Groq Code Companion — Complete Source Bundle

Two production-ready, Vercel-deployable implementations of the Groq Code Companion assistant, sharing the same persona and task modes.

Folder	What it is	Deploy target
web-nextjs/	Next.js 14 web app, streaming responses, Monaco editor, sidebar history	Vercel (recommended), any Node host
telegram-bot/	Telegram bot, same persona, command-driven, per-chat memory	Vercel webhook, Docker, VPS polling

🚀 Quick start

Web app (Next.js + streaming)

cd web-nextjs
npm install
cp .env.example .env.local   # paste GROQ_API_KEY
npm run dev                  # http://localhost:3000

Deploy to Vercel: push to GitHub → import at https://vercel.com/new → add GROQ_API_KEY env var → Deploy.

Telegram bot

cd telegram-bot
npm install
cp .env.example .env         # paste GROQ_API_KEY + TELEGRAM_BOT_TOKEN
npm start                    # polling mode, runs anywhere

Deploy to Vercel: push to GitHub → import → add env vars → register webhook (see telegram-bot/README.md).

🧠 Shared persona

Both clients use the same system prompt (in lib/persona.ts for the web app and src/persona.js for the bot), so the assistant behaves identically across surfaces. The persona mirrors the "Groq Code Companion" custom agent — professional, formal, technically precise, optimized for developers.

✨ Common feature set

• 5 task modes: Generate · Debug · Review · Refactor · Explain
• Code-context attachment for richer prompts
• Multi-language support (TS/JS/Python/Go/Rust/Java/C++/SQL/Bash/...)
• Conversation memory per session / per chat
• Server-side API key (never exposed to clients)

🔁 Default model

Both clients default to llama-3.3-70b-versatile on Groq — the strongest general/coding model on the platform. Override with GROQ_MODEL env var.

📄 License

MIT — use freely.