afterparty.digital

Every event ends. The connections shouldn't.

An AI-powered post-event platform. An organizer uploads an event's attendee list, and every attendee gets a personal, time-limited "afterparty" page: who was in their room, the specific people they should reach out to and why, a copy-ready intro for each, and an editable connection graph that dissolves in 30 days. Organizers get a one-screen connection-ROI dashboard.

Built for the DeveloperWeek NY 2026 hackathon. See PLAN.md for the full build log and SUBMISSION.md for the project write-up.

The honesty rule (the core design principle)

With only an attendee list there is no interaction data, so the product never claims to know who met whom. It knows who was present and recommends who's worth meeting. Every connection edge is recommended, marked (you added them), or confirmed (you're following up) — never "they talked." This honesty is the whole differentiator versus during-event networking apps.

What's built

Attendee page (/p/[token], an unguessable tokenized link)

• Vibrant hero: "here was your room," attendee count, your interest clusters
• People you should meet — top matches with a grounded reason each
• Copy intro — a ready-to-send, personalized outreach message per person
• Editable relationship graph — tap to open details, hover to trace links; confirm follow-ups, search-and-add anyone you missed
• Private notes per connection
• Take it with you — export your connections as a vCard before the page dies
• 30-day countdown, framed as privacy-by-design

Organizer dashboard (/dashboard/[eventId])

• ROI stats: attendees, connections mapped, avg intros/attendee, interest clusters
• The full room mapped as a force graph, communities separated and colored
• Clickable cluster legend + searchable attendee table

Ingestion (/upload) — robust CSV/JSON parser that maps messy real-world headers (Full Name/first+last, Job Title/Specialty, social handles → URLs, interest lists), then runs clustering + matchmaking automatically.

Stack

Layer	Choice
Framework	Next.js 16 (App Router, src/) · React 19 · TypeScript
Styling	Tailwind v4 · shadcn/ui · sonner
Database	Supabase (Postgres) via @supabase/supabase-js
AI core	Deterministic clustering + matchmaking (zero API calls); optional Anthropic Claude for interest enrichment
Graph viz	react-force-graph-2d
Hosting	Vercel; whole site password-gated via src/proxy.ts

The matchmaking engine is deterministic and key-free — free, reproducible, and it sends no attendee data to any third party. Claude is an optional enhancement (interest inference), not a dependency.

Local setup

npm install
cp .env.example .env.local        # set the Supabase URL + keys
# apply supabase/migrations/*.sql once (Supabase SQL editor or CLI)
npm run seed:demo                 # optional: seed a demo event from a local CSV
npm run dev                       # http://localhost:3000

• npm test — parser, clustering, and matchmaking unit tests
• Database schema: supabase/migrations/ (see supabase/README.md)
• Deployment + security model: DEPLOY.md

Privacy & security model

• Unguessable URLs — attendee pages use 96-bit tokens; dashboards use the event UUID. No browsable directory of people exists.
• Token-scoped writes — every edit action validates the page token server-side.
• Ephemeral by design — events carry a dissolves_at; pages go dark after it.
• Password gate — set SITE_PASSWORD and the whole deployment sits behind HTTP Basic Auth (plus noindex), so non-public demo data is never exposed.