https://github.com/vimalyad/doc-buddy
DocBuddy is a full-stack, production-ready Retrieval-Augmented Generation (RAG) application. Designed as a personalized version of Google's NotebookLM, it allows users to upload their own documents (PDFs, TXTs, CSVs) and have highly grounded, context-aware conversations with them.
https://github.com/vimalyad/doc-buddy
Last synced: 15 days ago
JSON representation
DocBuddy is a full-stack, production-ready Retrieval-Augmented Generation (RAG) application. Designed as a personalized version of Google's NotebookLM, it allows users to upload their own documents (PDFs, TXTs, CSVs) and have highly grounded, context-aware conversations with them.
- Host: GitHub
- URL: https://github.com/vimalyad/doc-buddy
- Owner: vimalyad
- Created: 2026-05-06T14:35:51.000Z (about 2 months ago)
- Default Branch: main
- Last Pushed: 2026-05-06T19:40:29.000Z (about 2 months ago)
- Last Synced: 2026-05-06T19:40:34.475Z (about 2 months ago)
- Language: TypeScript
- Homepage:
- Size: 219 KB
- Stars: 0
- Watchers: 0
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# DocBuddy
DocBuddy is a full-stack RAG (Retrieval-Augmented Generation) application designed to mirror the Google NotebookLM experience. It allows users to upload documents and have grounded, citation-rich conversations with their data.
## ✨ Key Features
- **Intelligent RAG Pipeline**: End-to-end document processing from ingestion to grounded answer generation.
- **Interactive Citations**: Hover over AI-generated citations (e.g., `[1]`, `p. 3`) to see the exact text snippet retrieved from your document.
- **Multi-Format Support**: Seamlessly parses and indexes **PDF, TXT, and CSV** files.
- **Smart Query Rewriting**: Automatically optimizes user questions into descriptive semantic search queries for higher retrieval accuracy.
- **Premium UI/UX**: A responsive, dark-mode dual-pane interface with auto-focusing chat and floating toast notifications.
- **Zero-Persistence Option**: Configure the backend to wipe state on restart or persist data for production.
## 🧠 The RAG Pipeline
### 1. Ingestion & Chunking
DocBuddy uses a **Recursive Character Splitting** strategy:
- **Chunk Size**: 2,000 characters
- **Overlap**: 400 characters
- **Strategy**: The splitter recursively tries to break text at double newlines, single newlines, and spaces. This ensures that paragraphs and sentences are kept together, providing the LLM with coherent context.
### 2. Embedding & Storage
- **Model**: `sentence-transformers/all-MiniLM-L6-v2` (via Hugging Face Inference API).
- **Vector Store**: **Qdrant Cloud**. We use 384-dimensional vectors with Cosine similarity to find the most relevant document chunks.
### 3. Retrieval & Generation
- **Retrieval**: The system fetches the top 5 most relevant chunks for every query.
- **LLM**: Powered by **Llama 3** (via Groq API) for lightning-fast, high-quality reasoning.
- **Groundedness**: A strict system prompt ensures the AI only answers based on the provided context and cites its sources using bracketed markers.
## 🛠 Tech Stack
- **Frontend**: React, TypeScript, Vite, Tailwind CSS, Lucide React.
- **Backend**: Node.js, Express, TypeScript, LangChain.js.
- **Database**: Qdrant Cloud (Vector), Local JSON (Metadata).
## ⚙️ Setup & Installation
### Backend
1. `cd backend`
2. `npm install`
3. Create a `.env` file with:
```env
GROQ_API_KEY=your_key
HUGGINGFACEHUB_API_TOKEN=your_token
QDRANT_URL=your_qdrant_url
QDRANT_API_KEY=your_qdrant_key
PORT=5000
```
4. `npm run build && npm start`
### Frontend
1. `cd frontend`
2. `npm install`
3. Create a `.env` file with:
```env
VITE_API_URL=http://localhost:5000
```
4. `npm run dev`
---