ADP fork of Langflow — adds multi-tenant orgs, template management, Data Mapper, ADP Assist, and a distributed (Arq + Redis) flow-execution runtime. The effective main branch is platform-multi-tenant.

This README covers how to run and deploy this fork. For feature docs, see the spec tree under docs/superpowers/specs/.

Services in this fork

Service	What it is	Default port	Depends on
Backend API (langflow run / make backend)	FastAPI app, flow runtime, REST API	7860	Postgres, Redis (if distributed execution is on)
Frontend (make frontend)	Vite/React UI	3000 (dev)	Backend
Worker (uv run langflow worker)	Arq consumer for webhook/schedule/MCP flow runs	n/a	Redis, Postgres
Redis	Arq broker + cache + concurrency signals	6379	—
Postgres	Primary database (SQLite fallback for local)	5432	—

See deploy/README.md for the full deployment reference (env vars, docker-compose, queue topology).

Running locally

Three terminals, one command each. This is the exact flow used in development.

0. Prerequisites

• Python 3.11+ (floor was raised from 3.10 for the pandas 3.0 bump) and uv
• Node.js 18+ and npm
• A running Redis (brew services start redis or docker run -p 6379:6379 redis:7)
• A running Postgres (optional; SQLite is the default fallback)
• A .env file at the repo root — copy .env.example and fill in at least LANGFLOW_DATABASE_URL, LANGFLOW_REDIS_URL, and LANGFLOW_DISTRIBUTED_EXECUTION=true

First-time only:

make init

1. Backend (:7860)

make backend env=.env

Runs uvicorn against langflow.main:create_app. Reads the .env file passed in, so all LANGFLOW_* vars apply. Override port with port=8080, workers with workers=4, etc. See the backend README.

2. Frontend (:3000)

make frontend

Installs deps and runs npm start. Proxies API calls to the backend via BACKEND_URL (default http://localhost:7860). See the frontend README.

3. Worker (task runner)

uv run langflow worker --queue runs:default --concurrency 8

Arq worker that executes webhook/schedule/MCP-triggered flow runs out-of-process. Requires Redis (via LANGFLOW_REDIS_URL) and the same Postgres the backend uses. Run one process per queue — see deploy/README.md for the full queue list and horizontal-scaling notes.

Deploying

The canonical deployment is deploy/docker-compose.yml (Traefik + backend + frontend + Postgres + Redis + 3 workers + optional pgadmin/prometheus/grafana). Start there:

cd deploy
cp .env.example .env   # edit values first
docker compose up

Full env-var table, port map, and horizontal-scaling guidance live in deploy/README.md.

Project layout

Path	Purpose
src/backend/base/	langflow-base package — FastAPI app, services, CLI, workers, Alembic migrations
src/lfx/	lfx package — graph engine, components, lfx serve/run CLI, worker settings
src/frontend/	React UI (Vite, Tailwind v4)
src/sdk/	Python SDK for calling deployed flows
deploy/	Production docker-compose stack
docker/	Dockerfiles + dev compose variants
docker_example/	Minimal single-container example
docs/	Docusaurus site + spec tree

Further reading

• AGENTS.md — contributor/AI-agent context (authoritative)
• DEVELOPMENT.md — upstream dev notes
• CONTRIBUTING.md — upstream contribution flow (note: this fork does not PR upstream)
• deploy/README.md — deployment reference
• src/backend/base/README.md — backend dev/deploy
• src/frontend/README.md — frontend dev/deploy
• src/lfx/README.md — lfx CLI + worker