modelpark

ModelPark

ModelPark provides a versatile platform to share and manage your ML models directly from your machine, offering a convenient Python API to manage these tasks programmatically, including controlling access and publishing applications.

This library provides a more Pythonic way of managing your applications with ModelPark compared to using the CLI directly.

See ModelPark website and platform for more details.

Features

• Share models directly from the Python API.
• Publish and manage applications using the ModelPark Python API.
• Configure access management according to your needs through Python methods.

Installation

To install ModelPark, you can use pip:

pip install modelpark

Configuration

Ensure Python and pip are installed on your machine. This API interfaces with the ModelPark CLI but manages interactions programmatically through Python.

Usage

Here's how you can use the ModelPark Python package:

Initialize and Login

from modelpark import ModelPark

mp = ModelPark() # downloads the modelpark CLI binary/ executable to your home folder as "~/modelpark'
mp.login(username="your_username", password="your_password")
mp.init()

clear cache while init (remove existing modelpark CLI binaries from system)

from modelpark import ModelPark

mp = ModelPark(clear_cache=True)

Register an Application

Register an app running on a certain port

mp.register(port=3000, name="my-app", access="public") 
# access='private' if private (not visible/ accessible in modelpark dashboard)

Register a password protected app running on a certain port

mp.register_port(port=3000, name="my-app", access="public", password='123')

Register an app running on a certain port

mp.register_port(port=3000, name="my-app", access="public")

Register a streamlit app that is not run yet (this starts the app as well)

mp.run_with_streamlit_and_register(port=3000, name="my-app", file_path="~/my-app/streamlit-app.py", access="public", framework="streamlit")
# generic registration also works >> 
# mp.register(port=3000, name="my-app", file_path="~/my-app/streamlit-app.py", access="public", framework="streamlit")

Register a streamlit app that is not run yet

mp.register(port=3000, name="my-app", file_path="~/my-app/streamlit-app.py", access="public", framework="streamlit")

Register a Fast API app while deploying

add register_port within startup_event() function in FAST API app

@app.on_event("startup")
async def startup_event():
    mp.register_port(port=5000, name="my-fast-api", access="public") 

List Registered Applications

mp.ls()
# or mp.status()

Make an API Call to a Registered Application

from modelpark import APIManager
mp_api = APIManager()

user_credentials = {'username': 'your_username', 'password': 'your_password'}
app_name = 'my-app'
payload = {'key': 'value'}  # Payload required by the application

# Make the API call
response = mp_api.make_api_call(app_name, user_credentials, payload)
print(response.json())  # Assuming the response is in JSON format

Stop and Logout

mp.stop()
mp.logout()

Kill an Application

mp.kill(name="my-app")

Kill all the registrations in this session

mp.kill(all=True)