- Get started
- Label Studio features
- Billing & Usage
- Release notes
Security and Privacy
- Secure Label Studio
Install and Upgrade
- Install and upgrade Label Studio
- Database setup
- Start Label Studio
Deploy and Install Enterprise
- Install Label Studio Enterprise
- Set up persistent storage
- Set up an ingress controller
- Install with Docker
- Deploy on Kubernetes
- Install on airgapped server
- Install on Amazon EKS
- Available Helm values
- Set up user accounts
- Manage access
- Set up authentication
- Set up SCIM2
- Import data
- Import pre-annotations
- Cloud storage setup
Labeling and Projects
- Project setup
- Manage data
- Set up your labeling interface
- Label and annotate data
- Review annotations
- Annotation statistics
- Custom agreement metric
- Export annotations
Machine Learning Setup
- Machine learning integration
- Write your own ML backend
- ML Examples and Tutorials
- Active learning loop
- Troubleshoot machine learning
- Webhook Setup
- Webhooks Event Reference
- Custom Webhooks
- Python SDK Tutorial
- Backend API
- Frontend library
- Frontend reference
- Update scripts and API calls
Label Studio uses a database to store project data and configuration information.
The SQLite database works well for projects with tens of thousands of labeling tasks. If you want to annotate millions of tasks or anticipate a lot of concurrent users, use a PostgreSQL database. See Install and upgrade Label Studio for more.
For example, if you import data while labeling is being performed, labeling tasks can take more than 10 seconds to load and annotations can take more than 10 seconds to perform. If you want to label more than 100,000 tasks with 5 or more concurrent users, consider using PostgreSQL or another database with Label Studio.
Label Studio uses SQLite by default. You don’t need to configure anything. Label Studio stores all data in a single file in the specified directory of the admin user. After you start Label Studio, the directory used is printed in the terminal.
You can also store your tasks and completions in a PostgreSQL database instead of the default SQLite database. This is recommended if you intend to frequently import new labeling tasks, or plan to label hundreds of thousands of tasks or more across projects.
Run the following command to launch Label Studio, configure the connection to your PostgreSQL database, scan for existing tasks, and load them into the app for labeling for a specific project.
label-studio start my_project --init -db postgresql
You must set the following environment variables to connect Label Studio to PostgreSQL:
DJANGO_DB=default POSTGRE_NAME=postgres POSTGRE_USER=postgres POSTGRE_PASSWORD= POSTGRE_PORT=5432 POSTGRE_HOST=db
When you start Label Studio using Docker Compose, you start it using a PostgreSQL database:
docker-compose up -d
If you’re using a Docker container, Heroku, or another cloud provider, you might want your data to persist after shutting down Label Studio. You can export your data to persist your labeling task data and annotations, but to preserve the state of Label Studio and assets such as files that you uploaded for labeling, set up data persistence.
Mount Docker volumes on your machine to persist the internal SQLite database and assets that you upload to Label Studio after you terminate a Docker container running Label Studio.
If you’re starting a Docker container from the command line, use volumes to persist the data. See the Docker documentation for Use volumes. For example, replace the existing volume flag in the Docker command with a volume that you specify:
docker run -it -p 8080:8080 -v <yourvolume>:/label-studio/data heartexlabs/label-studio:latest
If you’re using Docker Compose with the config included in the Label Studio repository, you can set up Docker volumes in the
docker-compose.yml file for Label Studio:
version: "3.3" services: label_studio: image: heartexlabs/label-studio:latest container_name: label_studio ports: - 8080:8080 volumes: - ./mydata:/label-studio/data volumes: mydata:
For more about specifying volumes in Docker Compose, see the volumes section of the Docker Compose file documentation.
Host a PostgreSQL server that you manage and set up the PostgreSQL environment variables with Label Studio to persist data from a cloud provider such as Heroku, Amazon Web Services, Google Cloud Services, or Microsoft Azure.