MLCore Python Client

A powerful Python client for MLCore, providing a database-like interface for managing datasets and machine learning models. Built for data scientists and engineers who want to automate their machine learning workflows.

Links

• GitHub Repository: Amanbig/MLCore
• Docker Hub: procoder588/mlcore

Installation

pip install mlcore-client

Connection

Connect to your MLCore instance using a connection string (URI) or explicit parameters.

from mlcore import MLCore

# Option 1: Connection string (Recommended)
# Format: mlcore://user:password@host:port
client = MLCore("mlcore://admin:password@localhost:8000")

# Option 2: Explicit parameters
client = MLCore(
    host="localhost", 
    port=8000, 
    email="admin@example.com", 
    password="password"
)

Asynchronous Support

For asyncio applications, use MLCoreAsync:

from mlcore import MLCoreAsync

client = MLCoreAsync("mlcore://admin:password@localhost:8000")
await client.connect()
# ... use await with all methods ...
await client.close()

API Documentation

📊 Dataset Manager (client.datasets)

Manage data lifecycle from raw files to cleaned versions.

🤖 Model Manager (client.models)

Train, evaluate, and deploy machine learning models.

📈 General Methods

Usage Examples

End-to-End Pipeline

# 1. Prepare Data
file_info = client.datasets.upload_file("raw_data.csv")
ds = client.datasets.create(name="Training Set", file_id=file_info["id"], ...)
cleaned_ds = client.datasets.clean(ds["id"], strategy="drop_nulls")

# 2. Train Model
model = client.models.train(
    dataset_id=cleaned_ds["id"],
    algorithm="random_forest",
    target_column="label",
    hyperparameters={"n_estimators": 200}
)

# 3. Monitor Specs
print(f"Model trained with {model['accuracy']}% accuracy")

# 4. Predict
result = client.models.predict(model["id"], inputs={"feature_1": 0.5, "feature_2": 1.2})
print(f"Prediction: {result['predictions']}")

Features

• Database-like Connection: URI-based connection strings (mlcore://).
• Session Persistence: Automatic token management and re-auth logic.
• Pandas Ready: One-click conversion from server data to DataFrames.
• Async First: First-class support for httpx-based asynchronous I/O.
• Developer Friendly: Strictly typed and linted with Ruff.

License

MIT