braintrust-langchain

0.3.0 · deprecated · verified Sun Apr 12

This library provided LangChain integration for Braintrust, a platform for evaluating and managing LLMs. It is now deprecated as of version 0.3.0. The LangChain integration has been moved into the main `braintrust` package. Users should install and use `braintrust` instead, which is actively developed with frequent releases.

Warnings

breaking This `braintrust-langchain` package is deprecated. Its functionality has been migrated to the main `braintrust` SDK.
Fix: Uninstall `braintrust-langchain` (`pip uninstall braintrust-langchain`) and `pip install braintrust` instead. All future updates and features will be in the main `braintrust` package.
breaking LangChain integration imports have changed. Direct imports from `braintrust_langchain` are no longer valid.
Fix: Update imports to use `from braintrust import auto_instrument` for global instrumentation, or `from braintrust.langchain import BraintrustCallbackHandler` for manual callback handling. Refer to the `braintrust` SDK documentation for the latest usage.
gotcha Braintrust requires an API key for authentication. If not provided, tracing and logging will not function.
Fix: Ensure the `BRAINTRUST_API_KEY` environment variable is set before running your application. Alternatively, pass the `api_key` argument directly to `braintrust.init_logger()` or `braintrust.set_api_key()`.

Install

pip install braintrust Install the replacement package
pip uninstall braintrust-langchain Uninstall the deprecated package

Imports

auto_instrument
```
from braintrust import auto_instrument
```
The global handler for LangChain is now part of the main `braintrust` package via `auto_instrument()`.
BraintrustCallbackHandler
```
from braintrust.langchain import BraintrustCallbackHandler
```
The `BraintrustCallbackHandler` class has moved from `braintrust_langchain` to `braintrust.langchain`.

Quickstart

This quickstart demonstrates how to use Braintrust's LangChain integration with the main `braintrust` package. It initializes the Braintrust logger, enables automatic LangChain instrumentation, and then executes a simple LangChain example. All interactions will be automatically logged to your Braintrust project.

import os
from braintrust import init_logger, auto_instrument
from langchain_core.prompts import ChatPromptTemplate
from langchain_openai import ChatOpenAI

# Ensure your Braintrust API key is set
os.environ['BRAINTRUST_API_KEY'] = os.environ.get('BRAINTRUST_API_KEY', 'YOUR_BRAINTRUST_API_KEY')

# Initialize the logger with your project name (required)
# and auto-instrument LangChain
init_logger(project="My LangChain Project")
auto_instrument()

prompt = ChatPromptTemplate.from_template("What is 1 + {number}?")
model = ChatOpenAI(model="gpt-3.5-turbo")

chain = prompt | model

response = chain.invoke({"number": "2"})
print(response.content)

# The LangChain interaction will be automatically traced and logged to Braintrust.

view raw JSON →