RAGMantra is a local-first application that enables users to perform summarization, semantic search, and question answering using Local LLMs — ensuring complete data privacy and customizability.
Modern LLMs like ChatGPT often store and use your prompts for future training. If you're working on sensitive, novel, or proprietary content — this can be a serious data privacy risk.
Risk of intellectual property leakage
Lack of transparency in how prompts are stored or used
Inability to customize models for personal or enterprise use
🖥️ Runs completely locally — no cloud API calls
🔐 Guarantees no prompt data leaks
🔄 Easily customizable and extendable for your own RAG workflows
Secure enterprise document chatbots
Research summarization
Local knowledge base assistants
Offline exam prep tools
- Clone the repository.
git clone https://github.com/Joyee2004/RagMantra.git
cd RagMantra
- Set up the backend.
uv venv --python 3.12.0
uv is fast and efficient package manager. Refer https://docs.astral.sh/uv/ to learn more. To install uv, simply run :
pip install uv
.venv/Scripts/activate
uv pip install -r requirements.txt
cd .\backend\
cd .\models\
uvicorn main:app --reload
- Set up frontend on another terminal.
cd RagMantra
cd .\rag-client\
npm install
npm run dev
If frontend takes lot of time to render, then follow the below steps:
npm run build
npx serve -s dist -l 5173
Click on the link http://localhost:5173/ on the terminal. And you are ready to use RagMantra!!
RAGMantra runs LLMs entirely locally using Ollama, a robust and efficient tool to host state-of-the-art models on your own machine.
Steps to set up Ollama:
-
Download ollama from https://ollama.com/
-
On your terminal , run
ollama pull cas/nous-hermes-2-mistral-7b-dpo:latestNote : Download model as per your task and hardware requirements. -
To see the list of downloaded models:
ollama list
Here's a short demonstration of our project Insert gif or link to demo
Harshit Gangwar : https://github.com/harshit-G51102
Joyee Biswas : https://github.com/Joyee2004