About this project
Larry is an intelligent document assistant that helps you find information in your PDF files. Unlike other tools that often make up answers, Larry only responds when the information is actually available in your documents. If it can't find what you're looking for, it tells you.
Creator
Created by Imanol Maiztegui • imanol.mzd@gmail.com
Source Code
The source code for this project is available on GitHub
Technologies
Larry uses S3 storage, a message queue, and a worker to process PDFs into vector embeddings. Questions are answered using vector search and an external LLM.
| Frontend | Dev | Prod |
|---|---|---|
| Web Application | Next.js + TypeScript + pnpm | Vercel |
| Backend | Dev | Prod |
|---|---|---|
| API Server | Node.js + TypeScript + pnpm | AWS EC2 (Docker) |
| File Storage | MinIO | AWS S3 |
| Message Queue | LocalStack SQS | AWS SQS |
| Background Worker | Node.js + TypeScript + pnpm | AWS EC2 (Docker) |
| Real-time Updates | Redis | Upstash Redis |
| Database | PostgreSQL + pgvector | NeonDB + pgvector |
Additional tools: Prisma (ORM), WebSockets,OpenAI text-embedding-3-small, and OpenAI gpt-4o-mini.
AI Tools
- ChatGPT - Architecture, Design and Copywriting
- Claude on Cursor - Detailed features
Next Incoming Features
- Stripe integration - Charge per question asked / File uploaded
- Google Login
- Support for .xlsx, .csv, .docx files
- PDF reader when file clicked, with source highlighted
- Improved styling and copywriting
Test PDFs
You can use the following test PDFs to try Larry: