Release ObsidiAI 0.1.9 · KNN-07/ObsidiAI

ObsidiAI 0.1.9

Attach PDFs through the computer file picker, drag and drop, or clipboard file paste.
Extract embedded text locally in page order. Preview the extracted text before sending; only that text, under the original filename, enters model context and saved history.
PDF text works with text-only models. No OCR or page/image rendering is included; scanned PDFs without embedded text need OCR first. Complex layouts and tables may not preserve reading order.
Enforce limits of 20 MiB, 500 pages, and 200,000 extracted characters per PDF. Reject invalid, unsupported, empty-text, and oversized documents rather than silently truncating them. Password-protected PDFs may be rejected.
Update attachment picker hints and the user guide. Bundle exact-pinned unpdf@1.8.1; no separate PDF worker installation is needed.

Download obsidiai-0.1.9.zip, not GitHub's automatic source archive.
Replace main.js, manifest.json, and styles.css in <vault>/.obsidian/plugins/obsidiai/. Preserve data.json and history.json if present.
Reload Obsidian or disable/re-enable ObsidiAI after the current run settles.

Requirements remain Obsidian desktop 1.11.4+ and embedded Node 22.19.0+. The three individual plugin files are also attached.

Production build, all 168 tests, release metadata/assets validation, and generated-bundle syntax checks passed.
Attachment tests cover compressed multipage extraction, basename retention, invalid signatures, damaged/mislabeled PDFs, empty text, and byte/page/extracted-text limits.
An isolated CommonJS bundle smoke check extracted text from a real PDF through the attachment input path using the production bundler configuration.
Native Obsidian file-picker, drag/drop, clipboard, and preview interactions for PDFs have not been verified.