Mastering structured output with LangChain

Articles

The notebooks in this repo are described in the following series of blog posts:

1. Mastering Structured Output in LLMs 1: JSON output with LangChain
2. Mastering Structured Output in LLMs 2: Revisiting LangChain and JSON
3. Mastering Structured Output in LLMs 3: LangChain and XML

Setup and Installation

This project uses Poetry for dependency management. Follow these steps to get started:

1. Install Poetry if you haven't already:

curl -sSL https://install.python-poetry.org | python3 -

2. Clone this repository and navigate to the project directory:

git clone git@github.com:adocherty/langchain-structured-output-evaluation.git
cd langchain-structured-output-evaluation

3. Create a virtual environment and install dependencies:

poetry install

4. Activate the virtual environment:

poetry shell

5. Launch Jupyter to run the notebooks:

jupyter notebook

Running Examples

These examples can be run on Binder:

See index.ipynb for a description and link to the notebooks.

Notes

httpx has been fixed to version 0.27.2 to avoid a bug due to changed parameters https://community.openai.com/t/error-with-openai-1-56-0-client-init-got-an-unexpected-keyword-argument-proxies/1040332

Requirements

• Python 3.12
• Poetry for dependency management
• Ollama installed locally for running local LLMs
• Anthropic API key for Claude access (optional)
• Fireworks.ai API key for cloud models (optional)