This is essentially a (vibe-coded?) wrapper around PaddleOCR: https://github.com/PaddlePaddle/PaddleOCR
The "guts" are here: https://github.com/majcheradam/ocrbase/blob/7706ef79493c47e8...
Claude is included in the contributors, so the OP didn’t hide it
At this point it feels like HN is becoming more like Reddit, most people upvote before actually checking the repo.
Most production software is wrappers around existing libraries. The relevant question is whether this wrapper adds operational or usability value, not whether it reimplements OCR. If there are architectural or reliability concerns, it’d be more useful to call those out directly.