Parse RAGFlow model name format: ragflow/{endpoint_type}/{id}/{model_name} Args: model: Model name in format ragflow/cha

Helper function to route data/content parameters correctly for httpx requests This prevents httpx DeprecationWarnings that cause memory leak

Read the first WS frame and return (model, raw_message), or None on error. Sends an appropriate error frame and closes the socket before returnin

Resolve the Entra user object ID from request data or auth context. Returns the strongest available identity walking down four sources, in

Build a ModelResponse-like object with a single assistant message.

Rotate a single key using existing regenerate_key_fn and call the rotation hook

Sanitize ``error_information`` before it lands in ``LiteLLM_SpendLogs.metadata``. Provider errors are stored verbatim via ``str(original_exc

Scan tool call arguments with allow/block/mask treatment (in-place modification). Each tool call is sent as a ``tool_event`` using the canoni

Helper function to set metadata fields that require premium user checks Args: object_data: The team/key/organization/project data ob

Determine if tool calls should be converted to JSON mode

Replace ``b64_json`` payloads in image-gen responses before save. Image-edit and image-generation responses carry the full base64 PNG/JPEG (1

Transform a chat completion chunk to a response API chunk. This currently handles emitting the OutputTextDeltaEvent, which is used b

Watch engine PID via os.waitpid() in a dedicated thread. The thread blocks on os.waitpid(pid, 0) which is a kernel-level wait and wit

Unwrap encrypted_content to extract model_id and original content. Returns: Tuple of (model_id, original_encrypted_content).

Triggers a budget alert if the token has reached EMAIL_BUDGET_ALERT_MAX_SPEND_ALERT_PERCENTAGE (default 80%) of its max budget. This is a

Close the dedicated HTTP clients used by this logger.

Count tokens using Google Gen AI Studio countTokens endpoint. Args: contents: The content to count tokens for (Google Ge

Convert a plain-text or multipart content message to OCI format.

Add a failure callback to litellm._async_failure_callback

Add provider specific params to optional_params

Fold a worker-produced VCR outcome into the controller's session stats. No-op outside the xdist controller path: * On a worker, ``_session_s

Merge key metadata tags into request_data before _tag_max_budget_check.

Assert tool_event has canonical PANW schema (no legacy keys).

Passes the entire stream to the guardrail This is useful for guardrails that need to see the entire response, such as PII masking.

Test access control validation

Async function to read a secret from AWS Secrets Manager Returns: str: Secret value Raises: ValueErr

Async function to rotate a secret by creating a new one and deleting the old one. This allows for both value and name changes during

Internal method to check the type of the result & cache used and adds the result to the cache accordingly Args: result:

Helper method to attach object_permission to a dictionary if object_permission_id is set. This function: 1. Checks if the dictionary has

Return a Prisma `where` fragment that scopes a managed-resource listing to records the caller is allowed to see. - ``{}`` means no scoping (p

if 'api_base' in request body. Check if complete credentials given. Prevent malicious attacks. Supplying an ``api_key`` is necessary but not

Enforce key/team tool allowlist (metadata.allowed_tools). No DB in hot path — effective allowlist is read from valid_token.metadata and valid

Special route. Allows UI link share user to update their password. - Get the invite link - Validate it's still 'valid' - Check if us

Main cleanup function. Deletes old spend logs in batches. If pod_lock_manager is available, ensures only one pod runs cleanup.

Close the aiohttp client session and transport if we own them.

OpenAI tool invokes: { "role": "assistant", "content": null, "tool_calls": [ { "id": "call_abc123",

Create an Anthropic-formatted error response dict. Anthropic error format: { "type": "error", "error

Generate a Responses API config class dynamically from JSON configuration. Parallel to create_config_class() but for /v1/responses endpoints.

Route success logging to async and/or sync handlers for this request. ``prefer_async_handlers`` only bypasses the sync-SDK-only shortcut (e.g

Convert GitLab path IDs like 'invoice/extract' → 'gitlab::invoice::extract

Field-level opt-in gate: a non-admin team member may only set `access_group_ids` on a (team) key if their team has opted in by adding

Get the info on the mcp server specified by the `server_id` Parameters: - server_id: str - Required. The unique identifier of

Get all updates from the queue and return all updates aggregated by daily_transaction_key. Works for both user and team spend updates.

Drain state aggregator and apply to LiteLLM_AdaptiveRouterState. Returns number of cells flushed.

Create a LiteLLMSendMessageResponse from a dict. Args: response_dict: Dict with A2A response structure reque

Build from one ``standard_logging_guardrail_information`` entry. Reads the canonical, provider-agnostic ``StandardLoggingGuardrailInformation

Get the latest health check for each model. Uses DB-level DISTINCT ON (model_id, model_name) with ORDER BY checked_at DESC (