Repository navigation
This release upgrades AnythingLLM and improves the reliability of stack-check.sh, especially during cold starts.
Highlights
- Updated AnythingLLM from 1.16.2 to 1.18.0 across the full, chat-ui, and voice-chat stacks.
- Increased stack-check.sh's default request timeouts to 120 seconds for LLM routing and 60 seconds for embeddings and TTS, with configurable request and connection timeouts.
- Prevented false health-check passes after failed HTTP transfers, inaccessible container engines, or checks with no recognized running services.
- Recognized stopped services in active Compose projects now cause the check to fail. Absent optional services are still skipped.
- Local HTTP probes bypass proxies and curl configuration files, with clearer HTTP status and timing diagnostics.
- The routing check prefers the first downloaded model's
ollama-chat/<model>alias when available, retaining the legacyollama/<model>fallback. - Added focused offline regression tests and updated troubleshooting documentation.
The timeout settings affect only stack-check.sh, not application request timeouts or Docker health-check settings.
Updating existing deployments
Back up persistent data before upgrading, then follow the update instructions in the README for this release.
Some container images use moving tags. Deploying this tagged snapshot later may therefore pull newer images.
See the changelog for details.