Example: Integrating LangGraph with TensorZero

Overview

This example demonstrates how to integrate LangGraph with TensorZero to create a simple chatbot that can call external APIs.

Getting Started

TensorZero

We provide a simple TensorZero configuration with a function chatbot that uses GPT-4o Mini and has access to a tool called temperature_api.

Prerequisites

Install Python 3.10+.
Install the Python dependencies. We recommend using uv: uv sync
Generate an API key for OpenAI (OPENAI_API_KEY).

Setup

Set the OPENAI_API_KEY environment variable.
Launch the TensorZero Gateway: docker compose up
Run the chatbot using python main.py. The script is interactive.

Sample Run

                +-----------+
                | __start__ |
                +-----------+
                      *
                      *
                      *
                 +---------+
                 | chatbot |
                 +---------+
               ...          ...
              .                .
            ..                  ..
+-----------------+           +---------+
| temperature_api |           | __end__ |
+-----------------+           +---------+

[User]
My name is Gabriel.

[Assistant]
Nice to meet you, Gabriel! How can I assist you today?

[User]
What is my name?

[Assistant]
Your name is Gabriel.

[User]
What is the weather in NYC?

[Tool Call: temperature_api]
{"location":"New York City"}

[Tool Result: temperature_api]
25

[Assistant]
The current temperature in New York City is 25°C. If you need more weather information, feel free to ask!