- Add repairTruncatedCardsJson to recover complete cards when output
is cut off mid-card (e.g. token limit reached)
- Log raw LLM response to console on parse failure for debugging
- Make error panel text selectable and add "copy error" button
Change-Id: I4e76121138888234d42f84f3695cbbfd6beba886
When a note has no cached parallel notes, the empty state now shows a
clickable "Generate" button so users can trigger generation directly
without going through the command palette. Closes#2.
Change-Id: I879897168ef77c44e0ab570e9c094fa9af9f66d4
Implement SSE streaming for API backends that support it (OpenAI Chat
Completions and Anthropic Messages). During generation the view shows
a live streaming preview with accumulated text. Adds a streaming toggle
in settings (enabled by default). Non-streaming-capable formats (Gemini,
OpenAI Responses) and CLI backends continue using the existing path.
Change-Id: I3361a5ca5aec3e25a7d16e4fb585185f6fc0695e
