IGen – Full Stack Generative AI System

This monorepo contains the complete IGen system: a modern frontend, robust backend, and a custom Stable Diffusion inference engine, all orchestrated for scalable, cloud-native deployment.

---

🗺️ Architecture Overview

[Frontend (React+Vite)] ⇄ [Backend (Django+Gunicorn)] ⇄ [Stable Diffusion Inference (Flask, RunPod)]

• Frontend: React SPA (Vite), deployed on Vercel
• Backend: Django app (igen), deployed on Render, proxies API calls
• Inference: Custom Stable Diffusion server (Flask), deployed on RunPod GPU pod

---

📦 Monorepo Structure

IGen/
├── frontend/         # React + Vite UI (Vercel)
├── backend/          # Django API & proxy (Render)
├── StableDiffusion/  # Custom SD inference server (RunPod)
└── README.md         # (this file)

---

🚀 Deployment Overview

Component	Stack	Deployment
Frontend	React, Vite	Vercel
Backend	Django, Gunicorn	Render
Inference Engine	Flask, PyTorch, SD	RunPod (GPU)

• Persistent Volumes: RunPod pod mounts /runpod-volume/weights and /runpod-volume/outputs for model and output persistence.
• API Flow: Frontend → Backend (Django) → Inference (Flask/RunPod)

---

Application Frontend UI Image

---

🔗 Component Summaries

Frontend (details)

• React + Vite SPA
• Calls backend API for prompt submission and image retrieval
• Deployed on Vercel

Backend (details)

• Django app with custom igen module
• Handles API routing, logging, and proxies requests to inference server
• Deployed on Render with Gunicorn

Stable Diffusion Inference (details)

• Custom pipeline: CLIP, VAE Encoder/Decoder, UNet
• Loads weights from HuggingFace (model, tokenizer)
• Flask API exposes /inference endpoint
• Deployed on RunPod GPU pod with persistent storage

---

🧠 Tech Stack

• Frontend: React, Vite, JavaScript/JSX, Vercel
• Backend: Django, Gunicorn, Python, Render
• Inference: Flask, PyTorch, custom SD pipeline, RunPod
• ML Models: CLIP, VAE, UNet (from Umar Jamil’s implementation)
• Tokenizer/Weights: HuggingFace

---

🛠️ Quickstart

1. Frontend
   • See frontend/README.md
2. Backend
   • See backend/README.md
3. Stable Diffusion
   • See StableDiffusion/README.md

---

📡 API Flow Diagram

User
  │
  ▼
[Frontend (React)]
  │  REST
  ▼
[Backend (Django)]
  │  Proxy
  ▼
[Stable Diffusion (Flask, RunPod)]

---

📝 Logging & Debugging

• Frontend: Browser console, UI error messages
• Backend: Django logs, Render dashboard
• Inference: Pod logs (stdout), output images in persistent volume

---

🙏 Acknowledgements

• Umar Jamil for Stable Diffusion implementation
• HuggingFace for model weights/tokenizer
• OpenAI CLIP
• RunPod, Vercel, Render

---

📄 License

This project is for research and educational purposes. See the root LICENSE file for details.