Docsmart is a SaaS AI application featuring real recurring subscription payments. It uses OpenAI Embeddings, Pinecone Serverless Vector Database, and Langchain to create an AI Model based on your PDF in which you can chat and interact. It also uses TailwindCSS for styling, Framer Motion for animations, and Three.js for the interactive 3D Models.

Frontend: Next.js 15, React 19, Typescript, TailwindCSS, Framer Motion

3D Models: Three.js for real-time interactive visuals

Authentication: Clerk with secure OAuth + user sessions

AI Model: GPT-4 Turbo via OpenAI API (128k context)

Embeddings & Search: OpenAI Embeddings + Pinecone Serverless

AI Orchestration: Langchain

Payments: Stripe (Subscriptions + Webhooks)

I built it as my submission for the 5-Day AI SaaS Challenge hosted by YouTuber Sonny Sangha—and it earned a nomination among the top projects, receiving enthusiastic feedback for both UI quality and product vision.

📄 Drag-and-drop document uploads Accepts PDFs and handles parsing and chunking using LangChain custom loaders.

💬 Real-time Q&A over documents Users can ask contextual questions. Uses Langchain pipeline to fetch relevant embeddings and generate grounded responses via GPT-4-Turbo.

⚡ Chat Memory & Follow-ups Past messages influence context dynamically. Ideal for multi-step workflows and follow-up refinement.

📚 Source-aware Answers Every AI-generated response includes traceable sources—an essential feature for compliance-heavy use cases.

🌙 Sleek UX Motion-rich UI with cool animated transitions, and error-state fallbacks—all handcrafted using Framer Motion + Tailwind + Three.js.

👉 Live Preview: docsmart-rsujays-projects.vercel.app 👉 Github Repository: github.com/R-Sujay/Docsmart 👉 Nomination: Challenge Finale 👉 Portfolio: sujay-r.vercel.app