SamurAIGPT/Vibe-Workflow

Vibe Workflow – Open‑source, self‑hosted node‑based AI workflow builder

What it is

• A free, MIT‑licensed web application that lets you visually compose pipelines for generative image and video creation. It mirrors the node‑editor experience of commercial tools like Weavy AI, Krea Nodes, Freepik Spaces and FloraFauna AI, but you run it on your own hardware.

Core components

• Frontend – a Next.js app (client/) that provides the drag‑and‑drop canvas.
• Workflow‑builder library – a reusable UI package (packages/workflow-builder/) that implements the node system.
• Backend – a FastAPI service (server/) that forwards node requests to a generative model provider (MuAPI/Vadoo AI) and returns results.

Key features

Feature	Details
Node‑based visual editor	Build, connect, and reorder modules (e.g., prompt, model, post‑process) without writing code.
Generative image & video	Uses MuAPI’s diffusion and video models; you supply your own API key.
Self‑hostable	Docker Compose or manual setup; no data leaves your infrastructure.
Extensible	Add custom nodes that call any external API or load LoRA/custom diffusion models.
API & embed support	The hosted version (muapi.ai/workflow) offers a REST API and embeddable widget; the open‑source repo can be extended similarly.
MIT license	Free to modify, redistribute, and integrate into other products.

Typical use cases

• Creative studios – automate large‑scale asset generation while keeping brand guidelines in‑house.
• Developers – embed a ready‑made visual workflow widget into SaaS products or internal tools.
• Researchers – experiment with new diffusion models or video generators without vendor lock‑in.
• No‑code creators – design complex generative pipelines through a UI rather than scripting.

Getting started (quick‑start)

1. Clone & install

   git clone https://github.com/samuraigpt/vibe-workflow.git
   cd vibe-workflow
   npm install   # installs client + library via workspaces
   

2. Add a MuAPI key

   cd server
   cp .env.example .env
   # edit .env → MU_API_KEY=your_key
   

3. Run locally
   • Frontend: npm run dev:app → http://localhost:3000
   • Backend:

     python -m venv venv && source venv/bin/activate
     pip install -r requirements.txt
     uvicorn app.main:app --reload --host 0.0.0.0 --port 8000
     

4. Or use Docker Compose

   cp .env.example .env   # set MU_API_KEY
   docker compose up --build
   

   Frontend on 3000, API on 8000.

Extending the platform

• Create a new node in packages/workflow-builder/ that defines input/output ports and the request it should make.
• Register the node on the backend so the FastAPI server knows how to handle the payload.
• Rebuild the library (npm run build:lib) and restart the client.

Why choose Vibe Workflow over the commercial alternatives?

• Open source & free – no subscription fees.
• Full control of data – everything runs on your servers.
• Plug‑and‑play with any model – swap MuAPI for another provider or host your own diffusion model.
• Reusable library – the node editor can be imported into other projects.

Resources

• Full story & motivation: [Medium article]
• Hosted SaaS version (if you just want a ready UI): https://muapi.ai/workflow
• Contributing guide: CONTRIBUTING.md
• License: MIT

---

Vibe Workflow is essentially a visual “pipeline builder” for generative AI, aimed at creators who want the power of node‑based editors without being locked into a paid SaaS platform.