IRIS

Overview

IRIS(Intent Revelation and Interaction System) is a browser-based research system that explores intelligent workflow automation through the Belief-Desire-Intention (BDI) model. It aims to facilitate seamless collaboration between users and AI systems by understanding and processing implicit user intentions through a novel approach to intent revelation and interaction.

Core Features

Content Collection

• Smart text selection and preservation
• Webpage screenshot functionality
• Contextual information capture
• User annotation support

Intent Processing

• Implicit intent extraction
• Hierarchical intent clustering
• Tree-based intent visualization
• Real-time intent updates

Smart Interaction

• Side panel quick access
• Floating tool window
• Content highlighting
• Cross-page content association

Technical Architecture

Frontend (Chrome Extension)

• Content Scripts
• Background Service
• Popup Interface
• Side Panel
• Network Visualization

Backend (FastAPI)

• Vector Embedding Service
• Intent Clustering Analysis
• RAG (Retrieval-Augmented Generation)
• API Services

System Requirements

Frontend Dependencies

• Chrome Browser
• Node.js environment

Backend Dependencies

• Python 3.8+
• FastAPI
• OpenAI API key
• Required Python packages (see requirements.txt)

Quick Start

1. Clone the repository

git clone https://github.com/jiaqi-xiao/MagicPocket.git

2. Setup Backend Environment

cd Back

# Create conda environment from freeze.yml
conda env create -f freeze.yml

# Activate the environment
conda activate HCIX

# Start the backend server
uvicorn main:app --reload --port 8000

3. Setup Frontend (Chrome Extension)

1. Open Chrome extensions page

   • Navigate to chrome://extensions/
   • Enable "Developer mode" in the top right

2. Load the extension

   • Click "Load unpacked"
   • Select the Front directory from the project

4. Configure the Extension

1. Open extension options
2. Set up required API keys
3. Configure any additional settings

Development Setup

Backend Requirements

• Anaconda or Miniconda
• Python 3.10
• FastAPI
• OpenAI API key
• Other dependencies as specified in freeze.yml

Frontend Requirements

• Chrome Browser
• Node.js (for development)

Future Work

• Advanced intent extraction algorithms
• Enhanced visualization techniques

License

This project is licensed under the MIT License