The Archive — AI-Powered Second Brain

A full-stack knowledge vault for saving notes, uploading PDFs, and asking questions over your own content using semantic search and vector embeddings.

live demo link : https://naveen-cs50-cse.github.io/TheArchive-AI.Powered/

The project includes:

• User authentication with signup/login and JWT-protected API routes
• Note capture, PDF ingestion, and document chunk embedding
• Semantic search and retrieval over stored content
• Retrieval-augmented generation (RAG) responses grounded in your archived notes
• Persistent chat memory for context-aware assistant replies

• Backend: Node.js + Express
• Database: Prisma ORM with SQLite (backend/prisma/schema.prisma)
• AI Embeddings: Google Gemini text-embedding-004
• PDF extraction: unpdf
• Frontend: static HTML/CSS/JavaScript (frontend/index.html, frontend/app.js, frontend/style.css)

backend/
  package.json
  server.js
  db.js
  clean.js
  middleware/
    auth.js            # JWT auth middleware
  routes/
    auth.js            # signup/login routes
    groq_ai.js         # answer generation / AI completion logic
    routes.js          # main notes/query/pdf API routes
  services/
    gemini_ai.js       # Google Gemini embedding provider
  prisma/
    schema.prisma      # data model definitions

frontend/
  index.html           # UI and layout
  app.js               # frontend logic and API calls
  style.css            # styling

README.md

• Register and login with email/password
• Store text notes into a personal archive
• Upload PDF files, extract text, and create searchable chunks
• Search by natural language query using embeddings
• Receive assistant answers grounded in your archive
• Reset the local archive from the frontend

The Prisma schema defines:

• User — users with name, email, password
• Note — raw note and PDF content linked to a user
• Chunk — text chunks with serialized vector embeddings
• Chat — assistant and user messages for chat history

1. Open a terminal in backend/.
2. Install packages:

npm install

3. Generate Prisma client:

npx prisma generate

4. Apply Prisma migrations (if available):

npx prisma migrate deploy

5. Create a .env file in backend/.

PORT=4000
JWT_SECRET=your_jwt_secret_here
GEMINI_API_KEY=your_google_gemini_api_key

The current backend is configured to use SQLite via backend/prisma/schema.prisma.

From backend/:

npm run dev

Then open frontend/index.html in a browser, or serve the frontend/ directory with a simple static server.

• POST /auth/signup — create an account
• POST /auth/login — authenticate and receive JWT
• GET /api/notes — list current user notes
• POST /api/write — save a new text note
• POST /api/query — ask a semantic search question
• POST /api/upload-pdf — upload and process a PDF
• GET /api/reset — clear/reset stored data

1. Register or sign in from the frontend.
2. Save notes or upload PDFs.
3. Enter a search query and view the AI-assisted answer.
4. Use the “Consult Full Ledger” button to list saved notes.

• The backend stores embeddings as JSON strings in the Chunk model.
• The search route uses cosine similarity plus simple keyword boosting to retrieve relevant chunks.
• The assistant answer service combines retrieved context with recent chat history.

• Change the embedding provider in backend/services/gemini_ai.js to switch models.
• Customize prompt generation in backend/routes/groq_ai.js.
• Inspect the local database with npx prisma studio from the backend/ folder.

Built for The Archive project.