Pretendo

Pretendo is a full-stack AI chat app that turns a normal prompt into a reply with a chosen character, mood, and OpenRouter model. It is built as a monorepo: a polished Next.js client at / and a Python FastAPI service at /api, deployed together on Vercel Services.

This project is small on purpose, but it covers the parts that matter in a real product: API boundaries, model validation, local persistence, deployment, environment config, tests, and a UI that feels finished.

What It Shows

• Built and deployed a multi-service monorepo with Next.js and FastAPI on Vercel's experimental Services setup.
• Integrated OpenRouter with validated free-model selection and returned the exact model used for every reply.
• Added local-first chat history with browser SQLite through sql.js.
• Shipped a responsive chat UI with character, mood, and model controls.
• Hardened the API with CORS allowlists, backend-owned system prompts, request limits, provider timeouts, and focused tests.
• Kept the project production-minded with TypeScript checks, API tests, web tests, Vercel Analytics, and Speed Insights.

Tech Stack

Layer	Tools
Frontend	Next.js 16, React 19, TypeScript, Tailwind CSS v4, Phosphor Icons
Client storage	sql.js browser SQLite
Backend	FastAPI, Pydantic, Python 3.12
AI	OpenRouter, OpenAI-compatible client
Deployment	Vercel experimental Services, Vercel Analytics, Speed Insights
Tooling	pnpm, uv, ESLint, node:test, Python unittest

Run Locally

cd api
uv run fastapi dev

cd web
pnpm dev

Open http://localhost:3000.

Checks

cd api && uv run python -m unittest discover -s tests
cd web && pnpm test && pnpm exec tsc --noEmit && pnpm build