Model Parser Agent

Introduction

This is python AI agent that will use model_parser_mcp tool at MCP_URL.

Pre-requisite

1. model_parser_mcp is running. Refer to section below on how to setup model_parser_mcp.
2. google adk is install https://google.github.io/adk-docs/get-started/python/#installation
3. Create a .env file. This file contain mandatory environment variables.
   • GOOGLE_API_KEY Google Gemini API key

Setting Up model_parser_mcp

Pre-requisite

Create a .env file. This file contain mandatory environment variables. Configure

• DATABASE_URL=postgres://postgres:admin@localhost/modelruntime //Targeting DB that contained savedModel
• CACHE_SIZE=5 //Cache size

Docker-compose

Create a .env file and run the docker compose file as below:

version: '3'
services:
  model_query:
    image: "jsfong/model-parser-mcp:latest"
    ports:
      - "8001:8001"
    volumes:
      - ./.env:/.env

Release page

https://hub.docker.com/repository/docker/jsfong/model-parser-mcp/general

Configuration

Environment Variable

Var	Remarks
MCP_URL	Set url of mcp tool in agent.py (only support SSE)

Instruction

Edit the instruction variable in agent.py.

How to run

cd ../
adk web