PaperDock v0.1.24

Gemini WebChat release.

• Add Gemini WebChat as a built-in web model provider.
• Use text-context mode for Gemini so it can read the current paper without requiring the fragile PDF attachment flow.
• Fix Gemini answer capture so PaperDock keeps the generated answer instead of showing the submitted prompt or paper context.
• Restore Markdown structure for compact Gemini numbered summaries, limitations, learning advice, and follow-up questions.
• Preserve already generated WebChat content when a Gemini stream or page connection disconnects.
• Add Gemini WebChat contract and stream parser tests.

PaperDock v0.1.23

Paper context parity release.

• Align paper chat context assembly with llm-for-zotero: first paper turn sends full paper context, ordinary follow-up turns use focused retrieved evidence, and cache-capable providers can reuse full paper context.
• Send paper evidence as a separate Document Context system message while keeping user and assistant chat history as separate messages.
• Improve follow-up retrieval by anchoring the abstract once, enriching retrieval with only the last assistant answer, and excluding references from empirical evidence.
• Add input-window truncation disclosure so the model tells users when full-paper coverage had to be shortened.
• Add contract tests for paper context planning, streaming fallback, strict-temperature retries, and Zotero suggestion fallback.

PaperDock v0.1.22

PaperDock rebrand release.

• Rename the extension, repository, UI, update links, WebChat bridge, and native helper identifiers to PaperDock.
• Replace the extension icon with the new PaperDock icon.
• Publish the stable extension ZIP as PaperDock-v0.1.22.zip.
• Keep the core paper chat, PDF reading, WebChat, and Zotero workflows intact.

arXivMate v0.1.21

Preferred model and fixed-temperature model release.

• Add a preferred model profile that is used by default across paper/PDF chat pages.
• Remember the model a user manually selects in any chat page until they switch again.
• Add a settings action and list badge for setting the preferred model explicitly.
• Fix GPT-5/o-series compatible requests by forcing temperature: 1 when the model requires it.
• Lock the settings temperature control to the model-supported boundary for fixed-temperature models.
• Add contract tests for preferred model persistence and fixed-temperature request bodies.

arXivMate v0.1.20

Zotero suggestion, model settings, and README cleanup release.

• Improve Zotero collection suggestions with tolerant parsing and local fallback matching when model output is empty or unusable.
• Preserve usable Zotero suggestions even when the background response is empty by falling back in the paper panel.
• Clean up removed model-dropdown refresh code and guard against old custom model menu regressions.
• Restore stable PDF panel behavior and protect textarea/model-select interactions.
• Simplify the README into a shorter project introduction.
• Add contract tests for Zotero suggestion parsing, PDF iframe panel behavior, model select stability, and source cleanup.

arXivMate v0.1.19

Summary

• Add Zotero Connector integration for reading local Zotero collections and saving papers into a selected collection.
• Add collection search, AI collection suggestions, optional Zotero note, tags, and PDF attachment saving.
• Fix hover-based collection tree expansion/collapse and embedded PDF panel loading for Zotero helpers.

Install / Update

1. Download arXivMate-v0.1.19.zip from this release.
2. Unzip it.
3. In Chrome/Edge, open Extensions, enable Developer mode, and load the unzipped arXivMate folder.
4. For an existing unpacked install, replace the previous files, click reload on the extension card, then refresh open paper pages.

arXivMate v0.1.18

arXivMate v0.1.18

Highlights

• README is restructured so GitHub visitors immediately see supported browsers, supported paper websites, WebChat sites, installation, upgrade, data, and privacy notes.
• Supports Chrome, Microsoft Edge, Brave, Arc, and other Chromium browsers as unpacked extensions.
• Documents arXiv, ACM Digital Library, IEEE Xplore, regular/dynamic PDFs, common publisher PDF entry points, and local PDFs.
• Fixes ChatGPT Web quick overview mode by switching to fast mode after PDF upload.
• Fixes DeepSeek Web mode handling and prevents thinking-only content from ending a normal response.

Assets

• �rXivMate-v0.1.18.zip: recommended stable package for loading unpacked in Chrome / Edge.
• �rXivMate-v0.1.18.crx: packaged CRX for users who know how to install CRX files.

Upgrade

Do not uninstall the extension. Replace the existing loaded folder, reload the same extension card, and refresh open paper pages.

arXivMate v0.1.17

arXivMate v0.1.17

Chat navigation and WebChat thinking-status release.

Changes

• Add quick previous/next buttons in the paper chat composer for browsing the current conversation.
• Fix chat navigation during generation: pressing next at the newest message now returns to the live bottom and resumes auto-following streamed output.
• Show active thinking/waiting status for ChatGPT Web and DeepSeek Web, including elapsed wait time and received reasoning length when available.
• Add lightweight contract tests for chat navigation and WebChat thinking status.

Download

• arXivMate-v0.1.17.zip: recommended for unpacked Chrome installation.
• arXivMate-v0.1.17.crx: packaged extension file.

arXivMate v0.1.16

PDF sites, manual save, and Markdown rendering release.

What's changed

• Add ACM Digital Library detection and page/full-text fallback for DOI, PDF, EPDF, and Full HTML pages.
• Improve WebChat PDF preparation for ACM/IEEE pages by falling back to readable page text and generated context PDFs when direct PDF download is blocked.
• Stop auto-saving generated chat history. Conversations enter the review library only when the user clicks Save, while current-page follow-up context still works.
• Improve Markdown rendering for escaped math, math-only code spans, and compact section dividers.
• Rewrite the README opening so supported sites and PDF page types are visible first.

Verify

• node --check passed for 11 JS files.
• History persistence contract passed.
• Markdown/math rendering contract passed.
• git diff --check reported only CRLF warnings.

arXivMate v0.1.15 - WebChat PDF attachment and thinking display

arXivMate v0.1.15 - WebChat PDF attachment and thinking display

This release fixes WebChat paper reading for ChatGPT Web and DeepSeek Web.

• Upload PDFs as verified web attachments before the first paper turn.
• Reuse only verified pdfAttached WebChat sessions, so old text-only sessions no longer skip PDF upload.
• Persist verified WebChat attachment metadata locally without clearing model settings or paper chat history.
• Fold ChatGPT/DeepSeek thinking or reasoning output into a collapsible thinking block.
• Add a Markdown fallback renderer for arXivMate thinking blocks.

Upgrade

• Git users: git pull on the existing install folder, then reload arXivMate in chrome://extensions.
• ZIP users: download arXivMate-v0.1.15.zip, overwrite the existing arXivMate folder, then reload the same extension card in chrome://extensions.

After upgrading, refresh already-open paper, ChatGPT, and DeepSeek pages.