You RSVP to a 500-person event. You'll realistically meet ~10 people — but you have no idea who's worth your time, and the ones who matter are invisible in a flat attendee list. Researching everyone by hand (open each LinkedIn, guess their relevance, figure out an ice-breaker, hope for a mutual connection) is impossible at that scale, so most people wing it and miss the highest-value conversation in the room.
WarmPath does that research for you before you walk in: it finds who you should meet, why, a personalized opener, and your warmest intro path — turning a room full of strangers into a ranked, warm game plan.
Work the room before you walk in. Paste a Luma event you're attending, say who you want to meet, and WarmPath scrapes the attendees, enriches them from LinkedIn, ranks them with AI, and shows your warmest intro path to the people worth meeting — then lets you chat about the room and save targets.
🔗 Live: https://warmpath.butterbase.dev
- Scrape the event's attendees from Luma (custom Apify actor, using your logged-in session).
- Enrich each attendee from LinkedIn (Apify) → real title, company, richer headline.
- Rank them with AI against your goal + background — the ranking runs on RocketRide Cloud.
- Store the graph in Neo4j — people, events, and companies as nodes. Because every event you search is saved, WarmPath knows when an attendee also showed up at other events you've searched, and finds warm-intro paths through shared events / employers.
- Everything else lives in Butterbase — auth, your saved sessions/chats/people, the AI gateway, and the app itself.
Attendees, events, and companies are nodes; ATTENDS / WORKS_AT are edges. The agent traverses it with
Cypher: shortestPath for warm intros ("you both worked at Stripe" / "you both went to last month's
event") and shared-attendee queries to recommend related events. Every searched event is persisted,
so the graph accumulates cross-event overlap — that's what makes warm paths possible.
The entire product runs on Butterbase: auth (login required), database (searched sessions, chat threads, saved people — per user), the AI gateway (Claude, for ranking/openers/chat), payments (Stripe Connect subscription tier), and hosting (static frontend + all serverless edge functions).
The attendee-ranking pipeline (chat → llm_openai_api → response_answers) is deployed to RocketRide
Cloud and invoked by the app (via an Apify actor bridge) as a "Deep re-rank." Its LLM node points at
Butterbase's gateway, so RocketRide and Butterbase are composed in a single pipeline.
- Apify (our edge): the custom Luma scraper, LinkedIn enrichment, and the RocketRide bridge.
- Data flow: Luma → Apify (scrape + LinkedIn enrich) → Neo4j (graph) → RocketRide Cloud (rank) → Butterbase (auth, storage, AI gateway, serving).
- Full attendee lists require being registered to the event (Luma gates the guest list).
- Monetization is fully built and ready — a "WarmPath Pro" plan, a Butterbase billing/checkout edge function, a free-tier usage meter, and an "Upgrade to Pro" flow are all wired end-to-end. It just can't be activated: Butterbase payouts run on Stripe Connect, whose onboarding is restricted to US residents, and our team is non-US. So the payment integration is code-complete up to the Stripe onboarding step — everything before that works.