Atlas - AI Trading & Stock Analysis

Algorithmic trading and stock analysis platform powered by AI.

Running the Project

1. Run the Backend

Navigate to the backend folder, set up your Python virtual environment and run the Flask server:

# In the backend directory
.\.venv\Scripts\python.exe wsgi.py

2. Run the Frontend

Navigate to the frontend directory, install packages, and start the development server:

# In the frontend directory
npm install --legacy-peer-deps
npm run dev

3. Run the Market News Pipeline

The Market News warning cards and the news-driven stock predictions come from a three-stage Docker pipeline: news-streamer fetches articles, warning-streamer turns them into warnings, and prediction-streamer turns warnings into stock predictions. Redpanda carries messages between stages and MongoDB stores the results. With Docker Desktop running and your API keys in backend/.env:

docker compose --profile docker-news -f docker-compose.news.yml up --build -d

See docs/news-pipeline.md for setup, logs, one-off runs, clearing old data, rebuilding, and stopping.

Testing

Frontend, from the project root:

node --test frontend/tests/*.test.mjs
npm --prefix frontend run build
npm --prefix frontend run lint

Lint has pre-existing errors, so treat it as informational and avoid adding new ones in files you change.

Backend, from backend/, with the virtual environment's Python (.venv\Scripts\python.exe on Windows):

.venv/bin/python -m unittest discover -s tests

The tests run offline: they mock NewsData, NVIDIA, Redpanda and MongoDB.

AI Development Workflow

.ai-workflow/ runs a local Claude plan → Codex code → Claude review pipeline in an isolated Git worktree. Use --bug for bug fixes: Codex first reproduces the bug with a failing test and reviews the code to find the root cause, then Claude plans the fix. See .ai-workflow/SETUP.md.

python3 .ai-workflow/pipeline.py "Feature task with acceptance criteria"
python3 .ai-workflow/pipeline.py --bug "What happened, what you expected, and any error text"