bailo

Bailo Python Client

A simple Python API Wrapper for Bailo

Table of Contents

• Key Features
• Installing
• Getting Started
• Documentation
  • Building Locally
• Development
  • Install and add precommit
  • Install the package locally
  • Testing

Key Features

• Uploading and downloading model binaries
• Managing Models and Releases
• Managing Datacards
• Managing Schemas
• Managing Access Requests

The Bailo Python client aims to programmatically cover Bailo's core functionality by interacting with the endpoints in the backend. The functionality covered is that which a Data Scientist, Software Engineer or other similarly technical role might be expected to utilise, meaning that it does not have complete coverage of all endpoints, such as those relating to the discussion & approval of reviews & access requests. For these interactions, the web frontend is expected to be used.

Installing

[!IMPORTANT] Python 3.9 or higher is required

pip install bailo

Getting Started

from bailo import Client, Model
client = Client("http://localhost:8080")

# Create a model
yolo = Model.create(
    client=client,
    name="YoloV4",
    description="You only look once!"
)

yolo.card_from_schema("minimal-general-v10")

# Create a new release
my_release = yolo.create_release(version="0.1.0",
                                 notes="Beta")

# Upload a file to the release
with open("yolo.onnx") as f:
    my_release.upload("yolo", f)

Documentation

Documentation is rendered with Sphinx and served here.

Building locally

Refer to backend/docs/README.md for local build steps.

Development

Install and add precommit

If already working on Bailo you may be prompted to overwrite Husky. Follow the instructions given by Git CLI.

pip install pre-commit
pre-commit install

Install the package locally

pip install -e .

Testing

The package uses Pytest to test packages. Tests can be ran accordingly from within this directory. Tests are split into categories sections for automation purposes.

In order to run integration tests make sure Bailo is running on https://localhost:8080:

pytest -m integration

To run the mlflow tests, make sure that Bailo is running as above and mlflow is running on port 5050 e.g. mlflow server --host 127.0.0.1 --port 5050:

pytest -m mlflow

Run all other tests:

pytest