ePub, docx, and xlsx do not work

[raffits]

I have tried importing docx, xlsx, and ePub but they didn't work (pdf, txt, and csv no issues). I tried the PicoDocExample, same problem, failed on opening the files. The necessary identifiers are in my Info.plist. Also, the HTML did not do any conversion, raw html content. Any thoughts?

[ronaldmannak]

Hi @raffits, thanks for letting me know and apologies for the issue. I'm planning to update this repo and enable it in Pico AI Server in the next few weeks. I'll see if I can replicate the issues you mentioned then if that works for you.

[ronaldmannak]

Update: the import/conversion engine has been rewritten on main, and all four problems reported here are fixed:

• DOCX / XLSX / ePub now import and convert. The root cause was detection: .docx/.xlsx were declared as XML (they're ZIP), and type matching used exact equality. Detection is now content-based (magic bytes + ZIP central-directory inspection), and the converters parse the real format (DOCX via OOXML, XLSX via CoreXLSX, ePub via ZIP + the HTML→Markdown walker) instead of routing through NSAttributedString.
• HTML now converts to Markdown instead of returning raw markup, with an optional Reader-mode pass (pure-Swift Readability) that keeps the article and drops nav/ads.

Other changes: a stateless PicoDocs.convert(...) engine usable off the main actor, fetch/parse now throw (so try await is correct), and the WebKit/EPUBKit dependencies are gone. There's a CI test suite (swift test) covering the formats above, including a regression guard for the detection bug behind this issue.

It's on main now (not yet in a tagged release). If you're able to try it against the files that failed for you, that'd be a great sanity check — otherwise this can close once it ships in a release.

---

Generated by Claude Code