Summary
When a user adds a custom cloud model provider/base URL and key, chat/model routing can still surface the OpenHuman backend resource exhausted / budget-exhausted error instead of using the configured custom provider.
Reproduction
- Exhaust or otherwise hit the OpenHuman/TinyHumans hosted model budget.
- Add a custom cloud model provider in Settings -> AI, including a custom endpoint/base URL and API key.
- Select that custom provider/model for chat.
- Send a prompt.
Expected
The selected custom provider is used for inference. The exhausted hosted OpenHuman budget should not block custom-provider calls.
Actual
The app still reports a resource exhausted / budget-exhausted response, suggesting the request path still uses the hosted OpenHuman backend/provider for at least some selected custom-provider cases.
Notes
Related but not identical to #2075, which covers Composio URLs being built from an LLM base URL. This issue is specifically about model inference routing after adding a custom provider.
No API keys or provider secrets included.
Summary
When a user adds a custom cloud model provider/base URL and key, chat/model routing can still surface the OpenHuman backend
resource exhausted/ budget-exhausted error instead of using the configured custom provider.Reproduction
Expected
The selected custom provider is used for inference. The exhausted hosted OpenHuman budget should not block custom-provider calls.
Actual
The app still reports a
resource exhausted/ budget-exhausted response, suggesting the request path still uses the hosted OpenHuman backend/provider for at least some selected custom-provider cases.Notes
Related but not identical to #2075, which covers Composio URLs being built from an LLM base URL. This issue is specifically about model inference routing after adding a custom provider.
No API keys or provider secrets included.