edi-maas-sdk

MaaS Platform Python SDK — HTTP client for model and dataset management.

Install

pip install edi-maas-sdk

Quick Start

from edi_maas_sdk import DatasetCategory, DatasetFormat, DatasetSplit, EdiApi

api = EdiApi(endpoint="https://maas.example.com/maas-service", api_key="edik-...")

# List models
repos = api.list_repos("models")
for r in repos:
    print(r.name, r.status)

# Create a model (display_name required; name is auto-generated)
result = api.create_repo("models", display_name="My Model", description="A fine-tuned model")

# Create a dataset (display_name required, name auto-generated; optional enums)
result = api.create_repo(
    "datasets",
    display_name="My Dataset",
    dataset_category=DatasetCategory.TEXT_CORPUS,
    format=DatasetFormat.PARQUET,
    split=DatasetSplit.TRAIN,
    tags=["nlp"],
)

# List versions
refs = api.list_repo_refs("models", org_id="acme", name=result.name)
for ref in refs:
    print(ref.id, ref.version_name)

# Browse files
files = api.list_repo_tree("models", org_id="acme", name=result.name, version=refs[0].version_name, recursive=True)
for f in files:
    print(f.type, f.path, f.size)

# Download (returns presigned URL; optional expires in seconds, max 3600)
info = api.download_file("models", org_id="acme", name=result.name, version=refs[0].version_name, path="pytorch_model.bin", expires=1800)
print(info.url, info.expires_in)

# Upload
api.upload_file("models", org_id="acme", name=result.name, file_path="./weights.bin")

# Upload a whole folder (filename = path relative to folder)
api.upload_folder("models", org_id="acme", name=result.name, local_dir="./artifacts")

Configuration

API Reference

EdiApi(endpoint=None, api_key=None)

Create a client. Reads MAAS_ENDPOINT and MAAS_SDK_API_KEY from environment if not provided.

Methods

Enums

create_repo accepts typed enums for category / format / split fields. The server validates these and rejects unknown values with HTTP 422.

License

MIT