A conversational AI assistant for managing a simulated investment portfolio. Ask questions, get advice, and execute trades through natural language.
- Natural Language Trading: "Buy 20 shares of SPY" or "Invest $500 in gold"
- RAG-Powered Responses: Uses actual market data, not hallucinated prices
- MCP Tool Calling: Executes trades, checks portfolio, fetches prices
- Streaming Responses: Real-time token-by-token output
- Session Persistence: Chat history saved across sessions
- Multi-Asset Support: SPY, GLD, SLV, AGG, TLT, VNQ
# 1. Clone and configure
git clone <repository>
cp .env.example .env
# 2. Start all services
docker compose up --build
# 3. Open the app
# Frontend: http://localhost:3000
# API Docs: http://localhost:8081/docs| Service | Technology | Port |
|---|---|---|
| Frontend | React + Vite | 3000 |
| Inference API | FastAPI + SSE | 8081 |
| MCP Server | FastAPI + MCP | 8000 |
| RAG Pipeline | ChromaDB + Ollama | - |
| Database | PostgreSQL | 5432 |
| LLM | Ollama (Qwen2.5) | 11434 |
| Symbol | Description |
|---|---|
| SPY | S&P 500 ETF |
| GLD | Gold ETF |
| SLV | Silver ETF |
| AGG | Aggregate Bond ETF |
| TLT | 20+ Year Treasury Bond ETF |
| VNQ | Real Estate ETF |
We use historical 2024 data projected to 2025 for realistic price movements without requiring paid API subscriptions.
# Start services
docker compose up -d
# View logs
docker compose logs -f
# Stop services
docker compose down
# Rebuild after changes
docker compose up --build├── frontend/ # React UI
├── inference/ # Chat orchestrator + streaming API
├── mcp-server/ # Portfolio operations + MCP tools
├── rag-pipeline/ # Knowledge base + embeddings
└── shared/ # Common models and utilities
MIT