Fragwise is an open-source fragrance discovery platform with a built-in AI chatbot guide. It serves connoisseurs cataloging notes, accords, and houses, and beginners who just want to find a scent they will love. Project URL (once deployed): https://fragwise.app

In active development, runnable locally, not deployed yet. Phases 0 through 4 are complete: monorepo tooling with docker compose and a justfile, the Postgres + pgvector schema and fragrance ontology, catalog read endpoints, hybrid lexical + vector search, the LangGraph chat guide (Wisp), and the web UI (home, catalog and taxonomy pages, dark theme). Auth is in progress as phase 5; deployment to fragwise.app follows.

• Web: Next.js 16 + Tailwind CSS v4 + shadcn/ui
• API: Python FastAPI + LangGraph
• Database: Neon Postgres + pgvector
• Cache / rate limit: Upstash Redis
• Object storage: Cloudflare R2
• Auth: Clerk
• LLM: OpenAI
• Hosting: Vercel (web) + Fly.io (api), pay-as-you-go with scale-to-zero
• Self-hostable via docker-compose

Requires just, pnpm, uv (Python 3.12), and Docker.

cp .env.example .env   # fill in what you need
just install           # pnpm workspace + Python venv via uv
just db-up             # Postgres with pgvector via docker compose
just dev               # web on :3000, api on :8000

Run just with no arguments to list the full recipe set (tests, lint, migrations, db shells).

Licensed under Apache-2.0. Inbound contributions are licensed outbound under the same terms; see CONTRIBUTING.md.