Multi-Model AI Application with Human-in-the-Loop Workflows

Chamber is a professional AI application that enables multi-model reasoning with human-in-the-loop workflows. Multiple LLM models collaborate on complex tasks, with one acting as orchestrator to synthesize decisions and advance progress.

• Frontend: Tauri 2.0 + SvelteKit (SPA) + TailwindCSS
• Backend: Rust (Tauri commands) + Python sidecar (LangGraph + FastAPI)
• Storage: IndexedDB (Dexie) + Markdown files (configurable workspace)
• LLM Providers: Anthropic, Google Gemini, Ollama, x.ai

• ✨ Multi-step reasoning with ReAct pattern
• 🤝 Multi-model chamber (parallel reasoning → orchestrator synthesis)
• 👤 Human-in-the-loop approval for tool usage
• ⏸️ Pause/resume capability with state persistence
• ⚙️ Configuration-driven (no code changes needed)
• 🔐 Secure OAuth authentication - Sign in with Claude/Google accounts instead of API keys
• 🔑 Keychain storage - Credentials stored securely in platform keychain

Chamber supports multiple authentication methods for LLM providers:

Supported Providers:

• Anthropic Claude (Sign in with Claude account)
• Google Gemini (Sign in with Google account)

Benefits:

• No need to manage API keys manually
• Credentials stored securely in your system keychain
• Automatic token refresh in the background
• No plain text credentials in config files

How to Use:

1. Open Chamber and go to Settings
2. Find the Authentication section
3. Click "Connect with OAuth" for your provider
4. Complete the sign-in flow in the authorization window
5. Your credentials are now stored securely!

You can still use traditional API keys if you prefer:

In Settings:

1. Click "Use API Key" button for your provider
2. Paste your API key
3. Click Save

Via Environment Variables: Create a .env file in the python-sidecar directory:

• Node.js 18+
• Rust 1.70+
• Python 3.10+
• Tauri CLI

# Install dependencies
npm install

# Install Python dependencies (using uv)
cd python-sidecar
uv sync
cd ..

# Setup environment variables
cp .env.example .env
# Add your API keys to .env

Create a .env file in the python-sidecar directory:

ANTHROPIC_API_KEY=your_key_here
GOOGLE_API_KEY=your_key_here
XAI_API_KEY=your_key_here
OLLAMA_BASE_URL=http://localhost:11434

This project uses git worktrees for parallel development, allowing you to run long-lived dev servers while developing features.

Current worktree structure:

D:/Projects/chamber     [main]  ← Production-ready code
D:/Projects/chamber-dev [dev]   ← Active development

Basic workflow:

# Terminal 1: Main worktree - Run dev servers (keeps running!)
cd chamber/
npm run tauri:dev

# Terminal 2: Dev worktree - Develop features
cd ../chamber-dev/
# Make changes, commit, test without stopping the dev server

Create additional worktrees:

# For experiments, PRs, or parallel work
git worktree add ../chamber-experiment -b experiment-branch
git worktree add ../chamber-staging staging

# See all worktrees
git worktree list

# Remove when done
git worktree remove ../chamber-experiment

For detailed git workflows, see GIT_WORKFLOWS.md.

• PKCE Flow: Uses Proof Key for Code Exchange (PKCE) for secure OAuth authorization
• Platform Keychain: Credentials stored using OS keychain services:
  • Windows: Credential Manager
  • macOS: Keychain Access
  • Linux: Secret Service API (libsecret)
• TLS Only: All OAuth communication over HTTPS
• Token Refresh: Automatic background refresh before token expiry
• Minimal Scopes: Only requests necessary OAuth scopes

• OAuth tokens are automatically refreshed 5 minutes before expiry
• Background refresh task runs every 5 minutes
• No manual token management required
• Tokens are never written to disk in plain text

If using API keys instead of OAuth:

• Store in .env file (already gitignored)
• Never commit .env to version control
• Consider using environment variable injection in production

Authorization page won't load:

• Check your internet connection
• Disable VPN temporarily
• Ensure popups are allowed for Chamber
• Try a different browser

Authorization fails:

• Make sure you're allowing all requested permissions
• Check that your account has access to the provider's API
• Try clearing your browser cache and cookies

Token refresh errors:

• Sign out and sign in again
• Check that your system keychain is accessible
• Verify your account is still active with the provider

Chamber's keychain storage is compatible with:

• Windows 10/11: Uses Windows Credential Manager
• macOS 12+: Uses Keychain Access
• Linux: Uses libsecret (gnome-keyring/kwallet)

For detailed platform-specific testing and verification procedures, see CROSS_PLATFORM_KEYRING.md.

Run the application in development mode:

# Terminal 1: Start SvelteKit dev server
npm run dev

# Terminal 2: Start Tauri (includes Rust hot reload)
npm run tauri:dev

# Terminal 3 (optional): Run Python sidecar standalone for testing
cd python-sidecar
uv run python -m chamber.main --host 127.0.0.1 --port 8765

Configuration is stored in workspace/config/chamber-config.yaml. Edit this file to:

• Change orchestrator and reasoning models
• Enable/disable specific models
• Configure tool approval settings
• Adjust workspace paths

User Input (SvelteKit UI)
  ↓
Svelte Stores + SessionManager
  ↓
Tauri Commands (Rust)
  ↓
Python Sidecar (FastAPI + LangGraph)
  ↓
Orchestrator → Parallel Reasoning → Synthesis → Tools
  ↓
WebSocket Stream ← Real-time Updates
  ↓
UI Updates (Reactive Svelte)

# Frontend tests
npm run test

# Rust tests
cd src-tauri
cargo test

# Python tests
cd python-sidecar
uv run pytest

# Build for production
npm run build
npm run tauri:build

This creates platform-specific binaries in src-tauri/target/release/bundle/.

chamber/
├── src/                  # SvelteKit frontend
│   ├── lib/
│   │   ├── components/  # UI components
│   │   ├── stores/      # Svelte stores
│   │   ├── services/    # Service layer
│   │   ├── types/       # TypeScript types
│   │   └── db/          # Dexie schema
│   └── routes/          # SvelteKit pages
├── src-tauri/           # Rust backend
│   └── src/
│       ├── commands/    # Tauri commands
│       ├── services/    # Business logic
│       └── models/      # Data structures
├── python-sidecar/      # Python LangGraph backend
│   └── chamber/
│       ├── graph/       # LangGraph workflows
│       ├── models/      # LLM providers
│       ├── tools/       # ReAct tools
│       └── server/      # FastAPI + WebSocket
└── workspace/           # User workspace
    ├── sessions/        # Session MD files
    └── config/          # Configuration

MIT

Contributions welcome! Please read CONTRIBUTING.md for guidelines.