Changelog
Follow up on the latest improvements andΒ updates.
RSS
new
v1.46.0
Features π
- Images can be baked into AI answer, just tell AI to: "include images in your response"
- Added table of contents to search thread for easier navigation
Enhancements β¨
- Improve back button handling
- Improved search speed and stability with more relevant answers
- Improved "memory" of follow-ups
- Deleting items from list views now require a second step to prevent accidental deletes
Bug Fixes πͺ²
- Refetch library after uploading or deleting a document
new
1.44.7
Merry Christmas everyone! ππβοΈβοΈ
Features π
- Automatically generated search thread titles (editable)
Enhancements β¨
- Snappier experience when opening past searches
- Searches and follow-ups can be aborted midway through generation by firing up another query
Bug Fixes πͺ²
- Improved thumbnail selection when saving tweets
new
1.44.1
Features π
- Side by side search assistant when viewing library documents
- AI is now able to understand if it should be on writing mode or not (especially useful for transformative uses). You can still tell the AI to explicitly search for new sources if you wish.
- Refreshed settings page for user settings and workspace settings
- Added custom citation styles
Enhancements β¨
- Added a sneak peak of currently selected model at the model switcher button
- Improved responsiveness with tables generated by AI
- Improved system prompt instructions to generate answers with more hierarchy
Bug Fixes πͺ²
- Fixed result images height issue on Firefox browsers
- Reduced parsing errors when performing a search
new
1.41.1
Features π
- Added ability to convert an entire search thread into a note
Enhancements β¨
- Massively overhauled AI generation (switched from multiple stateless API calls to one singular streaming call)
- Improve confidence score calculation for follow-ups by considering more context from previous messages
Bug Fixes πͺ²
- Surface notifications banner when using E -> M shortcut
- Do not prepend extra hashtag to auto-generated tags
new
1.40.1
Features β¨
- Automatic tags generation when creating a note, link or file
- Collaborative search (multiple members of a workspace can ask follow up questions within a search thread)
Enhancements πͺ
- Huge performance upgrade to app performance especially on mobile devices (e.g. reduced UI flickering, responsiveness issues)
new
1.38.2
Features β¨
- Saving a new note without title will automatically generate a title
- Saving a search answer as a note will automatically generate a title
Enhancements πͺ
- Consolidated copy actions / buttons into one markdown copy
- Clicking on share will automatically copy the url for sharing
- Follow-ups now contain info badges as well
Bug Fixes πͺ²
- Fixed duplicate shortcuts in the "View all shortcuts" drawer
- Scrollable follow up search box
- Fixed history searches not returning any results
new
1.37.13
Features β¨
- Added new Gemini 2.0 Flash experimental model
- Initial mentions / sources are now carried over in follow ups so you don't have to re-mention them again
Enhancements πͺ
- Descriptions on AI models and what to choose
- Improve performance for library and listing pages (using react scan to reduce re-renders)
- Added partial slugs in url for better readability
- Add special weightage for mentioned documents, mentioned documents should always be returned now
- Uploading a new file / capturing links and notes will now show up instantly on the library, without needing a refresh
new
1.37.2
Features β¨
- Added ability to delete follow up messages
Enhancements πͺ
- Added new models Gemini-exp-1206, Llama-3.3-70B, Nova-Lite-v1, Beloga BLG-1.0 (routing model, will route to best fit model for query)
- Renamed creative mode to freestyle mode to avoid confusion
- Batch upload will not fail when one of them is unsupported or failed to upload
- Cosmetic changes to logged out state when viewing a shared resource
Bug Fixes πͺ²
- Fixed issue with rendering discolored cards in white mode
new
1.36.0
Features β¨
Follow-ups can now pull in more sources from the web and your library with respect to the current thread. This has been a widely requested feature that was a little tricky to get right but glad we took it in stages to test the waters before a full rollout π. Apologies for the slight delay on this π
new
1.35.4
Enhancements πͺ
- Added creative mode toggle to reduce ambiguity and improve access
- Enhanced answer generation with flexible structure depending on the prompt and context provided
- Improved overall UX / layout for search interface and library
- Added Qwen Marco, LearnLM, GPT4-turbo models
- Added a special tag for Believer plan subscribers π³
Bug fixes πͺ²
- Fixed issue with partial user creation
Load More
β