Deployment

M3 Forge provides multiple deployment options to suit various environments and use cases.

All deployment options require external PostgreSQL and S3-compatible storage services. When deployed alongside Marie-AI, these can be shared from the Marie-AI infrastructure stack.

Deployment Options

All-in-One

Single Docker container with everything bundled — simplest path to a running instance

Docker Compose

Multi-container deployment with individual scaling and production profiles

Gitea Integration

Self-hosted Git repository management for workflows and configurations

Comparison

	All-in-One	Docker Compose
Containers	1	3+ (API, frontend, m3-plugin-runner, traefik)
Best for	Demos, evaluation, small teams	Production, development, scaling
Scaling	Vertical only	Individual service scaling
Hot reload	No	Yes (dev profile)
Process management	supervisord	Docker orchestration
Reverse proxy	nginx (embedded)	Traefik (with dashboard)
Build targets	`allinone`, `prod`	Individual Dockerfiles

Prerequisites

All deployment options require:

Docker Engine 24.0+ and Docker Compose v2.20+
PostgreSQL 15+ with the following schemas: public, marie_scheduler, marie_studio
S3-compatible storage (AWS S3, MinIO, or any S3-compatible service)

Using Marie-AI Infrastructure

If you have the Marie-AI all-in-one stack running, you already have PostgreSQL and S3-compatible storage available:


# Marie-AI provides these services:
# PostgreSQL: localhost:5432 (user: postgres)
# S3 Storage: localhost:8000 (console at :8001)

Configure M3 Forge to point at the shared infrastructure:


DATABASE_URL=postgresql://postgres:123456@localhost:5432/studio
S3_ENDPOINT_URL=http://localhost:8000
S3_ACCESS_KEY_ID=MARIEACCESSKEY
S3_SECRET_ACCESS_KEY=MARIESECRETACCESSKEY

Standalone PostgreSQL + S3 Storage

If deploying M3 Forge independently, start external services first:


# PostgreSQL
docker run -d --name postgres \
  -p 5432:5432 \
  -e POSTGRES_PASSWORD=mysecret \
  -v pgdata:/var/lib/postgresql/data \
  postgres:17
 
# S3-compatible storage (MinIO example)
docker run -d --name minio \
  -p 9000:9000 -p 9001:9001 \
  -e MINIO_ROOT_USER=minioadmin \
  -e MINIO_ROOT_PASSWORD=minioadmin \
  -v miniodata:/data \
  minio/minio server /data --console-address ":9001"

Environment Variables

All deployment methods share the same environment variables. Copy .env.example to .env and configure:


cp .env.example .env

Required:

Variable	Description	Example
`DATABASE_URL`	PostgreSQL connection string	`postgresql://user:pass@host:5432/db`
`S3_ENDPOINT_URL`	S3-compatible endpoint	`http://localhost:9000`
`S3_ACCESS_KEY_ID`	S3 access key	`MARIEACCESSKEY`
`S3_SECRET_ACCESS_KEY`	S3 secret key	`MARIESECRETACCESSKEY`
`JWT_ACCESS_SECRET`	JWT signing secret	Generate with `openssl rand -hex 32`
`JWT_REFRESH_SECRET`	JWT refresh secret	Generate with `openssl rand -hex 32`

Optional:

Variable	Default	Description
`S3_BUCKET_NAME`	`marie`	S3 bucket name
`S3_REGION`	`us-east-1`	S3 region
`WEBAPP_DOMAIN`	`apps.localhost`	Domain for webapp subdomains
`RUNNER_API_KEY`	(none)	API key for m3-plugin-runner auth
`PORT`	`3500`	API server port

Database Migrations

All deployment options run Prisma migrations automatically on startup. To run manually:


# Inside the container
npx prisma migrate deploy --schema packages/@marie/db/prisma/schema.prisma
 
# Or via Docker exec
docker exec m3-forge npx prisma migrate deploy \
  --schema packages/@marie/db/prisma/schema.prisma