---
Source: https://docs.microblink.com/on-prem/docker-compose
Title: Docker Compose
Description: How to deploy the on-prem image on a single host using Docker Compose
---

# Docker Compose

The Docker Compose deployment runs the on-prem image as one container holding the API, the processing workers, and the runtime models.
Use it for a single host.
For a cluster, see [Kubernetes and Helm](./kubernetes.md).

## Prerequisites

- Docker 20.10.5 or newer
- Docker Compose 2.22.0 or newer
- A Linux AMD64 host
- A license key and application ID
- The image tag for the release you're deploying

The deployment files are in the [on-prem-ops repository](https://github.com/microblink/on-prem-ops):

```bash
git clone https://github.com/microblink/on-prem-ops.git
```

The Compose file and the environment template are under `deploy/docker-compose/`.

## Configure the deployment

Copy `docker-compose.yaml` and `.env.example` into your deployment directory, and rename `.env.example` to `.env`.

Keep `.env` out of version control and restrict access to it: it holds your license key.

```dotenv title=".env"
BLINKID_VERIFY_IMAGE=us-docker.pkg.dev/document-verification-public/on-prem/core
BLINKID_VERIFY_IMAGE_TAG=4000.0.2
BLINKID_VERIFY_API_PORT=8080

LICENSE_KEY=<license-key>
LICENSE_APPLICATION_ID=<application-id>
DOCVER_WORKFLOW=ExtractAndVerify
WORKER_COUNT=2

BLINKID_VERIFY_INFLIGHT_LIMIT=2
BLINKID_VERIFY_QUEUE_LIMIT=0
LOG_FILES_ENABLED=0

LIMITS_CPUS=4
LIMITS_MEMORY=8G
RESERVATIONS_CPUS=2
RESERVATIONS_MEMORY=4G
```

Set `BLINKID_VERIFY_IMAGE_TAG` to the version for the release you're deploying, and leave the image repository unchanged unless you were given a different registry.
Keep `DOCVER_WORKFLOW=ExtractAndVerify` unless you're running an extraction-only deployment.

See [Environment variables](./environment-variables.md) for every supported setting and its default.

Always keep the memory limit set.
Overload protection is a percentage of the container memory limit, and it's disabled when there is no limit, so without one the container can't reject requests before it runs out of memory.

The reference defaults reject excess requests immediately rather than building an in-container backlog.
Review the [capacity settings](./environment-variables.md#workflow-and-capacity) before changing the admission limits.

## Start the deployment

```bash
docker compose --file docker-compose.yaml up -d
```

The API is available on port `8080` by default.
Wait for it to become ready:

```bash
curl --fail http://localhost:8080/health/ready
```

Then send a request:

```bash
curl --fail --location --output sample-front.jpg https://storage.googleapis.com/microblink-data-public/microblink-api/test-set/blinkid/SGP_ID_FRONT_new/SGP_ID_FRONT_sample.jpg
```

```bash
curl --fail --location --output sample-back.jpg https://storage.googleapis.com/microblink-data-public/microblink-api/test-set/blinkid/SGP_ID_BACK_new/SGP_ID_BACK_sample.jpg
```

```bash
curl --request POST --form "imageFirstSide=@sample-front.jpg" --form "imageSecondSide=@sample-back.jpg" http://localhost:8080/api/v3/verify
```

## Logs and status

```bash
docker compose --file docker-compose.yaml ps
```

```bash
docker compose --file docker-compose.yaml logs -f blinkid-verify
```

The container writes component logs to standard output and standard error.
Optional file logging is described under [Logging](./environment-variables.md#logging).

## Stop the deployment

```bash
docker compose --file docker-compose.yaml down
```

## Update the deployment

Change the image tag to the release you're moving to, then recreate the container:

```bash
docker compose --file docker-compose.yaml pull
```

```bash
docker compose --file docker-compose.yaml up -d
```

Review the [changelog](./changelog.md) for configuration or API changes before updating.


Last updated on Aug 6, 2026