Worker Setup

The Marqov platform worker polls Supabase for pending jobs and executes them via Temporal and AWS Braket. It runs as a Docker container on AWS ECS Fargate.

Building the Docker Image

The Dockerfile is at the repository root. It uses python:3.12-slim and installs dependencies from platform/requirements-worker-full.txt.


docker build \
  --platform linux/amd64 \
  --provenance=false \
  --sbom=false \
  -t marqov-worker .

Required flags:

Flag	Reason
`--platform linux/amd64`	ECS Fargate requires amd64. Apple Silicon builds ARM by default.
`--provenance=false`	Prevents attestation manifest 403 errors on ECR push.
`--sbom=false`	Same as above.

The image exposes port 8080 and includes a health check that hits http://localhost:8080/health every 30 seconds.

Entrypoint: python platform/src/platform_worker.py

Pushing to ECR

ECR repository: 123456789012.dkr.ecr.us-east-1.amazonaws.com/marqov/platform-worker


# Login to ECR
aws ecr get-login-password --region us-east-1 | \
  docker login --username AWS --password-stdin \
  123456789012.dkr.ecr.us-east-1.amazonaws.com
 
# Tag the image
docker tag marqov-worker:latest \
  123456789012.dkr.ecr.us-east-1.amazonaws.com/marqov/platform-worker:latest
 
# Push
docker push \
  123456789012.dkr.ecr.us-east-1.amazonaws.com/marqov/platform-worker:latest

Deploying to ECS

Cluster: marqov-production Service: marqov-worker AWS Account: 123456789012 (your-account)


aws ecs update-service \
  --cluster marqov-production \
  --service marqov-worker \
  --force-new-deployment \
  --region us-east-1

The --force-new-deployment flag triggers a rolling update that pulls the latest image from ECR.

Health Check

The worker exposes a health check endpoint:

Endpoint: GET /health
Port: 8080
Success response: HTTP 200

The Dockerfile configures a Docker HEALTHCHECK:

Interval: 30s
Timeout: 10s
Start period: 5s
Retries: 3

ECS also uses this endpoint for service health monitoring.

Environment Variables

The worker requires the following environment variables (passed via ECS task definition or --env-file):

Variable	Required	Description
`NEXT_PUBLIC_SUPABASE_URL`	Yes	Supabase project URL
`SUPABASE_SERVICE_ROLE_KEY`	Yes	Supabase service role key
`TEMPORAL_ADDRESS`	Yes	Temporal server address
`TEMPORAL_NAMESPACE`	Yes	Temporal namespace
`TEMPORAL_API_KEY`	Production	Temporal Cloud API key
`AWS_REGION`	Yes	AWS region (default: `us-east-1`)
`AWS_ACCESS_KEY_ID`	Yes	AWS credentials
`AWS_SECRET_ACCESS_KEY`	Yes	AWS credentials
`BRAKET_S3_BUCKET`	Yes	S3 bucket for Braket results
`BRAKET_S3_PREFIX`	Yes	S3 prefix for results
`SENTRY_WORKERS_DSN`	No	Sentry DSN for error tracking

Local Development

For local development, run the worker directly:


cd /path/to/marqov-chewie
.venv/bin/python platform/src/platform_worker.py

Ensure you have a .env file or exported environment variables. The worker requires a running Temporal server (see Temporal Configuration).

Dockerfile Structure


FROM python:3.12-slim
WORKDIR /app

# 1. System dependencies (git, build-essential)
# 2. Install Python deps from requirements-worker-full.txt
# 3. Install marqov package (--no-deps to skip covalent)
# 4. Copy benchmarks/ and platform/src/
# 5. Set PYTHONPATH=/app
# 6. Expose 8080, configure healthcheck
# 7. CMD: python platform/src/platform_worker.py

The marqov package is installed into site-packages via pip install --no-deps --no-cache-dir . so imports resolve correctly without path hacks. The --no-deps flag is needed because requirements-worker-full.txt already includes the required dependencies, and a full install would attempt to install covalent (which is no longer needed).