Only observe DOM mutations when dialog is open, preventing
unnecessary observer activity when the dialog is closed.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Added `isEditorFullscreen` state to track fullscreen mode for the Markdown editor.
- Integrated `MutationObserver` to detect changes related to the `.w-md-editor-fullscreen` class.
- Adjusted dialog and editor styles dynamically based on fullscreen state using `cn`.
- Enhanced UX by resetting fullscreen state on dialog close.
- Fix AlertDialogAction closing before async delete completes by using
asChild with Button and event.preventDefault()
- Add delete button to SourceInsightDialog header for consistent UX
- Add confirmation dialog in SourceInsightDialog
- Pass onDelete callback from SourceDetailContent to SourceInsightDialog
Addresses PR #334 feedback from lfnovo and cubic-dev-ai.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Use model_copy() instead of creating new AIMessage to preserve
response_metadata, id, usage_metadata, etc. Also adds test coverage
for malformed thinking tags pattern.
Addresses PR #333 feedback from lfnovo and cubic-dev-ai.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add button to navigate from insight popup to its parent source
- Uses existing modal manager to open source detail view
- Shows button only when source_id is available
- Add pendingModelOverride state to useNotebookChat hook
- Store model selection when no session exists yet
- Apply pending model override when session is auto-created on first message
- Simplify ChatColumn by using new setModelOverride function
- Add nullable_fields support to ObjectModel base class
- Configure ChatSession to allow model_override to be cleared to null
- Fix JSX conditional that rendered "0" when message_count was 0
- Display model name instead of raw ID in session manager
Fixes issues:
1. Switching to default model now persists correctly
2. Session list shows human-readable model names
3. Sessions with 0 messages no longer show "0" badge
- Add fix for quotes in environment variables causing empty URL errors
- Add SurrealDB configuration section (single container includes DB, v2 only)
- Add network timeout fix for slow connections and Chinese users
- Update table of contents with new sections
Addresses feedback from issue #316
Add ability to delete insights from sources with a confirmation dialog.
Uses AlertDialog component for a native React experience instead of
browser confirm().
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add thinking content cleaning to notebook and source chat graphs.
Previously, models that output <think>...</think> tags (like DeepSeek)
or malformed variants without opening tags (like Nemotron) would leak
reasoning content into user-visible responses.
Changes:
- chat.py: Clean AI response content before returning messages
- source_chat.py: Same fix for source-specific chat
- text_utils.py: Handle malformed output where opening <think> tag
is missing but </think> is present
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Use QUERY_KEYS.sourcesInfinite for infinite scroll query key
Starting with ['sources', ...] ensures mutations that invalidate
['sources'] will also invalidate the infinite scroll cache
- Use httpx.Timeout for chat service with short connect (10s) and
long read (600s) timeouts. Prevents 10 min wait on connection errors
Added remark-gfm plugin to enable GitHub Flavored Markdown support,
which includes proper table rendering. Updated all ReactMarkdown
usages with table styling components:
- ChatPanel (source and notebook chat)
- StreamingResponse (search/ask results)
- SourceInsightDialog
- SourceDetailContent
- TransformationPlayground
Tables now render with proper borders, headers with background,
and horizontal scroll for wide tables.
Users reported "Unable to Connect to API Server" errors on startup
because the frontend started before the API finished initialization
(database migrations, etc.).
- Add wait-for-api.sh script that polls /health endpoint
- Update supervisord configs to use wait script instead of sleep 5
- Waits up to 5 minutes for API to be ready before starting frontend
- Applies to both single-container and multi-container deployments
Fixes#315
Users with Ollama reported timeout errors on notebook chat while the
backend was still processing. The answer would appear after refresh.
- Frontend axios timeout: 5 min → 10 min
- Backend chat service timeout: 2 min → 10 min
Local LLMs can take several minutes for complex questions with large
contexts, especially on slower hardware.
Previously, notebook sources were limited to 50 items due to API default.
This adds pagination with infinite scroll to the sources column:
- Add useNotebookSources hook with useInfiniteQuery for paginated fetching
- Update SourcesColumn with scroll detection to load more sources
- Fetch 30 sources per page, loading more as user scrolls near bottom
- Added styles to ensure `.w-md-editor-fullscreen` behaves statically and adjusts height dynamically with `!static` and `!h-[calc(90vh-(var(--spacing)*41))]`.
The search page now reads `q` and `mode` query parameters from the URL,
enabling proper integration with the command palette's search/ask actions.
Changes:
- Read URL params using useSearchParams hook
- Initialize form fields and active tab from URL params
- Auto-trigger search/ask when arriving with URL params
- Handle subsequent navigations while on the search page
