kreuzberg

Xberg ➡️ "The fast precise document-intelligence engine for every language: text, tables, metadata from 100+ formats, with OCR, transcription, code intelligence." ➡️ 107 formats · 141 file extensions · 371 code languages · 15 language bindings · 6 output formats · OCR · transcription · embedding" xberg-io/xberg Community Overview: lifehubber.com Resources are shared for discovery and are not independently vetted—please do your own due diligence. New resources are added regularly — feel free to join the sub for updates. Full searchable archive of all resources posted so far on our community site, LifeHubber: lifehubber.com 300+ open-ish AI models, agents, tools, datasets, and related resources, with filtering and sorting.

@burkov @ucixrn Can you shed more light on the problematic formats? (I'm guessing that you've nailed PDF and ePub) Have you heard of or tried using xberg-io/xberg for other formats?

All of them are listed as either "in development" or "early alpha", and every single person that has tried to do something even roughly resembling a task you would do with the original product can tell you that it's broken in a thousand ways. As it stands right now, it proofs little more than it can replicate the post-boot screen of the original products. Just because every tech news outlet was dumb enough to report it without trying it doesn't mean you have to join the echo chamber. --- There are definitely examples out there that do use a similar relentlessly agentic development style that have produced great results. e.g. xberg[0] (formerly kreuzberg), is a great document content extractor, and in a relatively short amount of time has become one of the most robust and performant. I've reported about a dozen edge cases of documents of various types that were unextractable, and the maintainer (or their agents) were able to add a fix in maximum a day turnaround every time. [0]: xberg-io/xberg