No model found
Confirm Ollama is running, the model exists, and the endpoint is reachable from Odysseus.
Troubleshooting
Fix Odysseus by symptom, not guesswork.
Most Odysseus problems come from one of four layers: Docker, model serving, GPU/runtime setup, or external tools such as search and databases.
Debug order
Check logs, confirm the service is running, test the model server, then change Odysseus settings.
Docker GPU not detected
Check NVIDIA drivers, WSL GPU visibility, and container runtime support before editing app config.
Admin password missing
Look for first-run output, generated secret files, or reset instructions from the official docs.
ChromaDB client conflict
A dependency mismatch can break memory or vector features if the wrong client package is installed.
SearXNG returns no results
Google can rate limit or block automated queries. Use sane rate limits and alternative engines.
localhost does not load
Confirm the app port, container bind address, firewall, and reverse proxy rules.