You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
5 AI Providers — Mistral, OpenAI, Anthropic, Gemini, and Ollama (local). Stream completions from any provider with hot-swappable model selection
Semantic Search — Documents are chunked and embedded at upload time; queries are embedded at search time with vector cosine similarity, falling back to TF-IDF
Ollama Auto-Detect — Automatically detects running Ollama instances on mount, dynamically fetches available models from /api/tags
Multi-Provider Settings UI — Per-provider API key sections (Mistral, OpenAI, Anthropic, Gemini), Ollama URL + Connect button with live status indicator
Provider-Grouped Model Selector — Colored dots in TopBar visually distinguish provider groups in the model dropdown
Quality & Reliability
191 Automated Tests — Full test suite covering stores (uiStore, ragStore, canvasStore), API config/providers, utilities, components, and hooks. All passing with Vitest + jsdom
TypeScript Strictness — Replaced all any types with proper types (Message, inline API response types, catch clauses with instanceof Error)
Bug Fixes — 6 bugs fixed during stress testing including missing finalize call in embed router, setContent typo in ProviderModelSelect, missing tools imports in providers, canvas.width typo in useOllamaDetect
CSP Hardening — Content Security Policy enforced at the Tauri level; meta tag removed from index.html
Cross-Platform CI/CD
GitHub Actions CI — Runs tsc --noEmit, vitest, and vite build on push/PR to main
Cross-Platform Build Matrix — Automatic MSI/EXE (Windows), DMG (macOS), and AppImage (Linux) builds on tag push
Download Links — README now lists platform-specific installer links