BerriAI/litellm · error · Exception

LLM Router not initialized. Ensure models added to proxy.

Error message

LLM Router not initialized. Ensure models added to proxy.

What it means

Raised by ManagedFilesHook.create_file_for_each_model when llm_router is None. In managed-files mode a single POST /v1/files upload is fanned out to one real upload per target model, and each goes through llm_router.acreate_file (routed from litellm/proxy/openai_files_endpoints/files_endpoints.py:243 when the 'managed_files' hook is active). With no router the proxy has zero deployments, so the hook fails fast before contacting any provider.

Solutions

1. Add at least one entry under model_list in the proxy config and restart so llm_router initializes
2. If models load from the database, check startup logs for router init errors and fix the root cause (DATABASE_URL set, migrations applied, DB reachable)
3. Wait for /health/liveliness to report the router ready before sending file uploads
4. Programmatic callers: pass a live Router (proxy_server.llm_router) instead of None

Example fix

# before (config.yaml)
model_list: []
general_settings:
  master_key: sk-1234

# after
model_list:
  - model_name: gpt-4o
    litellm_params:
      model: openai/gpt-4o
      api_key: os.environ/OPENAI_API_KEY
general_settings:
  master_key: sk-1234

from litellm.proxy import proxy_server

def managed_files_upload_ready() -> bool:
    router = getattr(proxy_server, "llm_router", None)
    return router is not None and len(router.get_model_names()) > 0

# only upload when ready
if managed_files_upload_ready():
    await client.files.create(...)

When it happens

Trigger: POST /v1/files with managed files enabled while proxy_server.llm_router is None: config.yaml has no (or an empty) model_list, DB-backed models failed to load at startup, or the request arrives before router initialization completes.

Common situations: First-run proxy yaml with general_settings configured but model_list omitted; models expected from the database while DATABASE_URL is unset so nothing loads; integration tests that call the hook directly without passing a Router.

Related errors

• DB not connected. This endpoint needs a database; set…
• LLM router not found. Please set it up by passing in a…
• 503
• api_base is required
• api_base is required for hosted_vllm embeddings

AI-assisted analysis of BerriAI/litellm@77b7c6c40c (2026-08-18). Data as JSON: /api/errors/fb918c11773a807b. Report an issue: GitHub.

Appendix: source

Thrown at enterprise/litellm_enterprise/proxy/hooks/managed_files.py:1017

            # Get all cache keys matching the pattern file_id:*
            for file_id in litellm_managed_file_ids:
                # Search for any cache key starting with this file_id
                unified_file_object = await self.get_unified_file_id(file_id, litellm_parent_otel_span)

                if unified_file_object:
                    file_id_mapping[file_id] = unified_file_object.model_mappings

        return file_id_mapping

    async def create_file_for_each_model(
        self,
        llm_router: Optional[Router],
        _create_file_request: CreateFileRequest,
        target_model_names_list: List[str],
        litellm_parent_otel_span: Span,
    ) -> List[OpenAIFileObject]:
        if llm_router is None:
            raise Exception("LLM Router not initialized. Ensure models added to proxy.")
        responses = []
        for model in target_model_names_list:
            individual_response = await llm_router.acreate_file(model=model, **_create_file_request)
            responses.append(individual_response)

        return responses

    async def acreate_file(
        self,
        create_file_request: CreateFileRequest,
        llm_router: Router,
        target_model_names_list: List[str],
        litellm_parent_otel_span: Span,
        user_api_key_dict: UserAPIKeyAuth,
    ) -> OpenAIFileObject:
        responses = await self.create_file_for_each_model(
            llm_router=llm_router,
            _create_file_request=create_file_request,

View on GitHub (pinned to 77b7c6c40c)