How gaurav9508/pdfintellect works
Project type and purpose
• PDFIntellect is a full-stack, AI-powered PDF note-taking web app.
• Frontend: Next.js (App Router) + TailwindCSS + ShadCN UI
• Backend: Convex (hosted database + serverless functions)
• Auth: Clerk
• AI: Google Gemini Pro API
• Payments: PayPal
• Purpose: Upload PDFs, extract/summarize content via AI, create/save notes, manage user plans.
File structure
• app/ – Next.js routes (auth, dashboard, workspace, global layout & providers)
• app/(auth)/sign-in – Clerk sign-in integration
• app/api/pdf-loader/route.js – Next.js serverless API to parse PDF
• app/dashboard/* – user dashboard pages/components (upload dialog, header, sidebar)
• app/workspace/[fileId]/* – per-PDF workspace with PdfViewer, TextEditor, Header
• convex/ – Convex backend:
• schema.js – data models
• fileStorage.js – upload URL generation & metadata mutations
• langchain/ – DB helpers for AI ingestion
• myAction.js (ingest) – calls Gemini, writes summary to DB
• user.js, notes.js – user & note management mutations/queries
• _generated/ – Convex-generated type definitions & API bindings
• configs/AIModel.js – AI model configuration
• lib/utils.js, middleware.js – shared utilities & Next.js middleware
• components/ui – reusable UI primitives (Button, Dialog, Input, etc.)
• next.config.mjs, postcss.config.mjs, jsconfig.json – build/deployment configs
Extract insights from README
• Key workflows: PDF upload → metadata stored via Convex → text extraction via /api/pdf-loader → summary via Convex+Gemini → notes stored in Convex DB
• Auth via Clerk with auto user creation in Convex
• Plan upgrade via PayPalButtons → Convex mutation
a. Main components:
• Client (Browser/Next.js React app)
• Next.js serverless layer (SSR pages + /api/pdf-loader route)
• Convex backend (DB + serverless functions)
• Clerk (Auth service)
• Google Gemini Pro API (AI service)
• PayPal (Payment service)
• File storage (Convex fileStorage or underlying S3)
b. Relationships & data flow:
1. User → Next.js pages (protected by Clerk middleware)
2. Next.js ↔ Clerk for auth tokens & session
3. Upload PDF:
• Frontend calls Convex fileStorage.generateUploadUrl
• Convex returns presigned URL → frontend uploads PDF to storage
• Convex mutation fileStorage.AddFileEntryToDb stores metadata
4. Extract & Summarize:
• Frontend calls GET /api/pdf-loader?pdfUrl=… → Next.js parses PDF client or server-side
• Next.js returns raw text → frontend invokes Convex action myAction.ingest(text, fileId)
• Convex function fetches Gemini API → stores summary & notes in DB
5. Read data: workspace pages query Convex for summary & note records
6. Plan upgrade: frontend embeds PayPalButtons → on success calls Convex user.userUpgradePlan
c. Architectural patterns & principles:
• Serverless functions for both frontend (Next.js API routes) and backend (Convex actions)
• Clear separation: UI (React components) vs. business logic (Convex functions) vs. data storage
• Event-driven (Convex actions trigger on client requests)
• Secure API keys via Convex env or .env files, no hardcoding
d. Key technologies:
• Next.js App Router (React, SSR/SSG, API routes)
• Convex SDK (db, actions, fileStorage)
• Clerk SDK (React auth hooks, middleware)
• Google Gemini Pro (fetch in Convex actions)
• PayPal JS SDK