Langflow deployment overview

You have a flow, and want to share it with the world in a production environment.

This page outlines the journey from locally-run flow to a cloud-hosted production server.

More specific instructions are available in the Docker and Kubernetes pages.

Langflow deployment architecture

Langflow can be deployed as an IDE or as a runtime.

The IDE includes the frontend for visual development of your flow. The default docker-compose.yml file hosted in the Langflow repository builds the Langflow IDE image. To deploy the Langflow IDE, see Docker.

The runtime is a headless or backend-only mode. The server exposes your flow as an endpoint, and runs only the processes necessary to serve your flow, with PostgreSQL as the database for improved scalability. Use the Langflow runtime to deploy your flows, because you don't require the frontend for visual development.

Package your flow with the Langflow runtime image

To package your flow as a Docker image, copy your flow's .JSON file with a command in the Dockerfile.

An example Dockerfile for bundling flows is hosted in the Langflow Helm Charts repository.

For more on building the Langflow docker image and pushing it to Docker Hub, see Package your flow as a docker image.

Deploy to Kubernetes

After your flow is packaged as a Docker image and available on Docker Hub, deploy your application by overriding the values in the langflow-runtime Helm chart.

For more information, see Deploy Langflow on Kubernetes.