Savant is an AI-powered PDF research assistant designed to help users extract meaningful insights from their PDFs. Built on Google's Gemini AI, Savant enables users to upload documents, chat with them, and even analyze resumes for ATS compatibility.
🔍 No more scanning through pages manually! With Savant, you get precise, AI-powered answers directly from your documents.
✅ AI-Powered PDF Chat – Upload a PDF and chat with it anytime to extract key information.
✅ Smart Resume Analysis – Check if a resume is ATS-friendly for job applications.
✅ Vectorized Search – Uses Google's text-embedding-004 model for accurate context-based responses.
✅ Cloud Storage – Securely uploads PDFs to AWS S3 for easy access.
✅ Powerful AI Retrieval – Queries are vectorized, matched in Pinecone DB, and processed by Gemini AI for highly accurate answers.
✅ User Authentication – Powered by Clerk, ensuring a secure and seamless login experience.
✅ Subscription System – Monthly Stripe payments for continued access.
- Frontend: Next.js, React, TypeScript, TailwindCSS, Shadcn UI
- Backend: Google Gemini AI, Pinecone DB, NeonDB, LangChain, Drizzle ORM
- Authentication: Clerk
- Cloud Storage: AWS S3
- Payments: Stripe
Landing page where after successful login, users can upload and interact with their PDFs.
Example chat session with an uploaded PDF.
- Node.js (v18 or later)
- NPM or Yarn
- An
.envfile with the required API keys (see below)
git clone https://github.com/yourusername/savant.git
cd savantnpm installCreate a .env.local file in the root directory and add:
# Clerk
NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY=
CLERK_SECRET_KEY=
NEXT_PUBLIC_CLERK_SIGN_IN_URL=/sign-in
NEXT_PUBLIC_CLERK_SIGN_UP_URL=/sign-up
NEXT_PUBLIC_CLERK_AFTER_SIGN_IN_URL=/
NEXT_PUBLIC_CLERK_AFTER_SIGN_UP_URL=/
# Database
DATABASE_URL=
# AWS S3
NEXT_PUBLIC_AWS_ACCESS_KEY_ID=
NEXT_PUBLIC_AWS_SECRET_ACCESS_KEY=
NEXT_PUBLIC_S3_BUCKET_NAME=
# Pinecone
PINECONE_API_KEY=
PINECONE_INDEX_NAME=
PINE_HOST_URL=
# Gemini AI
GOOGLE_API_KEY=
# Stripe
STRIPE_API_KEY=
STRIPE_PUBLISHABLE_KEY=
STRIPE_WEBHOOK_SECRET=
NEXT_PUBLIC_URL=npm run devThis will start the app on http://localhost:3000.
1️⃣ Upload a PDF → Your document is securely stored in AWS S3.
2️⃣ Chunking & Vectorization → The PDF is split into smaller chunks and embedded using Google's text-embedding-004 model.
3️⃣ Indexing in Pinecone → The vectorized data is stored in Pinecone DB for efficient retrieval.
4️⃣ Query Processing → When a user asks a question, the query is also vectorized and matched against stored data.
5️⃣ Gemini AI Response → The most relevant data is sent to Gemini AI, ensuring accurate, context-aware answers.
We welcome contributions! If you’d like to improve Savant, feel free to:
- Fork the repository
- Create a new branch
- Submit a Pull Request
This project is open-source under the MIT License.
💡 Savant – The Future of AI-Driven Document Research! 🚀