# How to Vibe Code Your Own Genei (and Stop Paying for It) > AI-powered research and summarization tool - Site: https://genei.io - Category: Web app - Verdict: **Solid side project** (75/100 vibecodeable) - Estimated effort: 2-3 weeks of focused work ## Verdict You can build a fully functional personal clone of Genei as a solid side project, with the single biggest catch being the complexity of clean PDF text parsing and precise source-to-note coordinate linking. Replicating Genei for personal research is an entirely achievable weekend-to-weeks project using modern AI models and frameworks. The core value—uploading PDFs, generating summaries via LLM APIs, semantic search, and linking snippets to a notepad—relies heavily on commodity building blocks like LangChain or direct LLM SDKs, vector extensions in Postgres, and a React frontend. The primary friction points are dealing with messy academic PDF layouts where figures and tables break text flow, and building a responsive PDF canvas viewer with bidirectional link highlights. ### What you can't replicate - The original brand equity and user base - Proprietary customer acquisition channels ## What it does Genei is an AI-powered research and summarization tool designed for students, academics, and content creators to extract highlights, generate summaries, and manage references. ### Core features - PDF and web page document ingestion and text extraction - AI-generated document overviews, summaries, and keyword extraction - Multi-document semantic search and Q&A mode - Structured notepad with note-linking back to source documents - Automatic reference and citation generation - Project and folder file management workspace ## The business ### Pricing - Basic Tier: £3.99 - £9.99/mo - Pro Tier: £15.99 - £29.99/mo ### Funding $125K raised. - Seed round in August 2021 Investors: Y Combinator Founded 2019. Team size: ~5 employees. ## The hard parts - Parsing poorly formatted academic PDFs while preserving structural outlines, tables, and figures - Maintaining accurate document chunk offsets to link notepad highlights back to exact source text coordinates - Orchestrating multi-document RAG and vector queries across large folder contexts efficiently ## How to vibe code Genei ### Prerequisites - Node.js (free): Required for running the Next.js development environment and package manager. - GitHub (free): Code repository hosting and deployment integration. - Anthropic API Key (Pay-as-you-go): Provides Claude model access for document summarization, paraphrasing, and Q&A search. ### Recommended AI tools - Claude Code: Best-in-class terminal coding agent for scaffolding the full-stack Next.js application, database schema, and API integrations. - Cursor: Excellent AI code editor for fine-tuning the split-screen PDF viewer and notepad UI components. ### Stack - Frontend: Next.js with Tailwind CSS - Backend: Next.js Server Actions / API Routes - Database: Supabase (Postgres with pgvector) - Auth: better-auth - Payments: None (personal use clone) - Other: pdf-parse or Unstructured for PDF text extraction, Anthropic API for LLM reasoning ### Hosting - Vercel (Hosting the Next.js frontend and serverless API functions.): $0/mo (Hobby tier) - Supabase (Storing relational application data, user files, and vector embeddings via pgvector.): $0/mo (Free tier) ### Build guide 1. **Project Scaffolding and Database Schema** — Initialize the Next.js project with Tailwind CSS, configure Supabase with pgvector, and set up better-auth for user authentication. ``` Create a new Next.js 16 project configured with TypeScript and Tailwind CSS v4. Set up Supabase as the database provider and install better-auth. Create database tables for users, projects, documents, and notes. The documents table should store file metadata, original text content, and vector embedding columns for RAG search. The notes table should support document ID references and text highlight offsets for bidirectional note-linking. ``` 2. **PDF Ingestion and Text Extraction Pipeline** — Implement file upload handling, secure storage, and a robust parser to extract clean text blocks and outlines from uploaded PDF documents. ``` Build a document ingestion pipeline in Next.js that accepts PDF file uploads. Store raw files in Supabase Storage. Use a robust text extraction library (such as pdf-parse or unpdf) to extract all text content from the PDF. Write server actions to chunk the extracted text, generate vector embeddings using an embedding model or LLM API, and save the chunks to the database with corresponding page numbers and character offsets. ``` 3. **AI Document Overview and Summarization Engine** — Integrate the Anthropic API to automatically generate document summaries, extract keywords, and construct structural outlines upon file upload. ``` Create an AI processing service using the Anthropic API (Claude 3.5 Sonnet) that analyzes uploaded document text chunks. Upon finishing upload and extraction, trigger an asynchronous background task that calls the LLM to generate: (1) a concise document summary overview, (2) a list of extracted keywords and key phrases, and (3) a hierarchical document outline with section titles. Store these structured outputs in the documents table and create a UI tab to display them. ``` 4. **Split-Screen Workspace and PDF Viewer** — Build a multi-pane research workspace featuring a PDF viewer on one side and an interactive notepad on the other. ``` Design a split-screen research workspace page in Next.js using Tailwind CSS. The left side should feature a PDF document viewer (using react-pdf or a custom canvas viewer) that renders document pages and supports text selection. The right side should contain a structured notepad interface allowing users to create notes, organize thoughts into headings, and drag-and-drop or click-to-add AI summaries and highlights directly into their notes. ``` 5. **Semantic Search and Q&A Mode** — Implement a query-based search interface allowing users to ask questions across individual documents or entire project folders using vector search and LLM synthesis. ``` Build a query-based search and Q&A interface for project folders and individual documents. Implement vector similarity search using pgvector in Supabase to retrieve the most relevant text chunks matching the user's prompt. Send the retrieved chunks along with the user question to the Anthropic API to synthesize an accurate answer with inline citations pointing back to source document titles and page numbers. ``` 6. **Reference Generation and Export Functionality** — Add automated citation formatting and export tools so users can copy references or export their notepad contents. ``` Implement a reference generation module that parses document metadata (title, author, year) to automatically format citations in standard academic styles (APA, MLA, Harvard). Add an export functionality allowing users to download their project notes, summaries, and bibliographies as Markdown, PDF, or plain text files. ``` ### Cost vs paying **Starting costs (one-time):** - Domain name (optional): $12/yr - AI coding tool subscription: $20 - Total: ~$32 one-time **Ongoing costs (monthly):** - Anthropic API token usage: ~$5 - $15/mo - Vercel & Supabase hosting: $0/mo - Total: ~$10/mo - Paying for the SaaS instead: £15.99 - £29.99/mo - Build time: 25-35 hours - AI tool credits: $20 one-time (Claude Pro) - Break-even: 1 month ## Sources - [Genei Official Website](https://genei.io) - [Tracxn - Genei Company Profile](https://tracxn.com)