ANPS-TradeMeUp

ANPS-TradeMeUp is a multi-agent news-to-market prediction system. It ingests financial news, extracts entities and events with LLM-supported analysis, scores impact and surprise, and produces market predictions with a Dash GUI and a FastAPI backend.

Screenshots

Dashboard

Live Charts

News Feed

Predictions

Trading Simulations

Statistics

Status

• Core application code is organized under src/
• Operational scripts are grouped under scripts/
• Legacy SQLite migration utilities are archived under archive/legacy/sqlite/
• The repository is now prepared for full-container workflows with postgres, api, gui, and worker services in docker-compose.yml

Quick Start

Full Docker stack

docker compose up --build

Rebuild only affected services after file changes:

./scripts/docker/rebuild-on-change.sh
./scripts/docker/rebuild-on-change.sh --watch

The script rebuilds only gui for pure GUI changes; for changes across multiple areas or shared build files, the full app stack (api, gui, worker) is rebuilt.

Services:

• GUI: http://localhost:8050
• API docs: http://localhost:8000/docs
• API health: http://localhost:8000/health
• PostgreSQL: localhost:5432

This default stack starts:

• postgres
• api
• gui

The optional continuous pipeline worker is not started by default.

Start it explicitly:

docker compose --profile worker up -d worker

For older docker-compose v1 environments, a compatible command is:

COMPOSE_PROFILES=worker docker-compose up -d --no-recreate worker

One-shot pipeline in Docker with OpenAI

Set OPENAI_API_KEY in .env.local, then run:

docker compose exec -T \
  -e LLM_PROVIDER=openai \
  -e OPENAI_MODEL=gpt-4o-mini \
  api python scripts/run_mvp_pipeline.py

If you use a custom OpenAI-compatible proxy, set OPENAI_BASE_URL to a full https://... URL. If it is empty or invalid, the app now falls back to the official OpenAI endpoint automatically.

Local development with Dockerized PostgreSQL

cp .env.example .env.local
docker compose up -d postgres

python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

alembic upgrade head
python scripts/runtime/run_dashboard.py

Windows:

python -m venv venv
.\venv\Scripts\activate
pip install -r requirements.txt
alembic upgrade head
python scripts\runtime\run_dashboard.py

Common Entrypoints

• Dashboard: python scripts/runtime/run_dashboard.py
• API: uvicorn src.api.main:app --host 0.0.0.0 --port 8000 --reload
• Continuous pipeline worker: python scripts/run_continuous_pipeline.py
• One-shot MVP pipeline: python scripts/run_mvp_pipeline.py
• Database connection check: python scripts/db/test_db_connection.py
• Database bootstrap helper: python scripts/db/init_database.py

Convenience wrappers remain available under scripts/runtime/ and scripts/setup/.

Repository Layout

.
├── src/                  # Application code
├── scripts/              # Runtime, setup, db, and utility scripts
├── tests/                # Test suite and validation checks
├── docs/                 # Setup, architecture, features, and reports
├── migrations/           # Alembic environment and SQL migrations
├── archive/              # Legacy and diagnostic assets kept for reference
├── docker-compose.yml    # Full local container stack
├── Dockerfile            # Shared application image
├── pyproject.toml        # Python project metadata and tooling
├── requirements.txt      # Pip-compatible dependency list
├── alembic.ini           # Alembic configuration
└── .env.example          # Environment template

Detailed folder notes: docs/STRUCTURE.md

Documentation

• Quickstart: docs/setup/QUICKSTART.md
• PostgreSQL setup: docs/setup/POSTGRESQL_SETUP.md
• Repository structure: docs/STRUCTURE.md
• Architecture snapshot: docs/architecture/1.0.3-TradeMeUp_Complete_System_Architecture.md

Notes

• src/config/settings.py loads environment values from .env.local
• The Docker stack does not require .env.local for the default local boot path
• OPENAI_MODEL defaults to gpt-4o-mini
• The archived SQLite migration scripts are intentionally no longer part of the default flow
• Historical reports under docs/reports/ may still describe older paths or workflows