Walid BENMAAROUF
Back to projects
React (Hooks, Context)Streaming Fetch APIReact-Markdown + GFMjsPDF

PaperMind

Chat with your PDFs using grounded RAG.

Live demo GitHub
OverviewDemoGalleryTechDetails

Overview

PaperMind is an intelligent application that lets users chat with their PDF documents.

Live demo: papermind-frontend.onrender.com

Instead of manually reading long research papers or technical documents, users can:

  • upload one or multiple PDFs,
  • ask questions in natural language,
  • and receive precise answers grounded only in the uploaded documents.

The system understands the documents, remembers past conversations, and highlights relevant information instead of giving generic answers.

It is designed for:

  • research papers,
  • technical documentation,
  • academic work,
  • multi-PDF analysis.

In short: PaperMind turns static PDFs into an interactive knowledge base.

Architecture

React Frontend
  ↓
FastAPI Backend (Streaming)
  ↓
Async Ingestion & Embedding Pipeline
  ↓
Qdrant Vector Database + OpenAI Models

At a glance

Slugpapermind
Tech21 items
RepoOpen

Demo

Open the live demo in a new tab.

Launch demo

Tech

Frontend

5 items
React (Hooks, Context)
dynamic, stateful chat UI
Streaming Fetch API
real-time token rendering
React-Markdown + GFM
rich answers (code, tables)
jsPDF
export answers to PDF
Responsive UI
desktop & mobile support

Backend

5 items
FastAPI (Python)
async, high-performance API
Server-Sent Events (SSE)
streaming LLM responses
Subprocess orchestration
non-blocking ingestion pipeline
PyMuPDF
PDF text extraction
WebSockets
live pipeline logs

Rag Ai

4 items
OpenAI text-embedding-3-large
3072-D semantic embeddings
GPT-4o-mini
HyDE generation, reranking, summarization
Hybrid RAG
query rewriting + HyDE + LLM rerank
Context-grounded answering
no hallucinated content

Vector Database

3 items
Qdrant (Cloud)
HNSW-indexed vector search
Multi-level vectors
document / chunk / topic
Payload filtering
topic, filename, discipline

Infrastructure Design

4 items
Stateless API
horizontally scalable
Async background workers
ingestion isolated from chat
JSON-based persistence
conversations & metadata
Secure key handling
server-side only

Details

Project info

Repositoryhttps://github.com/walid7shind

Next steps

  • • Enhance processing speed for real time use
  • • Implement on embedded electronic system
Back to projects