YafAI Studio

YafAI Studio is a local AI code generator. You describe a task in plain English, and a small open-source code model, DeepSeek Coder 1.3B Instruct, writes the code for it. A FastAPI backend runs the model on the CPU with Hugging Face Transformers, and a Next.js frontend provides the prompt interface.

Features

• Prompt-to-code generation with deepseek-ai/deepseek-coder-1.3b-instruct, running on the CPU with no GPU or paid API required
• REST API: a single POST /generate endpoint that takes a prompt and returns the generated code
• Web UI: a prompt box, a generated-code panel with one-click copy, and a dark/light theme toggle
• Model pre-download script that fetches the weights once before you start the server

Tech stack

Layer	Technology
Model	DeepSeek Coder 1.3B Instruct (Hugging Face Transformers, PyTorch)
Backend	Python, FastAPI, Uvicorn
Frontend	Next.js 16 (App Router), React 19, TypeScript, lucide-react

Getting started

Prerequisites

• Python 3.9+
• Node.js 20+ and npm
• About 3 GB of free disk space for the model weights, plus enough RAM to run a 1.3B-parameter model in float32

1. Backend

cd backend
python -m venv venv
source venv/bin/activate        # Windows: venv\Scripts\activate
pip install -r requirements.txt

python download_model.py        # one-time download of the model weights
uvicorn app:app --reload        # serves on http://127.0.0.1:8000

2. Frontend

cd frontend
npm install
npm run dev                     # http://localhost:3000

The UI sends requests to http://127.0.0.1:8000/generate, so start the backend first.

API

POST /generate

// request
{ "prompt": "Write a Python function that checks if a string is a palindrome" }

// response
{ "code": "..." }

Generation uses sampling (temperature 0.2, top-p 0.95) and stops after at most 128 new tokens. You can change these settings in backend/model.py.

Project structure

backend/
  app.py               FastAPI app, CORS, /generate endpoint
  model.py             Model loading and generate_code()
  download_model.py    Pre-downloads the model and tokenizer
  requirements.txt
frontend/
  app/page.tsx         Main UI (prompt, output, theme toggle)
  app/layout.tsx       Root layout and metadata
  lib/api.ts           API helper (reads NEXT_PUBLIC_API_URL)
  package.json

Notes

• CORS is open (allow_origins=["*"]) for local development. Restrict it before deploying.
• The backend URL is hardcoded in app/page.tsx. A helper in lib/api.ts reads NEXT_PUBLIC_API_URL instead, but the page does not use it yet.