CodeExplainr

Small Next.js + FastAPI app that can explain, analyze, and convert code using a local LLM via Ollama.

Tech Stack

Frontend

• Next.js (App Router), React 18, TypeScript
• Tailwind CSS + shadcn/ui (Radix) + lucide-react
• Client streaming over fetch; diff + analysis views in the UI

Backend

• FastAPI + Uvicorn (Python 3.10+)
• Pydantic for request/response models
• requests to call the Ollama HTTP API

LLM Runtime

• Ollama (local) with Code Llama models
  e.g. codellama and codellama:7b-instruct-q4_0

Requirements

• Node 18+
• Python 3.10+
• Ollama running on http://127.0.0.1:11434
• A model pulled (e.g. codellama or codellama:7b-instruct-q4_0)

ollama pull codellama

Backend (FastAPI)

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

Create backend/.env: OLLAMA_BASE_URL=http://127.0.0.1:11434 OLLAMA_MODEL=codellama

Run:

uvicorn main:app --reload --port 8000

Frontend (Next.js)

npm install
npm run dev
# opens http://localhost:3000 (may switch to 3001 if 3000 is busy)

Notes

Keep one Ollama server running (service or ollama serve, not both).

If the browser blocks requests, add your dev URL to FRONTEND_ORIGINS and restart the backend.

If Ollama throws a 500 about CUDA/“unable to allocate”, free some RAM or use the 7b-instruct-q4_0 tag.