ContextOS — The Graph Mind

A persistent context layer for Chrome that builds a living knowledge graph of your browsing — turning every AI conversation into searchable, connected memory.

ContextOS captures conversations from Claude, ChatGPT, and other AI tools as you browse, stores them in a local graph + vector database, and surfaces relevant context when you need it. Think of it as long-term memory for your AI assistant sessions.

Features

• Knowledge Graph — Entities, concepts, and relationships extracted from every AI conversation using KuzuDB
• Semantic Search — Vector-indexed memory with LanceDB for similarity-based retrieval
• Chrome Extension — Passively captures chat context from Claude.ai and ChatGPT with zero friction
• DSPy Extraction Pipeline — Kimball-style entity extraction optimized for consistency and accuracy
• Local-First — All data stays on your machine; no cloud sync required

Tech Stack

Layer	Technology
Backend	Python, FastAPI
Graph DB	KuzuDB
Vector DB	LanceDB
Extraction	DSPy (MIPROv2 optimized)
Extension	Chrome Manifest V3
Models	OpenAI GPT-4 / compatible

Getting Started

Prerequisites

• Python 3.11+
• Chrome browser

1. Server Setup

cd server
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

2. Configuration

Copy .env.example to .env in the server/ directory and fill in your keys:

cp server/.env.example server/.env

3. Start the Server

cd server
uvicorn main:app --port 8742

4. Load the Chrome Extension

1. Navigate to chrome://extensions
2. Enable Developer mode (top right toggle)
3. Click Load unpacked → select the extension/ directory

The extension will begin capturing AI conversations automatically once loaded.

Architecture

Chrome Extension (content scripts)
    ↓ HTTP POST
FastAPI Server (port 8742)
    ↓ extract entities/relationships
DSPy Pipeline → KuzuDB (graph) + LanceDB (vectors)
    ↓ query
Context API → returned to extension on demand

Project Structure

.
├── extension/          # Chrome extension (Manifest V3)
│   ├── content.js      # Page scraper for Claude/ChatGPT
│   └── popup/          # Extension UI
└── server/             # Python backend
    ├── main.py         # FastAPI app
    ├── graph/          # KuzuDB schema + queries
    ├── vectors/        # LanceDB indexing
    └── extraction/     # DSPy extraction pipeline

Status

Active development. Core extraction pipeline and Chrome extension are functional.

---

Built with DSPy, KuzuDB, and LanceDB.