🎙️ RepoTalk

Talk to any codebase — voice-powered AI code analysis.

Paste a GitHub URL, and have a real-time voice conversation with an AI that understands the entire repository. Ask about architecture, get code reviews, find bugs, or just explore how things work.

Built for the Mistral Worldwide Hackathon — Best Voice Use Case (ElevenLabs)

---

Features

• Instant repo analysis — Paste a GitHub URL and get a full codebase breakdown in seconds
• Voice conversations — Talk to the AI using your microphone, hear responses spoken back
• Deep code understanding — Powered by Mistral Large for accurate, context-aware answers
• Code reviews on demand — Ask for bug detection, architecture review, or improvement suggestions
• Zero setup demo mode — Works without API keys using simulated responses

How it works

┌──────────┐     ┌───────────┐     ┌──────────┐     ┌────────────┐
│  GitHub   │────▶│  FastAPI   │────▶│  Mistral  │────▶│ ElevenLabs │
│   URL     │     │  Backend   │     │  Large    │     │    TTS     │
└──────────┘     └───────────┘     └──────────┘     └────────────┘
     │                │                   │                 │
     │           Parse repo          Analyze &          Voice
     │           structure           answer             output
     │                │                   │                 │
     ▼                ▼                   ▼                 ▼
┌─────────────────────────────────────────────────────────────────┐
│                    🎙️ Browser UI (Voice + Text)                  │
│                                                                  │
│  [Mic] ──▶ Speech-to-Text ──▶ Chat ──▶ TTS ──▶ [Speaker]       │
└─────────────────────────────────────────────────────────────────┘

Tech Stack

Component	Technology
AI Engine	Mistral AI (mistral-large-latest)
Voice	ElevenLabs TTS + Web Speech API STT
Backend	Python, FastAPI
Repo Parsing	GitHub REST API via httpx
Frontend	Vanilla HTML/CSS/JS — dark terminal aesthetic

Quick Start

1. Clone

git clone https://github.com/nicovlr/repotalk.git
cd repotalk

2. Install dependencies

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

3. Set API keys

cp .env.example .env
# Edit .env with your keys

4. Run

cd backend
python main.py

Open http://localhost:8000 in your browser.

Demo mode

No API keys? No problem — the app runs in demo mode with simulated AI responses. Great for testing the UI and flow.

API Keys

Key	Required	Get it
MISTRAL_API_KEY	For AI analysis	console.mistral.ai
ELEVENLABS_API_KEY	For voice output	elevenlabs.io
GITHUB_TOKEN	Optional (higher rate limits)	github.com/settings/tokens

Project Structure

repotalk/
├── backend/
│   ├── main.py              # FastAPI app & endpoints
│   ├── github_fetcher.py    # GitHub API repo parsing
│   ├── code_analyzer.py     # Context builder for Mistral
│   ├── mistral_client.py    # Mistral chat + mock client
│   └── requirements.txt
├── frontend/
│   └── index.html           # Single-page UI
├── .env.example
├── .gitignore
├── LICENSE
└── README.md

Built for

🏆 Mistral Worldwide Hackathon — Paris, Feb 28 – Mar 1, 2026

Targeting the Best Voice Use Case prize by ElevenLabs.

Sponsors: Mistral AI, ElevenLabs, Hugging Face, NVIDIA, AWS, Supercell

License

MIT — see LICENSE