Braintrust Core

0.0.59 · active · verified Sat Apr 11

Braintrust Core is a Python package that provides shared core dependencies for other packages within the Braintrust ecosystem. It is primarily an internal utility library, with user-facing functionality for logging, tracing, evaluations, and CLI workflows typically exposed through the main `braintrust` SDK. The current version is 0.0.59, and it generally follows the release cadence of the broader Braintrust Python SDKs.

Warnings

gotcha The `braintrust-core` package is primarily an internal dependency. For Braintrust functionality like logging, tracing, evaluations, or CLI interactions, users should typically install and use the main `braintrust` package (e.g., `pip install braintrust autoevals`).
Fix: Use `pip install braintrust` and import from `braintrust` for Braintrust features. Only install `braintrust-core` if explicitly required as a dependency by another Braintrust package.
gotcha Functionality that might appear in a 'core' package can be subject to refactoring and consolidation into the main `braintrust` SDK. For instance, the `@braintrust/core` (TypeScript) package's functionality was explicitly folded into the main `braintrust` package. While this was for TypeScript, it indicates a pattern that Python users should be aware of regarding `braintrust-core`.
Fix: Always refer to the official Braintrust Python SDK documentation for the latest recommended import paths and usage patterns, focusing on the `braintrust` package for user-facing APIs.
deprecated Direct imports or reliance on specific modules within `braintrust-core` are not officially supported APIs and may change or be removed without prior notice, as it's an internal utility package.
Fix: Avoid directly importing from `braintrust-core`. If you find yourself needing functionality seemingly offered by `braintrust-core`, check the main `braintrust` SDK documentation first, as the desired feature is likely available there.

Install

pip install braintrust-core Install Braintrust Core

Imports

Eval
```
from braintrust import Eval
```
For core Braintrust evaluation functionality, import from the main `braintrust` package, not `braintrust-core`.
init
```
from braintrust import init
```
For initializing Braintrust logging and experiments, import from the main `braintrust` package.

Quickstart

This quickstart demonstrates the usage of the main `braintrust` SDK for defining and running evaluations. `braintrust-core` is an internal dependency and not typically used directly in user quickstarts. Ensure you also install `autoevals` for scoring functions: `pip install braintrust autoevals`.

import os
from autoevals import LevenshteinScorer
from braintrust import Eval

# Set your Braintrust API key from environment variables
# BRAINTRUST_API_KEY='sk-...' (Replace with your actual key or set in environment)
api_key = os.environ.get('BRAINTRUST_API_KEY', 'YOUR_API_KEY_HERE')

# Define an evaluation
eval_run = Eval(
    "Say Hi Bot",
    data=lambda: [
        {"input": "Foo", "expected": "Hi Foo"},
        {"input": "Bar", "expected": "Hello Bar"},
    ],
    task=lambda input: "Hi " + input,
    scores=[LevenshteinScorer],
    api_key=api_key # Pass API key here or ensure BRAINTRUST_API_KEY env var is set
)

# Run the evaluation
# To run this, you would typically execute it via the Braintrust CLI:
# BRAINTRUST_API_KEY=YOUR_API_KEY braintrust eval your_script_name.py
# Or, to run directly in Python, if your API key is configured:
# result = eval_run.run() 
# print(result)

print("To run this evaluation, set BRAINTRUST_API_KEY and use the Braintrust CLI: braintrust eval <your_script.py>")

view raw JSON →