People-first networking, powered by AI.
Connvo recreates the vibe of a coffee chat so you can instantly meet peers, mentors, and collaborators who match your goals, interests, and personality. Connections have never been easier.
Connvo exists for the moment when a “quick chat” turns into a partnership. We combine an adaptive personality quiz, intentional AI matchmaking, and live collaboration tools so every conversation feels effortless and human.
- Intelligent matching: Proprietary models pair you with people who complement your interests, goals, and communication style.
- Guided coffee chats: Dynamic icebreakers, suggested questions, and live prompts keep conversation flowing from the first minute.
- In-call collaboration: Shared notes, participant context, action items, and transcription live beside the call so everyone stays aligned.
- Always-on follow-up: Recaps, recommended next steps, and future matches arrive automatically after each call.
- Compliance-first: SOC- and FERPA-aligned guardrails, opt-in controls, and content moderation keep conversations on-mission.
An adaptive onboarding flow captures personality insights, availability, and goals. Connvo’s matching engine blends embeddings, heuristics, and feedback signals to introduce the right people at the right moment.
Inside every voice or video session you’ll find shared notes, participant cards, suggested prompts, and optional transcription. No more juggling tools—everything lives in one workspace.
A hybrid WebRTC + Stream media stack delivers reliable audio/video with automatic fallback modes. Launch calls instantly, whether you’re meeting one-to-one or with a small cohort.
After the call, Connvo generates beautiful summaries, action items, and context-aware follow-ups. The more you use the app, the smarter recommendations become.
| Category | Technologies |
|---|---|
| Monorepo 🚀 |  All-in-one monorepo management |
| Front-End ✨ |      Hosted on Vercel |
| Authentication 🔒 |  |
| Real-Time Communication 📡 |  Hybrid WebRTC + Stream media pipeline for resilient calls |
| Backend & Realtime 💾 |    Convex functions, cron jobs, realtime subscriptions, and ML embeddings |
| ML Matching Engine 🤖 |   Vector-based similarity (dot products, cosine/euclidean distance) + semantic analysis using OpenAI |
| Area | Technologies |
|---|---|
| Frontend | Next.js 16, React 19, Motion, TailwindCSS 4, Radix UI, Lenis |
| Backend & Realtime | Convex (database, functions, realtime subscriptions), Convex crons & tasks |
| Authentication | WorkOS AuthKit for SSO + passwordless flows |
| Media | Stream Video SDK + browser-native WebRTC fallback |
| AI & Matching | Custom Convex pipeline, OpenAI embeddings, similarity scoring, feedback loops |
| Docs & Tooling | Mintlify, Redocly, pnpm, Vitest, ESLint, Prettier |
The repo is a monorepo anchored around Convex. Frontend UI (Next.js) and backend logic (Convex functions) live side-by-side, sharing Zod schemas and generated TypeScript types.
- Node.js 20+
- pnpm (
corepack enableto install) - Convex CLI (
npm install -g convex) - Stream and WorkOS accounts (see
.env.example)
pnpm installCopy .env.example (and optionally .env.prod) to .env, then fill in:
- WorkOS credentials (
WORKOS_CLIENT_ID,WORKOS_API_KEY, etc.) - Stream video credentials (
STREAM_API_KEY,STREAM_SECRET) - OpenAI key for embeddings (
OPENAI_API_KEY) - Convex deployment URLs (
CONVEX_DEPLOYMENT,NEXT_PUBLIC_CONVEX_URL)
pnpm convex:devThis keeps database schema and function changes in sync with your deployment.
pnpm devNow visit http://localhost:3000 for the web experience and http://localhost:3000/app after signing in.
| Command | Description |
|---|---|
pnpm dev |
Run Next.js in dev mode with Turbopack |
pnpm convex:dev |
Start Convex locally (required for backend functions & codegen) |
pnpm build |
Production build |
pnpm start |
Run the built app |
pnpm lint |
ESLint |
pnpm test |
Vitest unit + integration suites |
pnpm type-check |
TypeScript project check |
pnpm update:api-docs[:env] |
Regenerate OpenAPI docs from Convex functions |
pnpm audit:docs |
Validate docstrings for API documentation |
- Matching pipeline playbook:
docs/MatchingPipeline.pdfoutlines the Convex-based matching architecture, embeddings strategy, and feedback loop. - API reference:
docs/api-reference/convex-openapi.yamlis regenerated from Convex docstrings viapnpm update:api-docs[:env]. - Authoring guides:
docs/API_DOCUMENTATION.mdcovers environment setup, validation commands, and troubleshooting tips for the docs pipeline. - Mintlify portal:
docs/mint.jsonfeeds our developer site. Runmintlify devto preview changes locally. - Automation: CI refreshes the OpenAPI bundle and Mintlify content on pushes to
mainwhen theUpdate API Docsworkflow secrets are configured.
Connvo’s matching engine blends structured preferences, embeddings, and feedback signals:
- Profile Vectorization: User intents, interests, and communication styles are embedded via OpenAI.
- Compatibility Scoring: Multi-objective scoring weights complementarity, availability overlap, and historical outcomes.
- Positive Feedback Loop: Follow-up rates, call duration, and sentiment annotations tune future matchmaking.
Supporting documentation lives in docs/MatchingPipeline.pdf, and the Convex implementation can be explored under convex/matching/.
- Expand cohort-based matching (3–5 person themed sessions).
- Bring end-to-end encryption to live calls.
- Deepen analytics and insights for community managers.
- Launch automated onboarding journeys for larger org roll-outs.
- Ship AI-guided prompts and collaborative call workspace.
- Build SOC- and FERPA-aligned moderation guardrails.
- Deliver Stream-powered video and transcription pipeline.
- Human-first design wins: Matching quality improves dramatically when we optimize for conversation outcomes rather than simple profile similarity.
- Convex + Next.js accelerates iteration: Sharing types, validation, and auth context between UI and backend reduced context switching.
- Guardrails matter: SOC/FERPA alignment shaped data flows early, making compliance a feature instead of a retrofit.
- Feedback fuels the loop: Instrumentation around call quality and follow-up behavior is core to improving the model—not an afterthought.
Next.js • Convex • WorkOS • Stream Video • OpenAI