LLM-Guided Chemical Process Optimization

A Multi-Agent Framework for Autonomous Process Constraint Generation and Optimization

Overview

This project presents a novel approach to chemical process optimization using large language models (LLMs) integrated within a multi-agent architecture. Each agent is assigned a specialized role—such as constraint generation, parameter suggestion, simulation, and validation—to collaboratively explore and optimize steady-state process conditions. The system is built on top of IDAES for high-fidelity process modeling.

Features

• Chemical Process Simulation: Built on IDAES-PSE for robust process modeling
• AI-Driven Optimization: LLM-powered agents analyze and optimize process parameters
• Multi-Agent Collaboration: AutoGen framework enables collaborative problem-solving

Prerequisites

• Python 3.11
• Conda package manager
• Access to OpenAI API

Installation

1. Clone the Repository

git clone https://github.com/tongzeng24/ProcessAgent.git
cd ProcessAgent

2. Create and Activate Environment

conda create --yes --name ProcessAgent python=3.11
conda activate ProcessAgent

3. Install All Dependencies

pip install -r requirements.txt

4. Install IDAES Extensions

idaes get-extensions --extra petsc

Or install packages individually:

pip install idaes-pse==2.8.0
pip install autogen-agentchat==0.5.1
pip install autogen-core==0.5.1
pip install autogen-ext==0.5.1
pip install openai==1.70.0
pip install tiktoken==0.9.0
pip install pandas==2.2.3
pip install pyyaml==6.0.2
# Then run: idaes get-extensions --extra petsc

Quick Start

1. Activate the environment:

   conda activate ProcessAgent

2. Configure your LLM API keys:

   Set your environment variable:

   export OPENAI_API_KEY="your-actual-api-key-here"

   Then update the config.yaml file:

   # Edit config.yaml and add your API key to the Model section:
   # api_key: "your-actual-api-key-here"

   Note: Make sure to add your actual OpenAI API key in the api_key field under the Model section in config.yaml before running the application.

Usage

Running the Complete Pipeline

To run the entire LLM-guided chemical process optimization pipeline:

python main.py

Configuration

All system settings can be customized by editing the config.yaml file:

# Example configuration sections:
ContextAgent:
  context_sampling_iterations: 5
  
Optimization:
  optimization_metric: "cost"  # Options: "cost", "yield", "yield/cost"
  initial_params: {
    "H101_temperature": 600,
    "F101_temperature": 325,
    "F102_temperature": 375,
    "F102_deltaP": -240000
  }
...

Key Configuration Options:

• Context Agent: Adjust sampling iterations and output paths
• Optimization: Set objective function and initial process parameters
• Model: Configure LLM model and API settings

Results will be saved to the Results/ directory as specified in your configuration.