Extract field values from data using JSONPath-like expressions. Supports simple expressions like: - "query" -> data["query"]

Extract images from a message. For data URLs (e.g., "data:image/png;base64,iVBOR..."), only the base64 data portion is extracted. This i

Extract model ID from a unified resource ID. Args: unified_id: The unified resource ID (decoded or encoded) Returns: Mo

Find first team with access to the requested model

Generate an API key based on the provided data. Docs: https://docs.litellm.ai/docs/proxy/virtual_keys Args: mod

Get access token and project id 1. Check if credentials are already in self._credentials_project_mapping 2. If not, load cre

Get list of allowed agent IDs for the given user/key based on permissions. Returns: List[str]: List of allowed agent IDs

Get api key metadata, falling back to deleted keys table for keys not found in active table. This ensures that key_alias and team_id are preserve

Returns deployments available for pass-through endpoints (based on load balancing strategy) Similar to get_available_deployment, but

Helper function to get the bedrock provider from the model handles 3 scenarions: 1. model=invoke/anthropic.claude-3-5-sonnet

Logic for returning complete model list for a given key + team pair

Get the complete URL for the RAGFlow API call. Constructs URL based on endpoint type: - Chat: /api/v1/chats_openai/{chat_id}

Calculate cost for code interpreter feature. Azure: $0.03 USD per session (from model cost map) OpenAI: $0.03 USD per sessio

" OpenAI: $2.50/1k calls Azure: $0.1 USD per 1 GB/Day (storage-based pricing) Doc: https://platform.openai.com/docs/pricing#

For a given model id, return the model info 1. Check if model_id is in model info 2. If not, check if litellm model name is

Used for enforcing `disable_end_user_cost_tracking` param. service_type: "litellm_logging" or "prometheus" - used to allow prometheus only d

Return CometAPI-specific error class

List all models available on the Ollama server via /api/tags endpoint.

Return the OCI inference base URL, respecting any explicit api_base override. If ``api_base`` already ends with a fully-formed OCI action path

Get optional parameters for the image edit API. Args: model: The model name image_edit_provider_config: The

- Check if org id in proxy Org Table - if valid, return LiteLLM_OrganizationTable object - if not, then raise an error Args:

Get the provider configuration for a given custom_llm_provider. Args: custom_llm_provider: The provider identifier (e.g.

Get the provider specific headers for the given custom llm provider. Supports comma-separated provider lists for headers that work a

Return the MCP servers published to the AI Hub via /v1/mcp/make_public. Default (litellm.public_mcp_hub_strict_whitelist=True): mirr

Get the required headers for the CountTokens API. Args: api_key: The Anthropic API key Returns: Dic

Sets the Appropriate OpenAI headers for the response and forward all headers as llm_provider-{header} Note: _response_headers Passed here s

Extract the router deployment model_id from litellm_params. Checks both litellm_metadata and metadata for model_info.id. Used by cost

Get router settings method, returns a dictionary of the settings and their values. For example get the set values for routing_strateg

Converts a list of messages to a string

Get current team information

Returns team membership object if user is member of team. Do a isolated check for team membership vs. doing a combined key + team + user + t

Returns: - List of model name strings - Empty list if no models set - If model_access_groups is provided, only return models that are

Create OpenAI client with generated key

Get the user id from the request

Handles any messages to chat completion str messages conversion Relevant Issue: https://github.com/BerriAI/litellm/issues/9494

Static method to initialize WebSearchInterceptionLogger from proxy config. Used in callback_utils.py to simplify initialization logi

Check if a value contains an Anthropic OAuth token (sk-ant-oat*).

True iff the exception means the database could not answer at the infrastructure level (connection refused, socket/interface failure,

Check if the client opted into MCP debug mode. Looks for ``x-litellm-mcp-debug: true`` (case-insensitive) in the request hea

Check if a license (or compound SPDX expression) is acceptable.

Return True when this is a GET list route whose results should be served from the DB (user-scoped) rather than forwarded upstream.

Check if a tool is a web search tool (native or LiteLLM standard). Detects: - LiteLLM standard: name == "litellm_web_search" - OpenA

Yield every prefix form that may appear in tool names for ``server``. Always includes the *current* prefix returned by ``get_server_prefix``.

List team guardrail submissions. Returns only guardrails with a team_id. Admins see all submissions. Non-admin users see submissions for tea

`(response_id, metadata_lookup_id)` for each listed model, de-duplicated by response_id while preserving order. For team-scoped rows

Factory for fake LiteLLM_SpendLogs rows.

Get model_id from logging object or request metadata. The router sets model_info.id when selecting a deployment. This tries multiple

Apply all xAI normalization passes in order.

Normalize request routes by replacing dynamic path parameters with placeholders. This prevents high cardinality in Prometheus metrics by col