Chalk Python SDK

2.115.4 · active · verified Mon Apr 13

Chalkpy is the Python SDK for Chalk, a feature store designed to simplify feature engineering and deployment for machine learning teams. It allows users to define feature pipelines using familiar Python functions and data structures, orchestrating them on a Rust-based engine for parallel execution. The library facilitates defining features with Pydantic-inspired classes and creating resolvers to compute them for both online inference and offline training. The current version is 2.115.4, with frequent updates indicated by its changelog.

Warnings

gotcha Authentication requires `CHALK_CLIENT_ID` and `CHALK_CLIENT_SECRET` environment variables to be set, or prior authentication through the `chalk cli login` command. Failing to do so will result in connection errors.
Fix: Set `CHALK_CLIENT_ID` and `CHALK_CLIENT_SECRET` in your environment, or run `chalk login` from the command line and follow the prompts to authenticate.
gotcha Specific functionalities require installation with 'extras'. For example, `chalkpy[runtime]` is recommended for notebook environments, and `chalkpy[chalkdf]` is needed for features utilizing Chalk DataFrames. A bare `pip install chalkpy` may not include all necessary components for certain use cases.
Fix: Install with the appropriate extras, e.g., `pip install "chalkpy[runtime]"` or `pip install "chalkpy[chalkdf]"`.
gotcha The `cache_nulls` parameter for features defaults to `True`, meaning Chalk will cache all values, including nulls. This can lead to unexpected caching behavior if not explicitly handled, as a null value will replace an existing cached value.
Fix: Be aware of the `cache_nulls` default behavior. Set `cache_nulls=False` to prevent updating null entries in the cache, or `cache_nulls="evict_nulls"` to evict entries that would have been null.

Install

pip install chalkpy Basic installation
pip install "chalkpy[runtime]" For notebook environments
pip install "chalkpy[chalkdf]" For chalkdf functionality

Imports

ChalkClient
```
from chalk.client import ChalkClient
```
features
```
from chalk import features
```
Used as a decorator to define feature sets.
online
```
from chalk import online
```
Used as a decorator to define online feature resolvers.
offline
```
from chalk import offline
```
Used as a decorator to define offline feature resolvers.
_
```
from chalk import _
```
Used to build features on feature classes, commonly imported as underscore.

Quickstart

This quickstart demonstrates how to initialize the ChalkClient and verify authentication. It requires `CHALK_CLIENT_ID` and `CHALK_CLIENT_SECRET` to be set as environment variables or for the user to have authenticated via the `chalk login` CLI command.

import os
from chalk.client import ChalkClient

# Ensure these environment variables are set or authenticate via `chalk login` CLI command
CHALK_CLIENT_ID = os.environ.get("CHALK_CLIENT_ID", "")
CHALK_CLIENT_SECRET = os.environ.get("CHALK_CLIENT_SECRET", "")

if not CHALK_CLIENT_ID or not CHALK_CLIENT_SECRET:
    print("Warning: CHALK_CLIENT_ID and CHALK_CLIENT_SECRET environment variables are not set. Authentication will likely fail. Please refer to Chalk documentation for authentication.")

try:
    client = ChalkClient(
        client_id=CHALK_CLIENT_ID,
        client_secret=CHALK_CLIENT_SECRET,
        branch='notebook' # or 'production', or a custom branch name
    )

    # Verify the setup
    whoami_response = client.whoami()
    print(f"Successfully authenticated as user ID: {whoami_response.user_id}")
except Exception as e:
    print(f"Authentication failed or API call error: {e}")

view raw JSON →