Changelog ChatAcademia

Stay up to date with all the latest updates, improvements, and new features we ship to help you do better research. We’re constantly improving ChatAcademia for academics and researchers.

Home

© 2026 ChatAcademia

Multilingual App, Reliable Chats & App Startup

Multilingual App, Reliable Chats & App Startup

ChatAcademia now supports a multilingual in-app experience, with browser-language detection, language switching, and localized interface copy across core research workflows. Chats, onboarding, PDF exports, and app startup are also more resilient in edge cases that previously caused internal errors or crashes.

Improvements

  • Choose your display language in the app, or let ChatAcademia start with your browser language automatically
  • Core app navigation, auth screens, account settings, agent routes, project panels, dialogs, buttons, tool renderers, dates, numbers, and empty states now use localized interface text
  • The language selector now supports a larger set of display languages with search, browser-language defaulting, and a layout that stays stable for longer language names
  • AI responses can receive the selected response language so generated research help stays aligned with the interface language
  • The ChatAcademia website now highlights multilingual app support on the main landing page with a searchable language directory
  • Chats and onboarding streams no longer fail when final response chunks arrive in an unexpected order
  • Memory setup during onboarding is more reliable when the backend is under database lock pressure
  • PDF exports now handle Greek letters and other scientific symbols by converting them to PDF-safe text instead of crashing
  • The app now handles browsers that block local storage more gracefully, reducing startup and settings-related failures

Smarter Errors & Guided Chat Recovery

Smarter Errors & Guided Chat Recovery

The app now routes known problems to the right surface — a focused modal, a clear toast, or inline text — instead of generic failures. When older chat history cannot be replayed by the current model, you get a dedicated explanation with a one-click path to start fresh in the same agent.

Improvements

  • Known server errors carry stable codes so the interface can show the right message and action (for example, insufficient credits opens the upgrade flow instead of a vague error)
  • Corrupted or incompatible chat history surfaces a dedicated modal that explains what happened and offers to open a new chat in the same workspace
  • Rate limits and permission issues use consistent, human-readable copy instead of opaque status text
  • Shared handling across agents and chats reduces mismatched error toasts between the sidebar, chat view, and project panels

Fixed Chats That Wouldn't Resume

Fixed Chats That Wouldn't Resume

Some chats with older history left over from interrupted conversations were failing to send a new message with an internal validation error. Those empty placeholder messages are now skipped automatically, so you can pick up where you left off without having to start a new chat.

Improvements

  • Resuming a chat no longer fails with a 'message must contain at least one part' error caused by gaps in older conversation history

Fewer Chat Timeouts & Clearer File Errors

Fewer Chat Timeouts & Clearer File Errors

Long chats on the pro model now recover automatically when one provider is slow, and any time we can't read your file we'll tell you exactly why — whether it's password-protected, too large, or has too many pages — so you know what to fix.

Improvements

  • Chats no longer fail with 'The operation timed out' when the pro model takes a long time to respond — they now automatically fall back to a backup model and continue
  • File reading errors now explain the actual problem (e.g. 'Document is password-protected — remove the password and try again', 'File is too large (47 MB; max 20 MB)') instead of a generic 'Bad Request' message
  • Academic database and grant searches surface specific, actionable errors when a provider rate-limits us, including when to retry

Faster Repeat Attachments, Reliable Streaming & Clickable Verdict Papers

Faster Repeat Attachments, Reliable Streaming & Clickable Verdict Papers

Chats with files attached from Zotero, Mendeley, Google Drive, or saved Notes now skip redundant downloads on every turn, so follow-up messages feel snappier. Long conversations on the fast model also recover more cleanly when something goes wrong mid-stream, and every paper analyzed by the Research Verdict agent now has a clickable title that opens the source.

Improvements

  • Files attached from Zotero, Mendeley, Google Drive, or saved Notes are reused across turns instead of being re-fetched from the source on every message — noticeably faster follow-ups in chats with attachments
  • Long-running conversations no longer fail intermittently with 'Failed after 3 attempts' errors caused by older history that some AI providers can't process
  • Every paper in the Research Verdict 'Papers Analyzed' tab now has a clickable title that opens the original source — previously some papers appeared as plain text