This project simulates Google Gemini Gems capabilities, providing a flexible interface for managing and executing custom AI instructions ("Gems"). It is designed to run as a containerized service, leveraging the latest Gemini 3.0 models with low-latency thinking capabilities.

Since the standard Gemini Gems interface does not currently support direct API access, this project emulates that functionality by:

1. Storage: Storing Gem definitions (system instructions, prompts) in a Supabase database.
2. Execution: Providing a FastAPI backend service that reads Gem definitions from Supabase and combines them with user input to execute requests via the Google Gemini API (using the latest google-genai SDK).
3. Management: Offering a React-based web interface for creating, editing, and testing Gems.
4. Optimization: Automatically handling Gemini 3.0 "thinking" artifacts and optimizing for low latency.

• Frontend: React (Vite) - Served statically by FastAPI in production
• Backend: FastAPI (Python)
• Database: Supabase (PostgreSQL)
• AI Model: Google Gemini (Supports 1.5, 2.0, and 3.0 series)

• Initial Project Setup
• Dockerize the application (Multi-stage build)
• Configure Nginx Proxy Manager for secure deployment
• Setup private networking (Docker shared_net) to avoid exposing ports on localhost

• POST /api/gems/execute: Execute a specific Gem by name.
  • Body: {"gem_name": "Name", "user_prompt": "..."}
• POST /api/gemini/generate: Direct access to the configured Gemini model.
  • Body: {"system_instructions": "...", "user_prompt": "..."}
• GET /api/gems: List all available Gems.

• Docker & Docker Compose (Recommended)
• Supabase Account
• Google Gemini API Key

• Keep SUPABASE_KEY and API_TOKEN server-only; never add them to VITE_* variables or frontend code.
• The backend fails to start if a service key is exposed to the frontend or bundled into dist.
• Before committing, run npm run secret:check to scan for leaked tokens in assets, source, and docs.

Create a .env file with the following:

GEMINI_API_KEY=your_gemini_key
# Model to use (defaults to gemini-3-pro-preview if unset)
GEMINI_MODEL=gemini-3-pro-preview
SUPABASE_URL=your_supabase_url
SUPABASE_KEY=your_supabase_service_role_key
PORT=8000
NODE_ENV=production

This project is designed to be run via Docker Compose.

1. Build and Start:

   docker compose up -d --build

2. Access:

   • The service runs on port 8000 inside the container.
   • It connects to the external shared_net network.
   • Configure your reverse proxy (e.g., Nginx Proxy Manager) to point to the container name gemsapi on port 8000.

If you wish to run without Docker:

1. Install Node dependencies:

   npm install

2. Install Python dependencies:

   pip install fastapi uvicorn google-genai supabase python-dotenv slowapi pydantic httpx aiofiles

3. Run Frontend:

   npm run dev

4. Run Backend:

   uvicorn fastapi_server:app --reload