datalab-to/surya is a collection of Python tools that processes documents in more than 90 languages. The latest release is v0.22.1. The provided record contains no changelog for this version, so no newly added or changed features can be confirmed for this release.
Beyond OCR (optical character recognition), which reads the characters on a page, the project provides layout analysis, reading order detection, and table recognition. It does not merely transfer the contents of a page into characters, but also handles the structure in which information is arranged within a document. Layout analysis distinguishes the components of a page, and reading order determines the sequence in which those components are read. Because it provides document OCR and layout analysis together, it is focused on tasks that handle character recognition and page structure understanding in a single tool. The README summary also includes separate models for text line detection and OCR error detection. The code is licensed under Apache-2.0, and the model weights are licensed under a modified AI Pubs Open Rail-M license.
Therefore, the code and the weights cannot be treated as under the same license. The conditions for using the weights must be checked in the separate license. The provided information contains no detailed changes or performance evaluation for v0.22.1. It is difficult to judge the extent of the improvements in this version from this material alone, and the verifiable scope reaches no further than the document processing capabilities and the license difference between the two components.