Docker Deployment - Gnosis Prediction Market Agent

Overview

Docker deployment packages the agent and all its dependencies into a container image, ensuring consistent behavior across different environments. This is the recommended approach for production deployments.

Dockerfile Architecture

The project uses a multi-stage Docker build for optimal image size and build caching:

Build Strategy

Multi-Stage Build

The Dockerfile uses two stages:

Builder Stage

Installs Poetry and creates the virtual environment with all dependencies:

FROM --platform=linux/amd64 python:3.11.9-bookworm AS builder

RUN pip install poetry==1.8.2

ENV POETRY_NO_INTERACTION=1 \
    POETRY_VIRTUALENVS_IN_PROJECT=1 \
    POETRY_VIRTUALENVS_CREATE=1 \
    POETRY_CACHE_DIR=/tmp/poetry_cache

WORKDIR /app

COPY pyproject.toml poetry.lock ./

RUN --mount=type=cache,target=$POETRY_CACHE_DIR poetry install --no-root

Runtime Stage

Creates a minimal runtime image with only the virtual environment and source code:

FROM --platform=linux/amd64 python:3.11.9-bookworm AS runtime

RUN apt-get update && apt-get install -y ffmpeg libsm6 libxext6

ENV VIRTUAL_ENV=/app/.venv \
    PATH="/app/.venv/bin:$PATH"

WORKDIR /app

COPY --from=builder ${VIRTUAL_ENV} ${VIRTUAL_ENV}

COPY pyproject.toml poetry.lock ./
COPY prediction_market_agent prediction_market_agent
COPY scripts scripts
COPY tests tests
COPY tokenizers tokenizers

Building the Image

Basic Build

Build the Docker image from the project root:

docker build -t gnosis-prediction-agent .

Build with Cache

The Dockerfile uses Docker’s cache mount for faster builds:

docker build \
  --cache-from gnosis-prediction-agent:latest \
  -t gnosis-prediction-agent:latest \
  .

Build with Version Tag

Tag your builds with version numbers for better tracking:

docker build \
  --build-arg LANGFUSE_DEPLOYMENT_VERSION=$(git rev-parse HEAD) \
  -t gnosis-prediction-agent:v1.0.0 \
  .

Running Containers

Basic Run

Run a container with environment variables:

docker run \
  -e BET_FROM_PRIVATE_KEY="your_private_key" \
  -e OPENAI_API_KEY="your_openai_key" \
  -e runnable_agent_name="prophet_gpt4o" \
  -e market_type="omen" \
  gnosis-prediction-agent:latest

Run with Environment File

Use a .env file for cleaner configuration:

docker run --env-file .env gnosis-prediction-agent:latest

Never commit your .env file to version control. Keep it in .gitignore.

Interactive Mode

Run a container with an interactive shell for debugging:

docker run -it \
  --env-file .env \
  gnosis-prediction-agent:latest \
  bash

Override Command

Override the default command to run a specific agent:

docker run \
  --env-file .env \
  gnosis-prediction-agent:latest \
  python prediction_market_agent/run_agent.py coinflip omen

Environment Configuration

Required Variables

The container CMD uses environment variables to specify the agent:

CMD ["bash", "-c", "python prediction_market_agent/run_agent.py ${runnable_agent_name} ${market_type}"]

docker run \
  -e runnable_agent_name="prophet_gpt4o" \
  -e market_type="omen" \
  --env-file .env \
  gnosis-prediction-agent:latest

version: '3.8'
services:
  agent:
    image: gnosis-prediction-agent:latest
    environment:
      - runnable_agent_name=prophet_gpt4o
      - market_type=omen
    env_file:
      - .env

System Environment Variables

The Docker image sets several system-level environment variables:

string

default:"/app"

Python module search path, set to the application directory

string

default:"python"

Use pure Python implementation of Protocol Buffers

string

default:"1"

Disable transformers warnings (we only use for tokenization, not PyTorch)

string

default:"none"

Deployment version for Langfuse tracing (set via build arg in CI/CD)

System Dependencies

The runtime image includes system packages required by various Python libraries:

RUN apt-get update && apt-get install -y ffmpeg libsm6 libxext6

ffmpeg - Media processing (used by some AI models)
libsm6 - Session management library
libxext6 - X11 extensions library

Docker Compose

Single Agent Setup

Create a docker-compose.yml for easier management:

docker-compose.yml

version: '3.8'

services:
  prophet-agent:
    build: .
    image: gnosis-prediction-agent:latest
    environment:
      - runnable_agent_name=prophet_gpt4o
      - market_type=omen
    env_file:
      - .env
    restart: unless-stopped
    logging:
      driver: "json-file"
      options:
        max-size: "10m"
        max-file: "3"

Run with:

docker-compose up -d

Multiple Agents

Run multiple agents simultaneously:

docker-compose.yml

version: '3.8'

services:
  prophet-gpt4o:
    build: .
    image: gnosis-prediction-agent:latest
    environment:
      - runnable_agent_name=prophet_gpt4o
      - market_type=omen
    env_file:
      - .env
    restart: unless-stopped

  microchain:
    build: .
    image: gnosis-prediction-agent:latest
    environment:
      - runnable_agent_name=microchain
      - market_type=omen
    env_file:
      - .env
    restart: unless-stopped

  social-media:
    build: .
    image: gnosis-prediction-agent:latest
    environment:
      - runnable_agent_name=social_media
      - market_type=omen
    env_file:
      - .env
    restart: unless-stopped

With Database

Add a PostgreSQL database for agents that need persistence:

docker-compose.yml

version: '3.8'

services:
  database:
    image: postgres:15
    environment:
      - POSTGRES_DB=prediction_agent
      - POSTGRES_USER=agent
      - POSTGRES_PASSWORD=secure_password
    volumes:
      - postgres_data:/var/lib/postgresql/data
    restart: unless-stopped

  agent:
    build: .
    image: gnosis-prediction-agent:latest
    environment:
      - runnable_agent_name=prophet_gpt4o
      - market_type=omen
      - SQLALCHEMY_DB_URL=postgresql://agent:secure_password@database:5432/prediction_agent
    env_file:
      - .env
    depends_on:
      - database
    restart: unless-stopped

volumes:
  postgres_data:

CI/CD Integration

GitHub Actions

The project includes automated Docker builds in .github/workflows/python_cd.yaml:

name: Python CD

on:
  pull_request:
  push:
    branches: [main]

env:
  REGISTRY: ghcr.io
  IMAGE_NAME: ${{ github.repository }}

jobs:
  build-and-push-image:
    if: (contains(github.event.pull_request.body, 'build please') && github.event_name == 'pull_request') || (github.event_name == 'push' && github.ref == 'refs/heads/main')
    runs-on: ubuntu-latest
    permissions:
      contents: read
      packages: write
    steps:
      - name: Checkout Repository
        uses: actions/checkout@v3

      - name: Log in to the Container registry
        uses: docker/login-action@65b78e6e13532edd9afa3aa52ac7964289d1a9c1
        with:
          registry: ${{ env.REGISTRY }}
          username: ${{ github.actor }}
          password: ${{ secrets.GITHUB_TOKEN }}

      - name: Extract metadata (tags, labels) for Docker
        id: meta
        uses: docker/metadata-action@9ec57ed1fcdbf14dcef7dfbe97b2010124a938b7
        with:
          images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}

      - name: Build and push Docker image
        uses: docker/build-push-action@4a13e500e55cf31b7a5d59a38ab2040ab0f42f56
        with:
          push: true
          tags: ${{ steps.meta.outputs.tags }}
          labels: ${{ steps.meta.outputs.labels }}
          build-args: |
            LANGFUSE_DEPLOYMENT_VERSION=${{ github.sha }}

Image Registry

Images are automatically pushed to GitHub Container Registry (ghcr.io):

# Pull the latest image
docker pull ghcr.io/gnosis/prediction-market-agent:main

# Run the image
docker run --env-file .env ghcr.io/gnosis/prediction-market-agent:main

Trigger Builds

Automatic on Main

Pushes to the main branch automatically trigger builds

Manual PR Builds

Add “build please” to PR description to trigger a build:

## Changes
- Updated agent logic
- Fixed bug in market parsing

build please

Image Optimization

Size Reduction

The multi-stage build reduces image size significantly:

Without Multi-Stage

~2.5 GB (includes Poetry and build tools)

With Multi-Stage

~1.2 GB (runtime dependencies only)

Layer Caching

Optimize build times by ordering Dockerfile commands strategically:

Install system dependencies (rarely changes)
Copy pyproject.toml and poetry.lock (changes occasionally)
Install Python dependencies (cached until lockfile changes)
Copy source code (changes frequently)

Build Cache Mount

The Poetry cache is mounted during build to avoid re-downloading packages:

RUN --mount=type=cache,target=$POETRY_CACHE_DIR poetry install --no-root

Monitoring and Logs

View Logs

# Follow logs
docker logs -f <container_id>

# Last 100 lines
docker logs --tail 100 <container_id>

# With timestamps
docker logs -t <container_id>

Container Stats

# Real-time stats
docker stats <container_id>

# All containers
docker stats

Health Checks

Add a health check to your Docker Compose:

services:
  agent:
    build: .
    healthcheck:
      test: ["CMD", "python", "-c", "import sys; sys.exit(0)"]
      interval: 30s
      timeout: 10s
      retries: 3
      start_period: 40s

Troubleshooting

Build Failures

If the build fails, try clearing the build cache:

docker builder prune
docker build --no-cache -t gnosis-prediction-agent .

Platform Issues

The Dockerfile specifies linux/amd64. If you’re on ARM (M1/M2 Mac), you may need:

docker build --platform linux/amd64 -t gnosis-prediction-agent .

Out of Memory

Increase Docker memory limits in Docker Desktop settings or add resource limits:

services:
  agent:
    build: .
    deploy:
      resources:
        limits:
          memory: 4G

Missing Environment Variables

Ensure all required variables are set. Check logs:

docker logs <container_id> 2>&1 | grep -i "error\|missing"

Next Steps

Cloud Deployment

Deploy containers to Google Kubernetes Engine (GKE)

Environment Config

Complete environment variable reference

Local Development

Run agents locally without Docker

Contributing

Contribute to the project on GitHub

​Overview

​Dockerfile Architecture

​Build Strategy

​Multi-Stage Build

​Building the Image

​Basic Build

​Build with Cache

​Build with Version Tag

​Running Containers

​Basic Run

​Run with Environment File

​Interactive Mode

​Override Command

​Environment Configuration

​Required Variables

​System Environment Variables

​System Dependencies

​Docker Compose

​Single Agent Setup

​Multiple Agents

​With Database

​CI/CD Integration

​GitHub Actions

​Image Registry

​Trigger Builds

​Image Optimization

​Size Reduction

Without Multi-Stage

With Multi-Stage

​Layer Caching

​Build Cache Mount

​Monitoring and Logs

​View Logs

​Container Stats

​Health Checks

​Troubleshooting

​Next Steps

Cloud Deployment

Environment Config

Local Development

Contributing

Overview

Dockerfile Architecture

Build Strategy

Multi-Stage Build

Building the Image

Basic Build

Build with Cache

Build with Version Tag

Running Containers

Basic Run

Run with Environment File

Interactive Mode

Override Command

Environment Configuration

Required Variables

System Environment Variables

System Dependencies

Docker Compose

Single Agent Setup

Multiple Agents

With Database

CI/CD Integration

GitHub Actions

Image Registry

Trigger Builds

Image Optimization

Size Reduction

Layer Caching

Build Cache Mount

Monitoring and Logs

View Logs

Container Stats

Health Checks

Troubleshooting

Next Steps