Verifies that searching "Writing" in Quick Open shows the exact title
match first, followed by prefix matches. Also adds Refactoring test
entries to mock data and demo vault.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Each CLI invocation is a fresh subprocess — --resume depends on session
persistence that doesn't reliably work with -p mode. Switch to prompt-
embedded history: prior exchanges are formatted into each request using
formatMessageWithHistory + trimHistory (already existed as dead code).
- Remove sessionIdRef and --resume session tracking
- Always send system prompt (each request is stateless)
- Split sendMessage into doSend(text, history) to handle retry correctly
- Retry now passes correct history (excludes the retried exchange)
- Tests verify history accumulation, clear, retry, and system prompt
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The previous approach embedded conversation history as formatted text
in the prompt (<conversation_history> tags). The claude CLI's -p flag
treats this as a single user turn, losing turn boundaries — so the
model had no real multi-turn context.
Switch to using --resume with the session_id returned by the CLI's
init event. This lets the CLI manage conversation state natively:
- First message starts a new session (with system prompt)
- Subsequent messages resume via --resume (no system prompt needed)
- Clear and retry reset the session_id for a fresh start
Extract makeStreamCallbacks() to keep useAIChat complexity below
CodeScene threshold.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The AI chat was using both --resume (CLI session resumption) AND formatted
conversation history in the prompt simultaneously. This dual-context approach
confused the model — it saw the conversation twice (from session + from prompt
markup), leading to "I don't have context" responses on follow-ups.
Fix: remove --resume entirely from chat mode. Each CLI call is now independent,
with full conversation history formatted into the prompt via
<conversation_history> markup. trimHistory handles graceful truncation.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add trimHistory() to keep most recent messages within 100k token budget
- Add formatMessageWithHistory() to prepend conversation context to each message
- Wire up in useAIChat.ts: sendMessage now includes full history
- Keep --resume as belt-and-suspenders for same-session continuity
- 14 new tests in ai-chat.test.ts and useAIChat.test.ts
* feat: add claude CLI subprocess backend for AI panels
Add claude_cli.rs module that spawns the local `claude` CLI as a
subprocess instead of calling the Anthropic API directly. This removes
the need for users to configure an API key — the CLI uses the user's
existing Claude authentication.
- find_claude_binary(): discovers claude in PATH or common locations
- check_cli(): returns install/version status for the frontend
- run_chat_stream(): spawns claude -p with stream-json for chat panel
- run_agent_stream(): spawns claude with MCP vault tools for agent panel
- Parses stream-json NDJSON events and emits typed Tauri events
- Remove http:default capability (no longer calling APIs from frontend)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* feat: replace Anthropic API with Claude CLI for AI chat and agent panels
Remove direct Anthropic API calls and the entire tool-use loop from the frontend.
Both AI Chat and AI Agent panels now delegate to the `claude` CLI subprocess via
Tauri commands, streaming NDJSON events back to the UI.
Key changes:
- ai-chat.ts / ai-agent.ts: replace SSE/API streaming with Tauri invoke + listen
- useAIChat / useAiAgent hooks: simplified to use CLI streaming callbacks
- AIChatPanel / AiPanel: remove API key dialogs, model selectors, undo support
- SettingsPanel: remove Anthropic key field (no longer needed)
- claude_cli.rs: add comprehensive tests (37 total, 90% coverage)
- mock-handlers: replace ai_chat with check_claude_cli / stream_claude_* stubs
- Net -432 lines across 17 files
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* fix: add --verbose flag to claude CLI subprocess args
* chore: exclude search.rs and lib.rs from coverage (qmd integration + Tauri setup)
---------
Co-authored-by: Test <test@test.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
- Wrap ref assignments in useEffect to fix refs-during-render in
useTabManagement, useKeyboardNavigation, and Editor
- Rewrite useAppKeyboard to define keyMap inside useEffect, eliminating
ref access during render
- Add eslint-disable for react-hooks/set-state-in-effect on legitimate
dialog reset patterns (CommitDialog, CreateNoteDialog, CreateTypeDialog,
QuickOpenPalette, AIChatPanel, useVaultLoader)
- Add eslint-disable for react-hooks/static-components on icon lookups
(NoteItem, NoteList PinnedCard) — stateless icon components from a
static map
- Add eslint-disable for react-hooks/purity on Date.now() in
NoteItem TrashDateLine — intentionally memoized on trashedAt
- Remove unused `model` param from buildSystemPrompt (ai-chat.ts) and
update callers
- Fix exhaustive-deps: suppress vault object dep in App.tsx git history
effect (vault object is unstable, loadGitHistory is the real dep)
- Remove unused `modifiedFiles` from useNoteListData params and caller
- Add missing `ref` to Sidebar useOutsideClick deps
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Rewrite AIChatPanel with:
- Context picker: current note auto-added, search to add more notes,
removable context pills, token estimation badge
- Real Anthropic API integration via Vite dev server proxy (SSE streaming)
- API key stored in localStorage, configurable via dialog
- Model selector: Haiku 3.5, Sonnet 4, Opus 4
- Mock fallback when no API key is configured
- Extracted useAIChat hook and ai-chat utilities for clean separation
Decision: Using Vite middleware proxy for dev mode to avoid CORS issues
with direct browser-to-Anthropic calls. Phase 2 will add Rust backend.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
