Skip to content

Quick Start


There are instructions to quickly database service stack in 4 steps and a working ECS cluster to process your video in 1 step.

What, what? Yes, it really is that simple.


  • Bash Shell for running setup and launch scripts
  • Docker - Required to launch the services
  • Git - Used to check out the code
  • AWS account for launching a processing cluster in the Elastic Container Service (optional)


Your AWS account must have full permission to provision EC2 instances, create buckets, and setup job queues and definitions to use the ECS Cluster. See deepsea-ai setup instructions to setup AWS permissions.

Checkout the code and change to that working directory

git clone
cd deepsea-ai-backend

Database service stack

This will create the database, graphql api, and workers in development mode and load a small sample database.

1 - Setup an environment


2 - Run

./bin/ postgres all

3 - Add an admin user

./bin/ gulpereel true

You should see a token output similar to the following: "token": "eyJhbGciOiJIUzI1NiIsI ...."

4 - Use that token to load some test data

Copy the entire token here - it is abbreviated for the sake of brevity in the example.

./bin/ eyJhbGciOiJIUzI1NiIsI... /srv/data/DEEPSEA_AI/tracks/V4244_20191205T165023Z_h264_1min.tracks.tar.gz

The database and API should be up now and a small sample should now be loaded to experiment with in GraphiQL.

That's it! Now, open http://localhost:4000/graphql and check-out the example queries and mutations.

Loading data

Loading data can occur a number of ways:

  1. From a file e.g. to load a test file in the docker image docker exec deepsea-ai-worker ts-node src/worker/cli/load.ts --force \ --token eyJhbGciOiJIUzI1NiIsI... \ --file /srv/data/DEEPSEA_AI/tracks/V4244_20191205T165023Z_h264_1min.tracks.tar.gz

  2. From data in an S3 bucket e.g. docker exec deepsea-ai-worker ts-node src/worker/cli/load.ts --force \ --token eyJhbGciOiJIUzI1NiIsI... \ --s3 s3://my-track-bucket Note you must have permission to read the bucket

  3. From a local file or S3 bucket through a queue. This is the preferred method as it allows for the load to be queued and processed in the background. docker exec deepsea-ai-worker ts-node src/worker/cli/load.ts \ --token eyJhbGciOiJIUzI1NiIsI... \ --s3 s3://my-track-bucket \ --queue In method 3, the load will be queued and processed in the background. See the Bull Dashboard for the status of the load.


To shut down the software, run ./bin/ postgres all

Your data will not be deleted. It is stored in a docker volume. To delete the data, run docker volume rm deepsea-ai-backend_deepsea-ai-data deepsea-ai-backend_deepsea-ai-db deepsea-ai-backend_deepsea-ai-redis

Deploy Cluster

This deploys an autoscaling detection and tracking workflow that leverages the AWS Elastic Container Service ECS. This deployment is simplified with the Cloud Development Kit CDK - a much simpler tool for deploying work in the cloud than navigating complex Kubernetes infrastructure or very long CloudFormation templates.

Important Points

  • You don't get charged for having a cluster, only when you use resources in your cluster
  • By default, all resources are removed when destroying the stack so be sure to remove any videos of track data that you might need before tearing the stack down.

1 - Build your cluster