FinAlly - AI Trading Workstation

A Bloomberg-style trading terminal with live streaming prices, a simulated $10k portfolio, and an LLM chat assistant that can analyze positions and execute trades from natural language. Built by coding agents as a capstone for an agentic AI coding course.

Status

• Done: market data (GBM simulator, Massive API client, shared price cache) in backend/app/market/. See planning/MARKET_DATA_SUMMARY.md.
• Planned: portfolio and watchlist API, SSE endpoint, frontend, chat, Docker, E2E tests. See planning/PLAN.md for the spec and build order.

Architecture

Single container on port 8000:

• Frontend: Next.js (static export), TypeScript, Tailwind, Recharts
• Backend: FastAPI (Python, uv) with SSE price streaming
• Database: SQLite, created and seeded at startup
• AI: Anthropic Python SDK with structured outputs
• Market data: built-in simulator by default, Massive API if a key is set

Configuration

Copy .env.example to .env:

Variable	Description
ANTHROPIC_AUTH_TOKEN, ANTHROPIC_BASE_URL, ANTHROPIC_MODEL	LLM access; required for chat unless LLM_MOCK=true
MASSIVE_API_KEY	Optional; empty uses the simulator
LLM_MOCK	Default false; true gives deterministic mock chat replies

Run

Available now (market data only):

cd backend
uv sync --extra dev
uv run pytest
uv run market_data_demo.py

Planned, once the app is built:

scripts/start_mac.sh     # or scripts\start_windows.ps1
# flags: --build, --open (PowerShell: -Build, -Open)
# open http://localhost:8000

Layout

backend/    FastAPI uv project (market data done)
planning/   Plan and agent docs
frontend/   Next.js project (planned)
test/       Playwright E2E tests (planned)
scripts/    Docker start/stop wrappers (planned)

License

MIT. See LICENSE.