Structured LLM Output

Overview

This project implements a structured LLM Output workflow. It demonstrates how Pydantic can structure output so it can be used for tasks such as user-input, data and LLM-ouput validation, as well as improving trace/span evals (among many others).

The repo is structured with:

• A Jupyter Notebook (pydantic_user_input_validation.ipynb) for ideation, design, and experimentation related to user-input validation and error handling

• A Jupyter Notebook (pydantic_llm_response_validation.ipynb) for ideation, design, and experimentation related to LLM-output validation and error handling

Getting Started

1. Install dependencies

pip install -r requirements.txt

2. Set Antropic or OpenAI API key In a .env file in the project root:

ANTHROPIC_API_KEY="your_api_key_here"
OPENAI_API_KEY="your_api_key_here"