LlamaIndex

Retrieval Harness vBeta

LlamaParse Retrieval Harness is an enhancement to LlamaParse Index, providing AI agents with advanced filesystem primitives for real-time document traversal, grep, and reading. It includes features like visual layout preservation and pipeline observability, enabling agents to navigate complex documents effectively. This development addresses the limitations of traditional RAG by providing deterministic, systems-level utilities for autonomous agents to actively interrogate, verify, and traverse documents in real time.

LiteParse v1.0.0

LiteParse is an open-source, high-performance document parsing tool designed specifically for AI agents. It offers a lightweight and efficient solution for processing documents locally, enabling faster and more reliable AI-driven document analysis.

LiteParse 2.1 v2.1

LiteParse 2.1 is an open-source, model-free tool developed by LlamaIndex for converting PDFs to markdown format. It offers a fast and efficient solution for document parsing without relying on cloud services or LLM tokens. The tool has been benchmarked against three standard tests—opendataloader-bench, olmOCR-bench, and ParseBench—achieving top overall scores in each when compared to other model-free approaches. Users can install LiteParse 2.1 via pip and utilize it through command-line interface (CLI) or as a Python library. The tool is also available as a Node.js library, Rust library, and in-browser via WASM, ensuring versatility across different platforms. For more information and access to the source code, visit the official LiteParse page on LlamaIndex's website.