Deployment

Run locally, deploy to Docker, or ship to the cloud.

Local development

The fastest way to run a pipeline. No containers, no infrastructure.

rf run

This reads flow.yaml in the current directory, validates schemas, and executes every node in topological order. Results land in each node’s artifacts/ directory. State is tracked in .rf/state.db.

For iterative development, validate before running:

rf validate        # check for schema errors, type mismatches, cycles
rf run             # execute the pipeline
rf inspect         # view results and execution history

Local mode is best for development, personal automation, and sensitive data that shouldn’t leave your machine.

Docker deployment

For repeatable execution on any machine or server.

Dockerfile

FROM ghcr.io/radh-io/radhflow:latest

COPY flow.yaml /rf/project/flow.yaml
COPY nodes/ /rf/project/nodes/
COPY data/ /rf/project/data/

WORKDIR /rf/project

CMD ["rf", "run"]

docker-compose.yaml

version: "3.9"
services:
  radhflow:
    image: ghcr.io/radh-io/radhflow:latest
    ports:
      - "8080:80"
    volumes:
      - ./flow.yaml:/rf/project/flow.yaml
      - ./nodes:/rf/project/nodes
      - ./data:/rf/project/data
      - rf-state:/rf/project/.rf
    environment:
      RF_MODE: local
      RF_LOG_LEVEL: info
    restart: unless-stopped

volumes:
  rf-state:

docker compose up -d

Volume mounts

Mount	Purpose
flow.yaml	Pipeline definition
nodes/	Node specs and implementations
data/	Input data files
rf-state	Persistent state across runs (.rf/ directory)

Environment variables

Variable	Default	Description
RF_MODE	local	Deployment mode: local, saas, enterprise
RF_PORT	80	HTTP port inside the container
RF_PROJECT_PATH	/rf/project	Pipeline workspace root
RF_CONFIG_PATH	/rf/config	Config and database directory
RF_LOG_LEVEL	info	Log verbosity: debug, info, warn, error
RF_CREDENTIALS_KEY	—	Encryption key for credential vault

Cloud deployment (Radhflow Cloud)

Radhflow Cloud provides managed infrastructure, scheduling, monitoring, and a credential vault. You deploy with a single command.

rf deploy

This pushes your pipeline to Radhflow Cloud, which handles:

• Scheduled execution (cron-based or event-triggered)
• Secret storage and injection
• Execution monitoring and alerting
• Artifact storage with versioning

Note

Cloud deployment is in beta. Self-hosted Docker is production-ready.

Radhflow Cloud runs on EU infrastructure (Hetzner and Scaleway). Data stays in EU jurisdiction. No US hyperscaler in the data path.

Component	Infrastructure
Compute	Hetzner Cloud (Nuremberg / Helsinki)
Object storage	Scaleway S3 (Paris / Amsterdam)
Container registry	GitHub Container Registry

CI/CD integration

GitHub Actions

name: Pipeline CI/CD

on:
  push:
    branches: [main]
  pull_request:
    branches: [main]

jobs:
  validate:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - name: Install Radhflow
        run: npm install -g @radh/flow-cli
      - name: Validate pipeline
        run: rf validate --strict

  run:
    needs: validate
    if: github.ref == 'refs/heads/main'
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - name: Install Radhflow
        run: npm install -g @radh/flow-cli
      - name: Run pipeline
        run: rf run
        env:
          API_KEY: ${{ secrets.API_KEY }}

Use rf validate in CI to catch errors on every pull request. Use rf run in CD to execute the pipeline on merge to main.

Monitoring and observability

Execution logs

Every rf run writes logs to .rf/runs/<run-id>/. Each run directory contains:

• log.ndjson — timestamped execution events
• summary.json — node statuses, durations, row counts
• Per-node artifacts in nodes/<slug>/artifacts/

Health checks

In Docker or cloud deployments, the Radhflow container exposes a health endpoint:

curl http://localhost:8080/health
# {"status":"ok","version":"0.3.0"}

Alerting on node failures

When a node fails, the executor:

1. Logs the error with stack trace to .rf/runs/<run-id>/log.ndjson
2. Marks the node status as error in summary.json
3. Halts downstream nodes that depend on the failed node
4. On Radhflow Cloud: sends an alert via configured webhook or email

Check execution status programmatically:

rf inspect --run latest --format json