Skip to main content

Deploy with an agent

Copy agent prompt

Before you start

Connect your AWS account and choose a project. Review the upstream deployment instructions for the release you plan to pin. See Langfuse environment variables.

Data dependencies

PostgreSQL, ClickHouse, compatible Redis or Valkey, and private S3-compatible object storage. See SQL databases for private writer access and connection details.

Deployment order

  1. Create PostgreSQL in the Spring Winter project and wait for its writer. Provision ClickHouse, the cache and object storage before starting either Langfuse process. Verify private access and the cache’s supported topology.
  2. Set DATABASE_URL and CLICKHOUSE_URL. Give the migration process a reachable CLICKHOUSE_MIGRATION_URL and the required grants. Configure the cache connection and LANGFUSE_S3_EVENT_UPLOAD_* settings using the pinned release’s configuration.
  3. Deploy matching Langfuse web and worker images. Set shared secrets consistently and set NEXTAUTH_URL to the HTTPS web URL. Use a supported Spring Winter worker source, such as a GitHub Dockerfile wrapping the pinned upstream worker image, or run the worker on the chosen external host.
  4. Run the release’s PostgreSQL and ClickHouse migration sequence. If the web image runs migrations on startup, wait for its readiness. Start the worker after migrations succeed and send traffic once both processes are healthy.
Use Runtime variables for runtime credentials. For a Spring Winter web server, set an app-supported health path that returns HTTP 200 without credentials. Check History before running the application check.

Verify the deployment

Submit a sample trace with a project key and confirm it appears in the UI after worker processing.