A visual, node-based automation platform for building, running, and monitoring multi-step workflows. Wire together reusable nodes on a drag-and-drop canvas — call APIs, transform data, branch and loop, run scripts, and send notifications — then schedule them, trigger them from webhooks/email/files, and inspect every run.

• Frontend — React 19 + Vite, ReactFlow canvas, Zustand, Monaco editor
• Backend — Node.js 22 + Fastify 5, SQLite (better-sqlite3), a worker-thread run engine with 34 built-in node types
• Sidecars — WhatsApp bridge, Whisper voice-to-text, optional Ollama / cloud LLMs

See DOCS.md for the full user guide and node reference.

Quick start (Docker)

git clone <your-repo-url> flowline && cd flowline
cp .env.example .env    # then edit secrets/URLs as needed
docker compose up --build

Then open http://localhost:3001. Services:

Service	Port	Purpose
app	3001	Backend API + built frontend
whatsapp-bridge	3002	WhatsApp send/receive bridge
voice-to-text	9000	Whisper transcription sidecar

Data (SQLite DB, files, license/instance state) persists in the workflow_data volume mounted at /app/data.

Secure your instance. Set API_TOKEN in docker-compose.yml — without it, anyone who can reach port 3001 can read workflows, run them, and export decrypted secrets. See the "Admin & Backup" section of DOCS.md.

Local development

# backend  (http://localhost:3001)
cd backend && npm install && npm run dev

# frontend (Vite dev server)
cd frontend && npm install && npm run dev

Copy .env.example to .env first — both npm run dev and Docker Compose read it for WEBHOOK_SECRET, optional API_TOKEN, sidecar URLs, and the premium license key.

Building

Each package builds independently. Node 22 is required (the backend compiles the native better-sqlite3 addon).

# backend → TypeScript compiled to backend/dist
cd backend && npm install && npm run build

# frontend → static bundle in frontend/dist
cd frontend && npm install && npm run build

Run the backend test suite (and lint the frontend) before shipping:

cd backend && npm test
cd frontend && npm run lint

The production Docker image performs both builds in separate stages and copies the frontend bundle into the backend's public/ directory, so a single app container serves the API and the UI together.

Deployment

The supported deployment path is Docker Compose. It builds three services from this repo — app (API + frontend), whatsapp-bridge, and voice-to-text — and provisions the persistent volumes.

git clone <your-repo-url> flowline && cd flowline
cp .env.example .env       # set WEBHOOK_SECRET and API_TOKEN at minimum
docker compose up --build -d

Then open http://localhost:3001.

Before exposing an instance publicly:

• Set API_TOKEN. Uncomment it in .env / docker-compose.yml. Without it, anyone who can reach port 3001 can read workflows, run them, and export decrypted secrets.
• Set WEBHOOK_SECRET and use the same value in every webhook trigger in the UI and in the WhatsApp bridge.
• Set CORS_ORIGIN if the UI is served from a different origin (otherwise requests are same-origin only).
• Put a TLS-terminating reverse proxy (nginx, Caddy, Traefik) in front of port 3001; the app speaks plain HTTP.

Persistent state lives in named volumes — the SQLite DB, files, and license/instance state in workflow_data (mounted at /app/data), WhatsApp auth in wa_auth, and Whisper models in vtt_models. Back up workflow_data regularly; see the "Admin & Backup" section of DOCS.md.

GPU note: the voice-to-text service requests an NVIDIA GPU (needs the NVIDIA Container Toolkit). To run Whisper on CPU, set WHISPER_DEVICE=cpu in .env and remove the deploy.resources block from that service in docker-compose.yml.

Updating: pull the latest changes and rebuild — git pull && docker compose up --build -d. Schema migrations run automatically on startup; the data volume is preserved across rebuilds.

Editions & licensing

Flowline is open-core:

• The core — everything except the premium directories below — is free software under the GNU AGPL-3.0. You may self-host, modify, and use it for any purpose, including commercially, subject to AGPL's terms (notably: if you offer a modified version as a network service, you must publish your source changes).

• The premium tier is a paid add-on under a separate commercial license (LICENSE-PREMIUM), unlocked at runtime by a signed license key:

  Premium feature	Directory
  LLM Assistant	backend/src/plugins/assistant/
  Housekeeping	backend/src/plugins/housekeeping/
  Artifact History	backend/src/plugins/artifact-history/

  Premium source ships in this repo for transparency, but using or enabling it requires a commercial license and key. How the key gate works (Ed25519-signed, instance-bound, expiring) is documented in PREMIUM-LICENSING.md.

Running without a license key gives you the full free edition — no premium surfaces, everything else works.

Documentation

• DOCS.md — user guide, node reference, keyboard shortcuts
• PREMIUM-LICENSING.md — premium licensing & key enforcement
• CONTRIBUTING.md — how to contribute (note the CLA)