Phys-Agent Usage Guide

Overview

The Phys-Agent is a physics research assistant MCP server for the TSMC CareerHack project. It enables AI assistants to search ArXiv, ingest physics papers, and answer questions with citations.

---

Installation

# Install all dependencies
pip install -r requirements.txt

---

Running the Server

Option 1: Direct Execution (stdio mode for VS Code)

python mcp_server.py

Option 2: Test without MCP

python test_run.py

---

VS Code / Antigravity MCP Configuration

Add the following to your .vscode/mcp.json or MCP settings file:

{
  "servers": {
    "physics-agent": {
      "command": "python",
      "args": ["/home/itzu/phys-expert/mcp_server.py"]
    }
  }
}

Or for the Antigravity settings format:

{
  "physics-agent": {
    "command": "python",
    "args": ["/home/itzu/phys-expert/mcp_server.py"]
  }
}

---

Available MCP Tools

Tool	Description
add_knowledge_topic(topic)	Downloads and studies physics papers from ArXiv on a topic
consult_physics_expert(question)	Queries the knowledge base with formatted citations
verify_source(paper_id)	Retrieves full paper details for citation verification
get_knowledge_stats()	Returns database statistics

---

Example Usage in AI Assistant

User: How does Lambertian reflectance affect lighting estimation?

AI: [Calls add_knowledge_topic("Lambertian reflectance lighting")]
    [Calls consult_physics_expert("Lambertian reflectance lighting estimation")]
    
    Based on the physics literature, Lambertian reflectance...
    
    Sources:
    [1] Paper Title, Page 5 (arXiv:2xxx.xxxxx)

---

Project Structure

phys-expert/
├── physics_knowledge_db.py   # Core knowledge base class
├── mcp_server.py             # MCP server interface
├── test_run.py               # Test script
├── requirements.txt          # Dependencies
├── README.md                 # This file
└── db/                       # ChromaDB persistent storage