Check for existing issues

• I have searched the existing issues and checked that this is not a duplicate.
• Related issues: #15228, #21665
• This appears distinct from #15228 because it is reproducible via direct proxy API calls and points to alias resolution happening after team/virtual-key authorization.

What happened?

When using LiteLLM Proxy with a model alias configured via the UI, the alias cannot be used in API calls if access is restricted by team / virtual key allowed models.

Example setup:

• Model alias in UI:
  • AgentX-LLM -> openrouter/qwen/qwen3.5-flash-02-23
• Team / API key has access to:
  • openrouter/qwen/qwen3.5-flash-02-23

Request:

curl -X POST "https://<proxy>/v1/chat/completions" \
  -H "Authorization: Bearer <API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "AgentX-LLM",
    "messages": [
      { "role": "user", "content": "say ping" }
    ]
  }'

Actual behavior:

• Request returns 401 team_model_access_denied
• Logs show LiteLLM checked the alias name directly:
  • Tried to access AgentX-LLM

Expected behavior:

• LiteLLM should resolve the alias first:
  • AgentX-LLM -> openrouter/qwen/qwen3.5-flash-02-23
• Then apply permission checks against the resolved model
• Since the resolved model is allowed, the request should succeed

Suspected root cause:

Current auth flow appears to be:

1. Receive model = AgentX-LLM
2. Check team / key allowed models
3. Reject because alias name is not in allowed_models
4. Alias is never resolved

Expected flow:

1. Receive model = AgentX-LLM
2. Resolve alias -> real model
3. Check permissions on resolved model
4. Forward request

Steps to Reproduce

1. Deploy LiteLLM Proxy with team-based access control and virtual keys enabled
2. Configure a real model such as openrouter/qwen/qwen3.5-flash-02-23
3. Create a model alias in the UI:
   • AgentX-LLM -> openrouter/qwen/qwen3.5-flash-02-23
4. Create a team / virtual key whose allowed models include only openrouter/qwen/qwen3.5-flash-02-23
5. Call /v1/chat/completions with model: "AgentX-LLM"
6. Observe 401 team_model_access_denied
7. Check logs and observe Tried to access AgentX-LLM

Relevant log output

{
  "error": {
    "message": "team not allowed to access model...",
    "code": "401"
  }
}

Tried to access AgentX-LLM

What part of LiteLLM is this about?

Proxy

What LiteLLM version are you on ?

main-v1.82.3-stable.patch.2 (self-hosted Docker Compose deployment)

Twitter / LinkedIn details

No response