No-Code / Low-Code Intelligent Workflow Platform

Overview

React + FastAPI app to visually compose workflows (User Query → KnowledgeBase → LLM Engine → Output) and chat through the pipeline. Uses PostgreSQL for workflows and logs, ChromaDB for vector search, Gemini for LLM and embeddings, SerpAPI for optional web search, PyMuPDF for text extraction.

Prerequisites

• Node 18+
• Python 3.10+
• PostgreSQL running and accessible

Environment Variables

Backend:

GEMINI_API_KEY=your_gemini_key
SERPAPI_KEY=your_serpapi_key
DATABASE_URL=postgresql://user:password@localhost/workflow_builder
# Optional
GEMINI_EMBEDDING_MODEL=text-embedding-004

Frontend (optional):

REACT_APP_API_BASE=http://localhost:8000

Install & Run

Backend:

pip install -r backend/requirements.txt
uvicorn backend.app.main:app --reload --host 0.0.0.0 --port 8000

Frontend:

cd frontend
npm install
npm start

Usage

1. Build Stack

• Drag components in the left panel onto the canvas, connect: userQuery → knowledgeBase (optional) → llmEngine → output.
• Configure nodes from the right panel.
• Click "Build Stack" to save the workflow. This assigns an ID.

2. Upload Documents (optional)

• POST file to /api/documents/upload/{workflow_id}.
• The file text is extracted (PyMuPDF), chunked, embedded (Gemini), and stored in ChromaDB under workflow_{id} collection.

3. Chat with Stack

• Click "Chat with Stack" and ask questions. The app will follow your edges, optionally retrieve context from the knowledge base and web search, then answer with Gemini.

Architecture Highlights

• Backend
  • FastAPI routers: workflows, documents, chat
  • WorkflowExecutor orchestrates nodes in order
  • DocumentProcessor manages extraction, embeddings, and ChromaDB
  • GeminiClient for LLM responses and embeddings
  • SearchService for SerpAPI web results
• Frontend
  • React Flow builder with configurable node types
  • Zustand store persists saved workflow ID for chat
  • API base configurable via REACT_APP_API_BASE

Notes

• ChromaDB persists to ./chroma_db.
• Ensure you set GEMINI_API_KEY and SERPAPI_KEY before running.
• Saved workflow ID is required before chatting.

Optional

• Add auth, workflow versioning, execution logs, chat history views.