Introduction

Figure 1. ChemEnzyRetroPlanner Agent introduction.

Agent Interface Deploy

The ChemEnzyRetroPlanner Agent is a fully open-source service powered by llama3.1. Due to the high computational requirements, we have not directly integrated this service into the ChemEnzyRetroPlanner website. Instead, we encourage users to deploy llama3.1 via Ollama. The ChemEnzyRetroPlanner Agent can utilize a self-deployed Ollama service along with ChemEnzyRetroPlanner's free API for inference. Before using this Agent service, you need to deploy it on your personal computer. The deployment steps are as follows:

git clone https://github.com/wangxr0526/ChemEnzyRetroPlanner_agent.git
cd ChemEnzyRetroPlanner_agent
conda env create -f envs.yml
conda activate retro_planner_agent_env
pip install -r requirements.txt
pip install chemprice
pip install -e ./agent
cd streamlit_app
streamlit run app.pyCopy

Deploying Ollama is straightforward. If your local computer can support running llama3.1:70b, you can follow the steps below to deploy it directly:

# Ubuntu 20.04:
curl -fsSL https://ollama.com/install.sh | sh
ollama pull llama3.1:70b
# Ollama Base URL: http://localhost:11434
# ChemEnzyRetroPlanner Base URL: http://c2400755y0.goho.co:57669Copy

If your Ollama service is running on another server, you need to enable external access for the Ollama API and map the port (e.g., bind a domain name). The simplest approach is to forward port 11434 on the server running Ollama to localhost:11434. You can use the following SSH command to forward the port to the computer where the Agent is running:

ssh -L 11434:localhost:11434 user@ip
# Ollama Base URL: http://localhost:11434Copy

Then, specify the base URL for Ollama and the base URL for ChemEnzyRetroPlanner, and you can run the Agent. If everything is done, click here.