resolution. Python classes t

Mint managed IDs for raw provider values listed in ``BUILTIN_OUTPUT_ID_FIELD_MAP`` and swap them into *body*. Returns the same *body* ob

The same as `litellm.litellm_core_utils.token_counter`. Kept for backwards compatibility.

Validate that MCP servers requested on a key are within the allowed scope. Rules: - If key is in a team: key's mcp_servers must be a sub

Make sure each key starts with the given namespace

Pseudo-destructor to be invoked to clean up global data structures when router is no longer used. For now, unhook router's callbacks

Get the registered MCP Servers from the registry and union with the config MCP Servers

Invoke an agent using the A2A protocol (JSON-RPC 2.0). Supported methods: - message/send: Send a message and get a response - messag

Pass through endpoint handler, makes the httpx request for pass-through endpoints and ensures logging hooks are called Args: request

Transform the image generation response to the litellm image response

Decode *managed_id*. Returns ``None`` for anything that is not a passthrough managed ID — raw OpenAI IDs, unified-endpoint IDs, garbage,

Flush the logging queue. Waits until every enqueued task has completed. ``queue.join()`` blocks on the queue's unfinished-task counte

Must return the base completion call args

Returns -> Deployment or None Raise Exception -> if model found in invalid format

Prints the given `print_statement` to the console if `litellm.set_verbose` is True. Also logs the `print_statement` at the debug level using

Count the number of tokens in a given text using a specified model. Args: model (str): The name of the model to use for tokenization. De

Use `/team/member_add` AND `/team/member/delete` to add/remove new team members You can now update team budget / rate limits via /team/updat

Whether ``model`` uses adaptive thinking (``output_config.effort``). The model cost map is authoritative: an explicit ``supports_adaptive_thi

Normalize the endpoint URL for a specific OpenTelemetry signal type. The OTLP exporters expect endpoints to use signal-specific path

Attempt to acquire the lock for a specific cron job using Redis. Uses the SET command with NX and EX options to ensure atomicity.

Validate tool calls against the blocking service (fail-open).

Arize is used mainly for LLM I/O tracing, sending router+caching metrics adds bloat to arize logs

Return the prefix for a server. When the short-prefix mode is enabled (``LITELLM_USE_SHORT_MCP_TOOL_PREFIX``) a three-character base62 ID is

Use this to create a new INTERNAL user with a budget. Internal Users can access LiteLLM Admin UI to make keys, request access to models.