Skip to main content

Deploy Langflow on Docker

This guide demonstrates deploying Langflow with Docker and Docker Compose.

Prerequisites​

Clone the repo and build the Langflow Docker container​

  1. Clone the Langflow repository:

    git clone https://github.com/langflow-ai/langflow.git

  2. Navigate to the docker_example directory:

    cd langflow/docker_example

  3. Run the Docker Compose file:

    docker compose up

Langflow is now accessible at http://localhost:7860/.

Configure Docker services​

The Docker Compose configuration spins up two services: langflow and postgres.

To configure values for these services at container startup, include them in your .env file.

An example .env file is available in the project repository.

To pass the .env values at container startup, include the flag in your docker run command:


_10
docker run -it --rm \
_10
-p 7860:7860 \
_10
--env-file .env \
_10
langflowai/langflow:latest

Langflow service​

The langflowservice serves both the backend API and frontend UI of the Langflow web application.

The langflow service uses the langflowai/langflow:latest Docker image and exposes port 7860. It depends on the postgres service.

Environment variables:

  • LANGFLOW_DATABASE_URL: The connection string for the PostgreSQL database.
  • LANGFLOW_CONFIG_DIR: The directory where Langflow stores logs, file storage, monitor data, and secret keys.

Volumes:

  • langflow-data: This volume is mapped to /app/langflow in the container.

PostgreSQL service​

The postgres service is a database that stores Langflow's persistent data including flows, users, and settings.

The service runs on port 5432 and includes a dedicated volume for data storage.

The postgres service uses the postgres:16 Docker image.

Environment variables:

  • POSTGRES_USER: The username for the PostgreSQL database.
  • POSTGRES_PASSWORD: The password for the PostgreSQL database.
  • POSTGRES_DB: The name of the PostgreSQL database.

Volumes:

  • langflow-postgres: This volume is mapped to /var/lib/postgresql/data in the container.

Deploy a specific Langflow version with Docker Compose​

If you want to deploy a specific version of Langflow, you can modify the image field under the langflow service in the Docker Compose file. For example, to use version 1.0-alpha, change langflowai/langflow:latest to langflowai/langflow:1.0-alpha.

Package your flow as a Docker image​

You can include your Langflow flow with the application image. When you build the image, your saved flow .JSON flow is included. This enables you to serve a flow from a container, push the image to Docker Hub, and deploy on Kubernetes.

An example flow is available in the Langflow Helm Charts repository, or you can provide your own JSON file.

  1. Create a project directory:

_10
mkdir langflow-custom && cd langflow-custom

  1. Download the example flow or include your flow's .JSON file in the langflow-custom directory.

_10
wget https://raw.githubusercontent.com/langflow-ai/langflow-helm-charts/refs/heads/main/examples/flows/basic-prompting-hello-world.json

  1. Create a Dockerfile:

_10
FROM langflowai/langflow-backend:latest
_10
RUN mkdir /app/flows
_10
COPY ./*json /app/flows/.
_10
ENV LANGFLOW_LOAD_FLOWS_PATH=/app/flows

The COPY ./*json command copies all JSON files in your current directory to the /flows folder.

The ENV LANGFLOW_LOAD_FLOWS_PATH=/app/flows command sets the environment variable within the Docker container. By pointing it to /app/flows, you ensure that the application can find and utilize the JSON flow files that have been copied into that directory during the image build process.

  1. Build and run the image locally.

_10
docker build -t myuser/langflow-hello-world:1.0.0 .
_10
docker run -p 7860:7860 myuser/langflow-hello-world:1.0.0

  1. Build and push the image to Docker Hub. Replace myuser with your Docker Hub username.

_10
docker build -t myuser/langflow-hello-world:1.0.0 .
_10
docker push myuser/langflow-hello-world:1.0.0

To deploy the image with Helm, see Deploy the Langflow production environment on Kubernetes.

Customize the Langflow Docker image with your own code​

You can customize the Langflow Docker image by adding your own code or modifying existing components.

This example Dockerfile demonstrates how to customize Langflow by replacing the astradb_graph.py component, but the pattern can be adapted for any other components or custom code.


_20
FROM langflowai/langflow:latest
_20
# Set working directory
_20
WORKDIR /app
_20
# Copy your modified astradb_graph.py file
_20
COPY src/backend/base/langflow/components/vectorstores/astradb_graph.py /tmp/astradb_graph.py
_20
# Find the site-packages directory where langflow is installed
_20
RUN python -c "import site; print(site.getsitepackages()[0])" > /tmp/site_packages.txt
_20
# Replace the file in the site-packages location
_20
RUN SITE_PACKAGES=$(cat /tmp/site_packages.txt) && \
_20
echo "Site packages at: $SITE_PACKAGES" && \
_20
mkdir -p "$SITE_PACKAGES/langflow/components/vectorstores" && \
_20
cp /tmp/astradb_graph.py "$SITE_PACKAGES/langflow/components/vectorstores/"
_20
# Clear Python cache in the site-packages directory only
_20
RUN SITE_PACKAGES=$(cat /tmp/site_packages.txt) && \
_20
find "$SITE_PACKAGES" -name "*.pyc" -delete && \
_20
find "$SITE_PACKAGES" -name "__pycache__" -type d -exec rm -rf {} +
_20
# Expose the default Langflow port
_20
EXPOSE 7860
_20
# Command to run Langflow
_20
CMD ["python", "-m", "langflow", "run", "--host", "0.0.0.0", "--port", "7860"]

To use this custom Dockerfile, do the following:

  1. Create a directory for your custom Langflow setup:

_10
mkdir langflow-custom && cd langflow-custom

  1. Create the necessary directory structure for your custom code. In this example, Langflow expects astradb_graph.py to exist in the /vectorstores directory, so you create a directory in that location.

_10
mkdir -p src/backend/base/langflow/components/vectorstores

  1. Place your modified astradb_graph.py file in the /vectorstores directory.

  2. Create a new file named Dockerfile in your langflow-custom directory, and then copy the Dockerfile contents shown above into it.

  3. Build and run the image:


_10
docker build -t myuser/langflow-custom:1.0.0 .
_10
docker run -p 7860:7860 myuser/langflow-custom:1.0.0

This approach can be adapted for any other components or custom code you want to add to Langflow by modifying the file paths and component names.

Search