Python Interface for MLC CHAT CLI

This simple code allows you to interact with the MLC LLM using the official CLI (command-line interface) app using python api.

Install

First of all, You should install MLC LLM follwing official instruction, and are able to run mlc_chat_cli in the terminal. After that, you can feel free to copy this python code to your project or directly install by running:

python setup.py install

Usage

This code provide several python api to interact with MLC LLM, take string as input and get response string. It is recommand to run this code under the mlc-llm root with a dist folder that contains libary files. Or you can specify the path of dist folder manually.

>>> from mlc_chatbot import ChatBot
>>> bot = ChatBot()
"""
you can also specify the folder of the library and executable path by:
>>> bot = ChatBot(dist_url='{PATH_TO_DIST}', cli_dir='{PATH_TO_CLI}')
for example, a Windows user:
>>> bot = ChatBot(dist_url='~/mlc-chat/dist', cli_dir='~/miniconda3/envs/mlc-chat/Library/bin/mlc_chat_cli.exe')
"""
# now chat with mlc llm using python interface!
>>> bot.send('hello!')
>>> 'Hello! How can I help you today?'
# restart a fresh chat 
>>> bot.reset()
# get status of llm
>>> bot.status()
# encode speed and decode speed (token/s)
>>> (39.4, 26.0)