Cloud Deployment

Relevant source files

• deploy/.env.example
• deploy/README.md
• deploy/docker-compose.override.yml
• deploy/docker-compose.yml
• deploy/prometheus.yml
• docker/build_and_push.Dockerfile
• docker/build_and_push_backend.Dockerfile
• docker/build_and_push_base.Dockerfile
• docker/build_and_push_ep.Dockerfile
• docker/build_and_push_with_extras.Dockerfile
• docker/cdk.Dockerfile
• docker/dev.Dockerfile
• docs/docs/Components/bundles-nextplaid.mdx
• docs/docs/Components/bundles-openai-compatible.mdx
• docs/docs/Components/bundles-openai.mdx
• docs/docs/Components/bundles-openrouter.mdx
• docs/docs/Components/bundles-vllm.mdx
• docs/docs/Components/components-embedding-models.mdx
• docs/docs/Components/components-models.mdx
• docs/docs/Components/human-input.mdx
• docs/docs/Deployment/deployment-docker.mdx
• docs/docs/Flows/human-in-the-loop.mdx
• docs/docs/Lfx/extensions-bundle-list.mdx
• docs/docs/_partial-global-model-providers.mdx
• docs/versioned_docs/version-1.11.0/Components/bundles-nextplaid.mdx
• docs/versioned_docs/version-1.11.0/Components/human-input.mdx
• docs/versioned_docs/version-1.11.0/Deployment/deployment-docker.mdx
• docs/versioned_docs/version-1.11.0/Flows/human-in-the-loop.mdx
• docs/versioned_docs/version-1.11.0/Lfx/extensions-bundle-list.mdx
• src/lfx/docker/Dockerfile
• src/lfx/docker/Dockerfile.dev

This page documents the technical implementation of Langflow's cloud deployment strategies, focusing on infrastructure-as-code (IaC) for AWS, containerized orchestration with Docker Compose, and specialized configurations for multi-worker environments and monitoring.

1. AWS Deployment (CDK/Fargate/Aurora)

Langflow provides a production-ready AWS deployment path using the AWS Cloud Development Kit (CDK). This setup automates the provisioning of a highly available, scalable architecture.

1.1 Architecture Overview

The AWS deployment follows a tiered architecture:

• Edge Layer: Amazon CloudFront serves static frontend assets from S3 and proxies API requests to the backend.
• Compute Layer: Langflow's FastAPI backend runs on AWS Fargate (Serverless ECS).
• Database Layer: Amazon Aurora (PostgreSQL-compatible) serves as the persistent store.
• Security: Database credentials and secrets are managed via AWS Secrets Manager.

1.2 System Entity Mapping (AWS CDK)

The following diagram maps high-level AWS services to the specific infrastructure components and code entities used in the deployment implementation.

Title: AWS Infrastructure Entity Mapping

Sources: docker/cdk.Dockerfile1-5

---

2. Docker Compose Stack (Deploy)

The deploy/ directory contains a comprehensive docker-compose.yml for production-grade container orchestration, utilizing a multi-service architecture for scalability.

2.1 Service Architecture

The stack uses Traefik v3.0 as a reverse proxy to handle routing and SSL termination via Let's Encrypt deploy/docker-compose.yml2-51 It defines multiple networks including a public Traefik network and a default internal network deploy/docker-compose.yml6-8

Title: Docker Compose Service Architecture

Sources: deploy/docker-compose.yml2-100 deploy/docker-compose.yml124-153

2.2 Key Components

• Backend: Runs the main Langflow application using the langflowai/langflow-backend image, communicating with PostgreSQL for persistence deploy/docker-compose.yml69-85
• Celery Worker: Executes long-running tasks using the langflow.worker.celery_app module, typically configured with eventlet pool for high concurrency deploy/docker-compose.yml133-142
• Database Management: Includes pgAdmin for PostgreSQL administration, accessible via the pgadmin. subdomain deploy/docker-compose.yml101-123
• Task Monitoring: Includes Flower, a real-time monitor for Celery, running on port 5555 deploy/docker-compose.yml143-157

2.3 Monitoring and Observability

The deployment setup includes a Prometheus configuration to scrape metrics from the system.

• Prometheus: Scrapes metrics from itself and the Flower service every 15 seconds deploy/prometheus.yml1-12
• Grafana & Loki: The stack supports log aggregation through a Grafana-Loki-Promtail pipeline. Dashboards are provisioned automatically for production logs deploy/observability/grafana-loki/grafana/provisioning/dashboards/dashboards.yml1-10
• Logging: Langflow uses structlog for structured logging, which can be formatted as container_json or container for cloud log aggregators.

Sources: deploy/docker-compose.yml1-160 deploy/prometheus.yml1-12 deploy/observability/grafana-loki/grafana/provisioning/dashboards/dashboards.yml1-10

---

3. Docker Image Variants

Langflow provides multiple Docker image variants based on the Red Hat Universal Base Image (ubi10/python-314-minimal) to support different deployment needs.

3.1 Dockerfile Comparison

Image Variant	Dockerfile Path	Key Characteristics
Full	docker/build_and_push.Dockerfile	Includes frontend assets, backend core, and all standard workspace members docker/build_and_push.Dockerfile1-152
Base	docker/build_and_push_base.Dockerfile	Lean core framework only; excludes extension bundles like lfx-duckduckgo docker/build_and_push_base.Dockerfile81-87
Backend-Only	docker/build_and_push_backend.Dockerfile	No frontend assets or Playwright; optimized for API-only deployments docker/build_and_push_backend.Dockerfile1-7
Extras	docker/build_and_push_with_extras.Dockerfile	Includes heavy dependencies like couchbase, cassio, and nv-ingest docker/build_and_push_with_extras.Dockerfile57-59

3.2 Runtime Environment Security

Production images run as a non-root user (UID 1000) docker/build_and_push.Dockerfile122 To prevent startup crashes, LANGFLOW_CONFIG_DIR (defaulting to /app/langflow) is pre-created with correct ownership to allow the container user to write secret keys and profile pictures docker/build_and_push.Dockerfile130-137

Title: Container Runtime Entity Mapping

Sources: docker/build_and_push.Dockerfile124-137 deploy/docker-compose.yml137 deploy/.env.example11-12

---

4. Persistent Database Configuration

While Langflow defaults to SQLite, production cloud deployments require a robust external database.

4.1 PostgreSQL Integration

For containerized environments, Langflow is configured to use PostgreSQL. The deploy/ stack uses postgres:15.4 deploy/docker-compose.yml87 Database connectivity is managed via the environment variable LANGFLOW_DATABASE_URL.

4.2 Multi-Worker Support (Redis & RabbitMQ)

Cloud deployments with Celery workers utilize RabbitMQ as the message broker and Redis as a result backend deploy/docker-compose.yml124-132

• Broker: BROKER_URL=amqp://langflow:langflow@broker:5672 deploy/.env.example45
• Result Backend: RESULT_BACKEND=redis://result_backend:6379/0 deploy/.env.example46

Sources: deploy/docker-compose.yml86-100 deploy/.env.example23-47

---

5. Deployment Workflow

5.1 Docker Quickstart

For rapid deployment using the provided example:

1. Navigate to the deploy/ directory.
2. Copy .env.example to .env and configure DOMAIN, STACK_NAME, and LANGFLOW_SUPERUSER_PASSWORD deploy/.env.example1-12
3. Run docker compose up -d to start the Traefik proxy, backend, database, and workers.

5.2 GCP (Cloud Run/Spot)

Langflow can be deployed to GCP Cloud Run by using the standard Docker images. Configuration requires setting LANGFLOW_DATABASE_URL to a Cloud SQL instance and ensuring LANGFLOW_CONFIG_DIR points to a persistent volume (like Cloud Storage FUSE) or is used for ephemeral storage.

5.3 Log Retrieval

In cloud environments, logs can be retrieved via API if LANGFLOW_ENABLE_LOG_RETRIEVAL is set to True. This utilizes an internal buffer defined by LANGFLOW_LOG_RETRIEVER_BUFFER_SIZE.

Sources: deploy/.env.example1-53 docker/build_and_push.Dockerfile138-151