Deploy and configure AngeMedia Gateway
Make credentials and data persistent first. Add more providers, concurrency and external access only after the basic deployment survives a restart.
Recommended deployment
Use the repository Docker Compose setup for a long-running service. It keeps Gateway, queues and dependencies in a reproducible configuration.
export ADMIN_USERNAME=admin
export ADMIN_DEFAULT_PASSWORD='replace-with-a-long-random-password'
export GATEWAY_API_KEY='replace-with-a-long-random-gateway-key'
docker compose up -d --build
Minimum required settings
| Variable | Purpose | Recommendation |
|---|---|---|
ADMIN_USERNAME | Web Studio administrator | Use a non-default name |
ADMIN_DEFAULT_PASSWORD | Initial administrator password | Use a long random value and never commit it |
GATEWAY_API_KEY | API mode and protected media access | Store separately from the administrator password |
Provider configuration
SILICONFLOW_API_KEY=
MODELSCOPE_API_KEY=
POLLINATIONS_API_KEY=
AGNES_API_KEY=
OPENAI_IMAGE_API_KEY=
OPENAI_IMAGE_BASE_URL=https://api.openai.com/v1
OPENAI_IMAGE_MODEL=gpt-image-2
Configure only providers that are actually used. An empty optional provider should not break other routes.
Timeouts
IMAGE_PROVIDER_TIMEOUT=300
VIDEO_PROVIDER_TIMEOUT=900
Video jobs usually take longer than image jobs. Change these values only when the provider is healthy but Gateway ends the request too early.
Persistent data
- Database: jobs, events, attempts and deduplication state.
- Uploads: reference images and temporary inputs.
- Generated media: localized images and videos.
- Local configuration: administrator, Gateway key and provider settings.
Post-start verification
- Check
/health. - Sign in to Studio and confirm Dashboard can read job and storage state.
- Test one provider connection.
- Generate one image and open it from Assets.
- Restart the containers and confirm credentials, jobs and assets remain.