AI Chat UI: streaming, markdown, code blocks, history, settings (v0.2.0)

[s1rry]

What

Milestone (v0.2.0): you can now chat with your repository. Sub-step 8b completes AI Chat.

• AI Chat view: repository-aware conversation with the configured provider.
  • Streaming replies token by token via a Tauri IPC Channel<String>.
  • Markdown rendering (react-markdown + remark-gfm) with syntax-highlighted code blocks (rehype-highlight + highlight.js).
  • Conversation history in a Zustand store, sent with each turn.
  • Choose a project folder to chat about; clear the conversation; "Thinking…" while the first token is pending.
• Provider settings screen: pick provider (Ollama, Claude, OpenAI, Gemini) and model; enter API keys (masked password field, persisted via the backend settings store; Ollama needs none).
• Typed lib/ipc wrappers: chat (Channel streaming) and settings (get/set).

All backend access flows through lib/ipc; new features/ai-chat and features/settings slices with their own stores.

Verification

• pnpm build (tsc strict + vite): clean, no any.
• Rendered live in the Vite preview: seeded a conversation showing a user bubble and a streamed assistant reply with bold, inline code, a bullet list and a highlighted Rust code block; and the settings screen with provider dropdown, model and a masked API key. No console errors. Screenshots captured.

Live streaming runs against the backend (chat command) in pnpm tauri dev.

This closes the AI Chat feature (8a backend, 8b UI) and the initial AI phase.

🤖 Generated with Claude Code