Intentionality

A productivity app for students that turns coursework into quests, with streaks, focus sessions, friends, and an AI planner.

Solo, design to deploy · 193 commits · Dec 2025 to Mar 2026

Scale: 300+ users

Results

• 300+ users
• 46 tables, all with row-level security
• 2 LLM providers with automatic failover

The problem

Students juggle coursework, calendars and habits across separate tools, and a plain to-do list makes procrastination easy. Intentionality makes the work feel like progress, and plans from the student's real calendar and courses.

What I built

• Quests, XP and levels, achievements, daily habits and streaks, a focus timer, analytics, friends, groups, and leaderboards.
• A Postgres schema of 46 tables across 44 migrations, with row-level security on every table.
• Kofi, an AI assistant for chat, a daily briefing, and turning a brain dump into structured tasks.
• Calendar sync from Google Calendar and Canvas or any ICS feed.

The hard part

Keeping the AI features usable on free-tier model quotas. I wrote a router that gives each feature a primary and a fallback provider, opens a per-provider circuit breaker after three straight failures, enforces per-user daily caps, and turns provider errors into friendly messages, so a rate limit never breaks the app.

How it fits together

• Web app to API routes
• Calendars to API routes (sync)
• API routes to Database (queries)
• API routes to AI router (plans)
• AI router to LLM providers (failover)

Tradeoff: two cheaper providers with failover instead of one paid one, so the app stays free to run.

Decisions

• Two free-tier providers with failover instead of one paid provider, so the app stays free to run and a rate limit never breaks it.
• Row-level security on every table, not just checks in API routes, so a bug in one route can't leak another student's data.

Tech: Next.js, TypeScript, Supabase, LLMs. Live Code

By David Antwi (david.antwi@yale.edu). Markdown version: /work/intentionality.md. Full profile: /llms.txt.