On-Premises Intelligence (OPI) - find and chat though your personal documents

Local, private Retrieval-Augmented Generation for personal data analysis.

Imagine having a smart assistant that could comb through your private collection of documents — anything from confidential contracts to specialised medical guidelines or complex tax records. By keeping everything on your own machine, OPI is designed to index those files securely and let you ask plain-English questions such as "Which clause covers early termination?" or "What's the deductible for procedure 99214?". It then attempts to surface the most relevant passages and draft a concise answer — all without sending any data to the cloud. Potential benefits:

• Privacy by default – sensitive data never leaves your firewall.
• Faster insight – cut hours of scrolling through PDFs down to seconds.
• Domain flexibility – use for legal, financial, medical, engineering and other private documents.

---

Key Features

• 100 % on-prem – your documents never leave the device.
• RAG + Smart File Search – ask questions or just search: the sidebar becomes a live file-system search as you type.
• Multimodel-ready – resaons about text and image data, powered by open source LMs and VLMs.
• One-page UI – chat, upload, and document browser all in a single responsive page.
• FastAPI + SSE – answers stream token-by-token for instant feedback.

---

🚀 Quickstart

The project is Python 3.10 or 3.11 compatible. The snippet below shows the recommended workflow with conda (feel free to use venv/pipenv if you prefer).

# 1. Create and activate an isolated environment
conda create -n opi python=3.11 -y
conda activate opi

# 2. Install the Python packages
pip install --upgrade pip wheel
pip install -r requirements.txt

# 3. Fire it up
./start_chat.sh

The helper script will:

1. start the FastAPI server on http://localhost:8000;
2. launch a tiny static file server on http://localhost:5173; and
3. open your default browser at frontend/chat.html (the single-page app).

---

Repository Layout (TL;DR)

api/                 # FastAPI application
frontend/            # Static web UI (single-page chat + upload)
opi_file_system/     # Uploaded files are stored here
src/                 # RAG pipeline, database interface, models
config.py            # All user-configurable parameters
start_chat.sh        # helper script launch backend + frontend
requirements.txt     # Python dependencies

---