llmstruct 0.1.0

A Python library for reliably extracting structured JSON from text and validating it with Pydantic models.

LLMStruct

llmstruct is a Python library for reliably extracting structured JSON from text and validating it with Pydantic models.

Features

• Extracts JSON objects or arrays of JSON objects from messy text.
• Validates extracted JSON against Pydantic models.
• Resilient to surrounding text and other noise.

Installation

You can install the library using pip:

pip install .

Or with uv:

uv pip install .

Demos

This project includes two demo scripts to showcase its capabilities by interacting with the Anthropic API.

Prerequisites

Before running the demos, you need to install the demo-specific dependencies and set your Anthropic API key.

1. Install demo dependencies:

   uv pip install -e ".[demos]"
   

2. Set the API Key: Export your Anthropic API key as an environment variable:

   export ANTHROPIC_API_KEY="your-api-key"
   

Simple Demo

The simple demo shows the basic usage of extract_json_from_text to extract a single JSON object from a response generated by the Anthropic API.

To run the simple demo:

python demos/simple.py

Advanced Demo

The advanced demo illustrates a powerful, real-world use case. It performs the following steps:

1. Fetches a large, complex JSON object for a product from the public Open Food Facts API.
2. Sends this raw JSON to the Anthropic API and instructs the LLM to extract and structure a concise summary.
3. Streams the LLM's response and uses llmstruct to parse the structured summary from the text.

The entire process is wrapped in a tenacity retry block, making the operation resilient to transient network issues or occasional LLM failures to generate a perfectly valid object on the first try. This demonstrates how llmstruct can be a key component in a robust data processing pipeline.

To run the advanced demo:

python demos/advanced.py

License

This project is licensed under the MIT License - see the LICENSE file for details.