A self-hosted, air-gapped AI workbench for confidential documents.
01
GenIQ is a self-hosted AI workbench designed so that confidential documents never have to leave the organisation's own machines. Instead of calling a cloud AI service, it runs a language model locally and lets a user work with their documents through a chat interface.
It was built for Smart India Hackathon problem statement SIH26117.
02
Organisations that handle sensitive material, such as government, PSU and defence-related bodies, often cannot send documents to external AI services.
The objective was to show that useful document-aware AI assistance can run entirely on local infrastructure, in an air-gapped style deployment.
03
A React and Vite front end lets a user upload documents and chat with an assistant. A FastAPI backend written in Python handles uploads, extracts text from PDFs using pypdf, and sends prompts to a locally running Ollama server hosting the Llama 3.2:3b model.
Because every component runs locally, no document content needs to reach an external service.
04
Team Leader & Project Developer. This was a team project. I contributed to development activities involving React/Vite, FastAPI, Ollama integration and PDF processing.
05
06
React + Vite UI
Upload and chat interface
FastAPI backend
Python API layer
pypdf
PDF text extraction
Ollama
Local model runtime
Llama 3.2:3b
On-device language model
07
Screenshots have not been added yet.
08
No public demo or repository link is available for this project yet.
09
Whether you're hiring, planning a website or exploring a collaboration, I'd like to hear about it.