Serenova

Breast health education, reconstruction visualization, and a source-grounded AI assistant, built for people navigating breast cancer and recovery.

🔗 Live demo: serenova-sage.vercel.app

⚠️ Disclaimer: Serenova is an educational tool. It does not diagnose, recommend treatment, or replace advice from a qualified healthcare professional. AR/VR previews are approximations, not predictions of surgical results.

---

Features

🩺 Breast health resources

• Landing site, self-check guide, and curated resources pages
• Clear, supportive, educational content

🧬 Reconstruction visualization (/regeneration)

• Interactive 3D viewer (react-three-fiber + drei) with orbit controls
• Procedure types: implant, DIEP, latissimus, fat grafting
• Controls for color, wireframe, and opacity
• AR / VR mode via WebXR: place the model in your room (AR) or view it in an immersive scene (VR)
• Capability-based detection with a 2D fallback on unsupported devices

💬 Grounded breast-health chatbot (/chat)

• Streaming chat powered by Google Gemini
• RAG (Retrieval-Augmented Generation): answers are grounded in cited excerpts from CDC, NCI, ACS, USPSTF, and FDA guidance
• Inline [S1]-style citations with source links
• "Grounded answers" toggle so users can compare with and without retrieval
• Supabase auth, per-user conversation history, and Upstash Redis rate limiting

---

Tech stack

Layer	Tools
Framework	Next.js 16 (App Router), React 19, TypeScript
UI	Tailwind CSS, shadcn/ui
3D / XR	react-three-fiber, drei, @react-three/xr
Auth & DB	Supabase (Auth + Postgres with RLS)
AI	Google Gemini (chat + text-embedding-004), Vercel AI SDK
Vector search	Pinecone (768-dim, cosine)
Validation & limits	Zod, Upstash Redis
Hosting	Vercel

---

Architecture

knowledge-base.ts ──chunk──► embed (text-embedding-004) ──► Pinecone (dim 768)
                                                                   │
user message ──embed query──► Pinecone topK ──► formatRagContext ─┘
                                                       │
                                          prepend to system prompt
                                                       │
                                          Gemini streamText ──► cited answer

reconstruction-mesh.tsx   ← procedural geometry + per-procedure profiles (shared)
        │
        ├── model-viewer.tsx   (desktop: OrbitControls, overlays)
        └── xr-viewer.tsx      (WebXR: immersive-ar / immersive-vr)

For the full design, see IMPLEMENTATION.md.

---

Getting started

Prerequisites

• Node.js 18+
• A Supabase project
• A Google AI Studio API key
• A Pinecone index with dimension 768 and cosine metric
• An Upstash Redis database

1. Clone and install

git clone https://github.com/amitkojha05/serenova-.git
cd serenova-
npm install

2. Configure environment

Create .env.local:

NEXT_PUBLIC_SUPABASE_URL=
NEXT_PUBLIC_SUPABASE_ANON_KEY=
SUPABASE_SERVICE_ROLE_KEY=        # optional, admin tasks
GOOGLE_GENERATIVE_AI_API_KEY=     # Gemini chat + embeddings
PINECONE_API_KEY=
PINECONE_INDEX=                   # dimension 768, cosine
UPSTASH_REDIS_REST_URL=
UPSTASH_REDIS_REST_TOKEN=
ADMIN_EMAILS=                     # comma-separated

3. Set up the database

Apply the schema in supabase/ to create the RLS-protected conversations and messages tables.

4. Seed the knowledge base

npm run rag:ingest

5. Run

npm run dev      # http://localhost:3000
npm run build    # production build
npm run lint

---

Trying the features

• RAG chat: open /chat, keep "Grounded answers: on", and ask "When should I start mammograms?". The answer should reflect USPSTF/ACS guidance with inline citations. Toggle grounding off to compare.
• AR / VR: open /regeneration → AR / VR Mode.
  • Chrome on Android (ARCore) over HTTPS: "View in AR" places the model in your room.
  • Meta Quest and other headsets: AR and/or VR buttons are enabled.
  • Desktop: buttons are disabled with an explanation; the 2D preview still renders.

WebXR requires HTTPS. Vercel provides this automatically.

---

Project structure

app/          Next.js routes and API handlers (/chat, /regeneration, /api/chat)
components/   UI, chat, and regeneration (3D / XR) components
hooks/        Custom React hooks
lib/          RAG retrieval, Supabase clients, rate limiting, helpers
scripts/      Knowledge base corpus and RAG ingestion
supabase/     Database schema and policies
types/        Shared TypeScript types
public/       Static assets

---

Knowledge base sources

The curated corpus paraphrases guidance from:

• CDC: overview, risk factors, symptoms, screening
• NCI (NIH): types, BRCA1/BRCA2 genetics, treatment overview
• ACS: screening guidelines, reconstruction options
• USPSTF: mammography screening recommendations
• FDA: breast implant safety information

Health guidance changes over time. Review the corpus against the live source pages periodically.

---

Roadmap

• Photo-based personalized 3D body meshes (replacing the procedural mesh)
• Manufacturer implant shape libraries as morph targets
• Soft-tissue deformation for more realistic previews
• Expanded and regularly refreshed knowledge base

---

Contributing

Issues and pull requests are welcome. For larger changes, please open an issue first to discuss what you'd like to change.

License

Released under the MIT License.