GenAI Workflow Builder

A platform that enables users to visually construct, manage, and interact with Intelligent RAG (Retrieval-Augmented Generation) workflows.

This application allows you to build custom AI "Stacks" by dragging and dropping components to connect Google's Gemini LLM with personal document knowledge bases, all through a seamless visual interface.

Key Features

• Visual Workflow Canvas: Built with React Flow for a high-performance, node-based experience.
• RAG-as-a-Service: Upload PDFs and extract knowledge using Google Cloud Embeddings (text-embedding-004).
• Graph-Based Engine: A FastAPI backend that parses visual connections and executes workflows as Directed Acyclic Graphs (DAGs).
• Intelligent LLM Integration: Uses the latest google-genai SDK with built-in retry logic for handling API Quota (429) errors and strict v1 API versioning to avoid 404 model errors.
• Full Persistence: Metadata stored in PostgreSQL; Vector embeddings stored in a persistent ChromaDB instance.

---

Tech Stack

• Frontend: React.js, Vite, React Flow, Tailwind CSS.
• Backend: FastAPI, SQLAlchemy, PyMuPDF (Text Extraction).
• AI/ML: Google Gemini (1.5 Flash), Google Cloud Embeddings.
• Vector Store: ChromaDB (Cloud-backed).
• Infrastructure: Docker, Docker Compose, Node-Serve.

---

Prerequisites

• Docker & Docker Compose installed on your machine.
• Google AI Studio API Key: Obtain one for free at aistudio.google.com.

---

Quick Start

The entire stack (Frontend, Backend, PostgreSQL, and ChromaDB) is containerized for a one-command setup.

1. Clone the repository:

   git clone https://github.com/amann-codes/workflowbuilder.git
   cd workflowbuilder

2. Configure Environment Variables: Create a .env file in the root directory (where docker-compose.yml is):

   GEMINI_API_KEY=your_actual_api_key_here

3. Launch the Stack:

   docker compose up --build

4. Access the Application:

   • Frontend UI: http://localhost:5173
   • Backend API: http://localhost:8000
   • Interactive API Docs (Swagger): http://localhost:8000/docs

---

Manual Local Setup (Development)

If you wish to run the components without Docker:

Backend

1. cd workflow-builder-be
2. Create and activate a venv: python -m venv venv -> source venv/bin/activate
3. Install dependencies: pip install -r requirements.txt
4. Set .env with GEMINI_API_KEY and DATABASE_URL=sqlite:///./test.db.
5. Run: uvicorn main:app --reload --port 8000

Frontend

1. cd workflow-builder
2. Install: npm install
3. Set .env with VITE_API_URL=http://localhost:8000.
4. Run: npm run dev

---

How to Use the Builder

1. Dashboard: Click "New Stack" and give your workflow a name.
2. The Canvas: Drag components from the left sidebar onto the workspace.
   • User Query: The entry point for your questions.
   • Knowledge Base: Upload a PDF.
   • LLM (Gemini): Configure your system prompt.
   • Output: The final chat destination.
3. Connections: Connect User Query -> LLM and Knowledge Base -> LLM. Finally, connect LLM -> Output.
4. Execution: Click the Save icon, then click the Play (Chat) button. Ask questions based on your uploaded document!