We are open sourcing parsebench, the first document parsing benchmark built for the agent era. It contains 2,000 human verified enterprise pages evaluated across five dimensions. It contains tables, charts, content faithfulness, semantic formatting, and visual grounding with over 160,000 rules in total.Agents are the new consumers of document parsing, and they need semantic correctness. If a table header is shifted one column, the agent silently pulls the wrong number. If a chart gets flattened to raw text, ...

Hi everyone, Felia from Lindex here and today I am really exciting to introduce you to a new tool that we are releasing today. Light pars an open-source parsing tool with spatial text extraction that doesn't have any LLM or cloud dependency. Lights has a really fast and fully local text extraction with a flexible OCR system that has a built-in sensible default with test and uh can also be extended through HTTP servers.It can generate screenshot and uh it does produce multiple output formats such as JSON and ...

Llama index simplifies the process of passing and indexing documents which is crucial for end to end AI development. My name is Manuel and I moved from Spain to the UK about 9 years ago to join entity data. I spent my entire career in IT consulting starting out as a software engineer and then gradually moving into project and program management roles before taking on more strategic positions in operations and transformation.At the beginning of 2025, I joined Entity Data's global AI office, a team that drive ...

Ever wondered how to automatically pull structured data out of messy PDFs, receipts, invoices, or text files? In this video, we'll walk you step-by-step through building your very own VLM-powered document parsing and information extraction agent. Notebook from this webinar: https://tinyurl.com/58xsy96n LlamaParse signup: https://www.llamaindex.ai/signup ...