Docker Deployment¶

Run ClawBio in a container for reproducible, isolated deployments. This guide covers Dockerfile patterns, docker-compose for multi-skill setups, and GPU passthrough.

Dockerfile Pattern¶

A typical Dockerfile for ClawBio:

FROM python:3.11-slim

WORKDIR /app

# Install system dependencies
RUN apt-get update && apt-get install -y --no-install-recommends \
    git curl && \
    rm -rf /var/lib/apt/lists/*

# Copy requirements and install Python deps
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

# Copy the full ClawBio repo
COPY . .

# Default: run the bot
CMD ["python3", "bot/roboterri.py"]

Note

This Dockerfile is an example pattern — it is not committed to the ClawBio repository. Adapt it to your needs.

docker-compose for Multi-Skill Setups¶

Run ClawBio alongside supporting services:

version: "3.8"

services:
  clawbio:
    build: .
    env_file: .env
    volumes:
      - ./data:/app/data          # Genomic input data
      - ./results:/app/results    # Skill output
    ports:
      - "8080:8080"
    restart: unless-stopped

  # Optional: local LLM with Ollama
  ollama:
    image: ollama/ollama
    volumes:
      - ollama-models:/root/.ollama
    ports:
      - "11434:11434"

volumes:
  ollama-models:

Volume Mounts¶

Mount	Purpose
`./data:/app/data`	Input genomic files (VCF, 23andMe .txt, FASTQ)
`./results:/app/results`	Skill output (reports, figures, JSON)
`./skills:/app/skills`	Custom skills (mount for live development)

Environment Variables¶

Pass your .env file or set variables directly:

docker run --env-file .env \
  -v $(pwd)/data:/app/data \
  -v $(pwd)/results:/app/results \
  clawbio

Key environment variables:

Variable	Description	Example
`LLM_API_KEY`	API key for your LLM provider	`sk-...`
`LLM_BASE_URL`	OpenAI-compatible endpoint	`https://api.openai.com/v1`
`CLAWBIO_MODEL`	Model to use	`gpt-4o`
`TELEGRAM_BOT_TOKEN`	Telegram bot token	From @BotFather
`RATE_LIMIT_PER_HOUR`	Per-user rate limit	`10`

GPU Passthrough¶

For ML-heavy skills (scRNA embedding, structure prediction), pass through your GPU:

NVIDIA¶

docker run --gpus all --env-file .env clawbio

Requires NVIDIA Container Toolkit installed on the host.

docker-compose with GPU¶

services:
  clawbio:
    build: .
    env_file: .env
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [gpu]

Building and Running¶

# Build
docker build -t clawbio .

# Run interactively
docker run -it --env-file .env clawbio

# Run a specific skill
docker run --env-file .env clawbio \
  python skills/pharmgx-reporter/pharmgx_reporter.py \
  --input data/patient.txt --output results/

# Run in background
docker compose up -d

Health Checks¶

Add a health check to your Dockerfile:

HEALTHCHECK --interval=30s --timeout=10s --retries=3 \
  CMD python -c "import clawbio; print('ok')" || exit 1

Next: Railway — deploy to the cloud with one click.