Setting up LangSmith tracing in LangChain - Try It Yourself

Setting up LangSmith tracing

📖 Scenario: You are building a simple Python program using LangChain to run a language model. You want to track and trace the calls your program makes to the language model using LangSmith tracing. This helps you see what inputs and outputs your program produces.

🎯 Goal: Set up LangSmith tracing in your LangChain program to automatically record all language model calls.

📋 What You'll Learn

Create a LangChain OpenAI language model instance

Create a LangSmith tracer instance

Attach the LangSmith tracer to the language model

Run the language model with a prompt and have tracing enabled

💡 Why This Matters

🌍 Real World

Tracing language model calls helps developers debug and understand how their AI programs behave. LangSmith tracing records inputs, outputs, and metadata for each call.

💼 Career

Many AI developer roles require setting up monitoring and tracing for language model applications to ensure reliability and improve performance.

1

Create the OpenAI language model instance

Write code to import OpenAI from langchain.llms and create a variable called llm that is an instance of OpenAI with temperature=0.

LangChain

# Import OpenAI and create llm with temperature=0
# Your code here

Need a hint?

Use from langchain.llms import OpenAI to import the model, then create llm = OpenAI(temperature=0).

2

Create the LangSmith tracer instance

Write code to import LangSmithTracer from langchain_experimental.langsmith and create a variable called tracer that is an instance of LangSmithTracer.

LangChain

from langchain.llms import OpenAI

llm = OpenAI(temperature=0)

# Import LangSmithTracer and create tracer instance
# Your code here

Need a hint?

Import LangSmithTracer and create it with tracer = LangSmithTracer().

3

Attach the tracer to the language model

Write code to set the llm instance's callback_manager to the tracer instance.

LangChain

from langchain.llms import OpenAI
from langchain_experimental.langsmith import LangSmithTracer

llm = OpenAI(temperature=0)

tracer = LangSmithTracer()

# Attach tracer to llm's callback_manager
# Your code here

Need a hint?

Assign llm.callback_manager = CallbackManager([tracer]) to enable tracing.

4

Run the language model with tracing enabled

Write code to call llm with the prompt "Hello, LangSmith!" and assign the result to a variable called response.

LangChain

from langchain.llms import OpenAI
from langchain_experimental.langsmith import LangSmithTracer
from langchain.callbacks.manager import CallbackManager

llm = OpenAI(temperature=0)

tracer = LangSmithTracer()

llm.callback_manager = CallbackManager([tracer])

# Call llm with prompt and save to response
# Your code here

Need a hint?

Call llm("Hello, LangSmith!") and save it to response.