Docker & Deployment

The system is designed with a container-first architecture to ensure consistency between development, backtesting, and production environments. Using Docker, you can orchestrate the multi-agent trading system, the FastAPI backend, and the React-based dashboard with a single command.

Infrastructure Overview

The deployment stack consists of three primary components:

1. Trading Backend (FastAPI): Orchestrates the LangChain agents, manages the vector store (RAG), and handles data ingestion from NewsAPI and CoinGecko.
2. Dashboard Frontend (React/Vite): Provides the real-time visualization of equity curves, trade logs, and agent reasoning.
3. Persistent Storage: SQLite database for trade history and a local volume for the vector database (ChromaDB) indexing.

Prerequisites

Before deploying, ensure you have the following installed:

• Docker (20.10+)
• Docker Compose (v2.0+)
• An API key for your chosen LLM provider (Groq or a local Ollama instance).

Configuration

Environment variables are managed through a .env file in the root directory. Copy the template and fill in your credentials:

# Core API Keys
GROQ_API_KEY=your_groq_key_here
NEWSAPI_KEY=your_newsapi_key_here

# Backend Configuration
DATABASE_URL=sqlite:///./trading_system.db
LOG_LEVEL=info

# LLM Selection (options: groq, ollama)
MODEL_BACKEND=groq 
OLLAMA_BASE_URL=http://host.docker.internal:11434

Deployment Steps

1. Build and Launch

To start the entire ecosystem, run:

docker-compose up --build

This command builds the frontend and backend images, links them via a virtual network, and exposes the following ports:

• Frontend Dashboard: http://localhost:5173
• Backend API (Swagger Docs): http://localhost:8000/docs

2. Initializing the Vector Store (RAG)

The system requires a knowledge base to inform agent decisions. Once the containers are running, you must index the trading documents:

docker-compose exec backend python -m agents.rag.indexer

3. Running a Backtest in Docker

You can trigger backtests within the containerized environment via the CLI. This will populate the dashboard with data:

docker-compose exec backend python run_backtest.py --days 30 --strategy multi-agent

Docker Compose Structure

The system uses a standard docker-compose.yml to manage service dependencies. Below is the service definition for the production-ready stack:

services:
  backend:
    build: 
      context: .
      dockerfile: backend.Dockerfile
    ports:
      - "8000:8000"
    volumes:
      - ./logs:/app/logs
      - ./knowledge:/app/knowledge
      - ./data:/app/data
    env_file: .env

  frontend:
    build:
      context: ./dashboard/frontend
      dockerfile: Dockerfile
    ports:
      - "5173:5173"
    environment:
      - VITE_API_BASE=http://localhost:8000
    depends_on:
      - backend

Production Considerations

Persistence

The system uses Docker volumes to ensure that trade history and backtest reports persist across container restarts. Ensure the ./logs and ./data folders on your host machine have appropriate write permissions.

Local LLM Connectivity

If using Ollama for local inference:

• On Linux: Ensure the backend container can reach http://localhost:11434.
• On macOS/Windows: Use http://host.docker.internal:11434 in your .env to allow the container to communicate with the host machine's Ollama service.

Resource Allocation

The Multi-Agent Orchestrator can be resource-intensive when running parallel "debate" phases between agents. It is recommended to allocate at least 4GB of RAM to the Docker engine for stable performance.