Upgrade StorageSkimRead, document-wide LLM skimming highlights for the PDF reader
I'd like to share a plugin I've been working on: SkimRead, which adds Semantic Reader–style skimming highlights to Zotero 9's built-in PDF reader.
It reads the whole document and selects the sentences that best convey its goal, method, results, and novelty, then shows them as colored overlays with margin labels. It highlights the paper's real sentences (guided AI selection), so there are no invented summaries.
Highlights
Whole-document selection, not sentence-by-sentence; chunked at chapter/section boundaries for long documents and books.
Label modes: the default Goal/Method/Result/Novelty scheme, your own custom labels, or zero-shot auto-discovery for non-standard documents.
Providers: Ollama (local), any OpenAI-compatible server (vLLM, LM Studio, OpenRouter), or your ChatGPT/Claude subscription via the Codex App Server and Claude Code CLIs.
Optional TL;DR summary of the paper, which can be saved as a Zotero note.
Highlights are ephemeral overlays by default and never touch your library; "Save as Zotero annotations" creates standard highlights that work with "Add Note from Annotations."
Privacy: local providers are the default and never send paper text off your machine. Cloud providers require an explicit opt-in, and there is no silent fallback.
Requires: Zotero 9 (macOS tested; the local/API providers should work cross-platform).
Install: download the latest .xpi from the releases page, then Tools → Plugins → Install Plugin From File.
https://github.com/adellife/zotero-skimread
It's designed to coexist with other Zotero AI plugins rather than duplicate them. It's open source (AGPL-3.0). Feedback, bug reports, and feature ideas are very welcome. Please note the usual caveat about automated selection bias; it's meant for fast skimming/orientation, not in-depth study.
It reads the whole document and selects the sentences that best convey its goal, method, results, and novelty, then shows them as colored overlays with margin labels. It highlights the paper's real sentences (guided AI selection), so there are no invented summaries.
Highlights
Whole-document selection, not sentence-by-sentence; chunked at chapter/section boundaries for long documents and books.
Label modes: the default Goal/Method/Result/Novelty scheme, your own custom labels, or zero-shot auto-discovery for non-standard documents.
Providers: Ollama (local), any OpenAI-compatible server (vLLM, LM Studio, OpenRouter), or your ChatGPT/Claude subscription via the Codex App Server and Claude Code CLIs.
Optional TL;DR summary of the paper, which can be saved as a Zotero note.
Highlights are ephemeral overlays by default and never touch your library; "Save as Zotero annotations" creates standard highlights that work with "Add Note from Annotations."
Privacy: local providers are the default and never send paper text off your machine. Cloud providers require an explicit opt-in, and there is no silent fallback.
Requires: Zotero 9 (macOS tested; the local/API providers should work cross-platform).
Install: download the latest .xpi from the releases page, then Tools → Plugins → Install Plugin From File.
https://github.com/adellife/zotero-skimread
It's designed to coexist with other Zotero AI plugins rather than duplicate them. It's open source (AGPL-3.0). Feedback, bug reports, and feature ideas are very welcome. Please note the usual caveat about automated selection bias; it's meant for fast skimming/orientation, not in-depth study.