Highlights
Connected chat models now appear consistently in Settings, Chat, GraphRAG, and the Prompt Library.
- Provider readiness and model verification now use one consistent, fail-closed policy, so incomplete or stale connections are not used silently.
- The Prompt Library recognizes profile-based providers and includes safer saving, generation, keyboard navigation, and recovery.
- Provider edits immediately refresh active RAG and GraphRAG runtimes, preventing outdated credentials or endpoints from being reused.
- When a saved default model is unavailable, the interface asks for an explicit selection instead of switching providers behind the scenes.
Privacy clarification
Settings and API keys are stored unencrypted in Obsidian local plugin storage and plugin data.json. They may be included in vault sync and local backups.