End-to-end personal finance platform with AI, in production
Personal finance app with Supabase auth, a balance and projection dashboard, transaction management, per-category budgets, recurring entries, import/export and a conversational AI assistant (Luna). Turbo monorepo with a Python backend and a Next.js frontend, continuously deployed on Vercel. The screenshots alongside are from the authenticated platform in production.
Luna: context assembled from the authenticated user's own transactions, categories and budgets, through parameterized queries filtered by user_id
LLM through the OpenAI SDK pointed at Groq (Llama 3.3 70B), a 500-token output cap and history truncated to the last 10 messages
Per-route rate limiting with slowapi: 30 req/h on chat, 10/h on insights, 5/h on the monthly report
Row level security in Postgres on Luna's digest tables, on top of Supabase auth
Turbo monorepo: FastAPI API + Next.js frontend, continuously deployed on Vercel
Answer: The SDK is the same; only the base_url changes. Llama 3.3 70B on Groq delivers far lower latency at a cost a personal project can carry, and switching providers is one line of configuration if the trade-off changes.
Answer: The month's financial context already fills the system prompt. With no cap, a long conversation drives cost per request up and answer quality down — 10 messages covers the continuity of a finance chat.
Luna — conversational AI finance assistant
Investments — total net worth and goals
Responsive layout on mobile
Home prices served by a FastAPI API on AWS: 9.32% MAPE against actual 2010 sales, every request logged, continuous deploy with a healthcheck
