LlamaParse

LlamaParse extracts structured data from diverse, complex documents via an API, optimizing them for large language models (LLMs).

LlamaParse, an API offered by LlamaIndex, efficiently transforms unstructured documents—like PDFs—into a structured markdown format. It's engineered to handle a variety of complex layouts, including those with intricate tables and figures, by converting them into a format ideal for LLM ingestion. This process bypasses the common issues of lost context and formatting that plague other parsing methods, ensuring high-quality data extraction. Developers integrate LlamaParse using a simple API key and Python client, specifying their desired output—either markdown or a LlamaIndex 'Document' object. The service supports both synchronous and asynchronous parsing, accommodating individual files or bulk uploads, and includes features like an 'until_page' parameter for partial document processing. This capability significantly streamlines the preparation of proprietary data for RAG applications, enhancing the accuracy and relevance of LLM outputs.

https://docs.llamaindex.ai/en/module_guides/loading/simple_directory/llama_parse/