FAQ
FAQ
The troubleshooting guide works from the failing boundary inward, which is the right approach when you do not yet know what broke. This page is for when you do: it is indexed by the thing you are staring at.
Configuration and startup
| Symptom | Page |
|---|---|
| The agent key I configured is not found | Agent not found |
| The server refuses to start after a config change | Server will not start |
| A tool I listed on the agent is not available to it | Tool not available |
Models
| Symptom | Page |
|---|---|
| The model ignores images I attach | Model ignores images |
| My local model never calls tools | Local model will not call tools |
A model I expected is missing from GET /api/v1/models | Model missing from the registry |
Retrieval and memory
| Symptom | Page |
|---|---|
| RAG returns nothing, or the agent ignores my documents | RAG returns nothing |
| The agent forgets things between conversations | Agent forgets between sessions |
Requests and runs
| Symptom | Page |
|---|---|
| Requests come back unauthorized | Requests are unauthorized |
| Files the agent produced have disappeared | Generated files disappear |
| A schedule did not fire | Schedule did not fire |
Next
For anything not listed above, these go wider:
- Troubleshoot a deployment - the boundary-by-boundary method.
- Observe runs - traces and audit entries for a specific run.
- - when the failure crosses subsystem boundaries.