Luma

Tasks, calendar, goals and a focus timer in one place, with XP, streaks and achievements to turn organization into a habit. The assistant doesn't just answer: it creates tasks, logs progress and plans your day.

• An AI assistant with 25 tools that change real data: create and reschedule tasks, log goal progress, plan the day. Replies stream over SSE, with up to 5 tool steps per turn.
• A gamification engine driven by data: points, levels, streaks, achievements, an in-app currency with a ledger, a shop, avatars and weekly leaderboards.
• Per-user timezones: hourly jobs close streaks and leaderboards and send notifications at each user's local midnight, daylight saving included.
• Google Calendar sync over OAuth, with tasks dragged straight onto the calendar.
• A modular Fastify API: 17 modules with the same shape (routes, controller, repository, schema), Zod validation, rate limiting and a single error type.

1. 01A separate API

   I split the API from the Next.js front end so a future native app can consume the same API, and because a dedicated server is more robust for scheduled jobs and business rules.

2. 02Game rules in the database

   XP values, triggers and rewards live in the database so the admin is in control: anything can be changed straight from the admin panel, whenever needed, without a deploy.

3. 03Keeping AI cost bounded

   A sliding window of the last 20 messages, the user's daily context cached for 5 minutes, small models, token usage logged per call, and a rate limit on the chat route.

4. 04One hourly job, converted per user

   Instead of one scheduled job per timezone, a single hourly job checks which users just crossed midnight in their own timezone.

5. 05Cutting scope on purpose

   I built subscriptions, trials and credits, then removed them to focus on the core loop, and merged the app and the API into one public monorepo.