Skip to content

Releases: WolframResearch/Chatbook

Experimental Release

31 Dec 14:17
32f56cf
Compare
Choose a tag to compare
Experimental Release Pre-release
Pre-release

This is an experimental release that's always updated with the latest build from the main branch.

Release v1.5.1

07 Oct 20:45
3926b64
Compare
Choose a tag to compare

What's Changed

  • Miscellaneous paclet optimizations by @rhennigan in #757
  • Fix 447719: let context menus be localizable by @rhennigan in #759
  • Fix 447860: let long menu item labels line wrap by @rhennigan in #760
  • Basic support for chatgpt-4o-latest by @rhennigan in #761
  • Serialize =[query] as natural language input by @rhennigan in #762
  • Fix Spacing Issues for Inline Code Cells by @rhennigan in #763
  • Fix: Correct tokenizer application for generic tokenizer by @rhennigan in #766
  • Implement Vector Databases by @rhennigan in #767
  • Log timing for different parts of a chat evaluation by @rhennigan in #768
  • Prompt generators for code assistance by @rhennigan in #769
  • Feature: Code assistance UI updates by @rhennigan in #770
  • NotebookEditor Tool (WIP) by @rhennigan in #771
  • More notebook editor tool implementation by @rhennigan in #772
  • Faster embeddings for Code Assistance by @rhennigan in #773
  • Added ShowContentSuggestions for BoxData cells by @rhennigan in #775
  • Install Persona/Tool from Definition Notebook by @rhennigan in #776
  • Bugfix: Don't show hidden tools in UI by @rhennigan in #777
  • Added support for TextData completions by @rhennigan in #778
  • Support new format of ChatSubmit stream data by @rhennigan in #779
  • Bugfix: Handle missing "ContentDelta" key in streaming data by @rhennigan in #780
  • Improved content suggestions by @rhennigan in #781
  • Initial work to support o1 models (still need ChatSubmit support) by @rhennigan in #782
  • Use local model for embeddings by @rhennigan in #783
  • More content suggestion improvements by @rhennigan in #784
  • Bugfix: Added unhandled argument pattern for formatRaw by @rhennigan in #786
  • Feature: Saving and Restoring Previous Chats by @rhennigan in #787
  • Added delete button to workspace chats and some other misc UI tweaks by @rhennigan in #788
  • Bugfix: Fixed duplicate incompatible definitions for stringToBoxes by @rhennigan in #789
  • Bugfix: Prevent ShowStringCharacters message for partial strings in code blocks by @rhennigan in #791
  • Custom progress display for inline and workspace chats by @rhennigan in #792
  • Bugfix: Workspace chat input alignment by @rhennigan in #793
  • Workspace chat feedback buttons by @rhennigan in #794
  • Make workspace chat cells fully editable by @rhennigan in #795
  • Added support for ListSavedChats[NotebookObject[...]] by @rhennigan in #796
  • Allow workspace chat window to be resized on Mac by @rhennigan in #797
  • Added "ForceSynchronous" setting to allow non-streaming models like o1-preview to work by @rhennigan in #798
  • Search Functionality by @rhennigan in #799
  • Bugfix: Fixed tool evaluations when ForceSynchronous is True by @rhennigan in #800
  • Bugfix: Handle changes to CellObject in cloud by @rhennigan in #802
  • Bugfix: Fix several version compatibility issues by @rhennigan in #803
  • Bugfix: Don't show unusable gpt-4o-realtime models in menus by @rhennigan in #804
  • Bugfix: The latest snapshot of gpt-4o is really bad at textual tool calling by @rhennigan in #805
  • Bugfix: Added workaround for GoogleGemini models failing in ChatSubmit by @rhennigan in #806
  • Avoid some errors if code assistance is programmatically enabled in an earlier WL version by @rhennigan in #807
  • Minor local model optimizations by @rhennigan in #808
  • Bugfix: Quiet RegularExpression::maxrec message during formatting by @rhennigan in #809
  • Bugfix: Fix some bad/weird spacing in inline code and WL code blocks by @rhennigan in #810
  • Testing local models for code completion by @rhennigan in #811
  • Feature: Added multi-cell content suggestions by @rhennigan in #812
  • Added some settings that are useful for automated testing by @rhennigan in #813
  • Bugfix: Added fallback definition for unrecognized boxes when exploding cells by @rhennigan in #815
  • Bugfix: GPT-4o often writes tool calls in a partial code block by @rhennigan in #816
  • Bugfix: If ChatObject[...] failed for some reason, save chat using the fallback dataset by @rhennigan in #817
  • Release v1.5.1 by @rhennigan in #818

Full Changelog: v1.5.0...v1.5.1

Release v1.5.0

07 Aug 17:13
f6143fb
Compare
Choose a tag to compare

What's Changed

Full Changelog: v1.4.8...v1.5.0

Release v1.4.8

01 Aug 15:14
6ddc474
Compare
Choose a tag to compare

What's Changed

  • Bugfix: Fixed an issue with % not giving the last ChatObject by @rhennigan in #748
  • Bugfix: $currentSettingsCache should only be reset to <||> if it currently exists. by @rhennigan in #749
  • Bugfix: Fixed bad parsing of code blocks to BoxData[{...}] that should be BoxData[RowBox[...]] by @rhennigan in #750
  • Bugfix: Fixed bad contexts for inline reference boxes by @rhennigan in #751
  • Minor fixes needed for ImportMarkdownString and ExportMarkdownString resource functions by @rhennigan in #752
  • Bugfix: Add ImageSize -> Automatic to the send chat button PaneSelector by @rhennigan in #753
  • Release v1.4.8 by @rhennigan in #754

Full Changelog: v1.4.7...v1.4.8

Release v1.4.7

25 Jul 18:35
1a317a5
Compare
Choose a tag to compare

What's Changed

  • Update the checkout and upload artifact actions to the latest versions by @rhennigan in #644
  • Bugfix: Set $ChatNotebookEvaluation when calling the WL tool function programmatically by @rhennigan in #646
  • Minor build fixes by @rhennigan in #647
  • Use front end resources for strings by @rhennigan in #648
  • Bugfix: Fixed an internal failure in model selection menu by @rhennigan in #650
  • Bugfix: Fixed an internal failure that occurs when an ExternalLanguage cell appears in chat history in cloud by @rhennigan in #652
  • Fold bad tool calls into a "discarded material" opener by @rhennigan in #654
  • Bugfix: PreemptProtect some FE operations in cloud by @rhennigan in #655
  • Fixed an issue where icons specified by a service were not being used in the UI by @rhennigan in #667
  • More serialization improvements by @rhennigan in #668
  • Bugfix: Suppress ServiceExecute::multser messages in chat evaluations by @rhennigan in #670
  • Bugfix: Localize preference tab labels by @rhennigan in #671
  • Bugfix: Allow popup menu text to be localizable by @rhennigan in #672
  • Bugfix: Allow localization in cell insertion menu by @rhennigan in #673
  • Bugfix: apply custom formatter function to static outputs as well as dynamic ones by @rhennigan in #674
  • Support localization of core persona names by @rhennigan in #675
  • Bugfix: Fixed $IterationLimit::itlim error that occurs when evaluating a raw string in the WL evaluator tool by @rhennigan in #677
  • Bugfix: Fixed an internal error in preferences panel due to non-string persona display names by @rhennigan in #679
  • Bugfix: Fix unevaluated FEPrivate`Part in tool manager panel by @rhennigan in #680
  • Bugfix: Fix CurrentChatSettings returning failures when used in CloudSubmit by @rhennigan in #681
  • Google Gemini updates by @rhennigan in #682
  • Misc formatting fixes and improvements by @rhennigan in #683
  • Added "BypassResponseChecking" option which lets you bypass the default error checking in the final result by @rhennigan in #684
  • Support for GPT-4 Omni by @rhennigan in #686
  • Bugfix: Workaround for web search changes by @rhennigan in #687
  • Reduce the height of the preferences panel in cloud and make it scrollable by @rhennigan in #688
  • Added option to allow evaluator tool to use the current session by @rhennigan in #689
  • Improved tool call prompting and switch to gpt-4o as default by @rhennigan in #690
  • Bugfix: Fix internal error for fine tune models that have a specified name by @rhennigan in #691
  • Allow ParentList specification in persona tools to allow inheriting tools from notebook, FE, etc. by @rhennigan in #692
  • Implement toolRequestParser in chatbook code to avoid dependency issues by @rhennigan in #693
  • Fix for LLMTool changes in 14.1 by @rhennigan in #694
  • Added "ConversionRules" option to customize serialization of lower level box structures by @rhennigan in #695
  • Fix a bad error message in conversion rules by @rhennigan in #696
  • Bugfix: Fixed unnecessary FE <-> Kernel traffic for localized persona names by @rhennigan in #697
  • Session evaluator improvements by @rhennigan in #698
  • More session evaluator work by @rhennigan in #699
  • Bugfix: Fixed some uses of Part with LLMTool to handle 14.1 changes by @rhennigan in #700
  • Set caller for any service execute evaluations by @rhennigan in #701
  • Minor evaluator and formatting improvements by @rhennigan in #702
  • Ensure text styling is applied to relevant inline cells by @rhennigan in #704
  • Added support for markdown delimiters and block quotes by @rhennigan in #705
  • Create a URI handle for any expressions that have non-trivial formatting by @rhennigan in #706
  • Bugfix: Allow whitespace to precede simple tool command syntax by @rhennigan in #707
  • Allow some registered services to be excluded from UI menus by @rhennigan in #708
  • Bugfix: Wrong short name for tools by @rhennigan in #709
  • Bugfix: Fixed an internal error that can occur when CellToChatMessage is used outside of a chat evaluation by @rhennigan in #712
  • Simplify shared contexts by @rhennigan in #713
  • Initial work towards a separate floating chat window by @rhennigan in #714
  • Bugfix: MX build issues and LLMTool fixes by @rhennigan in #715
  • Updated design for Wolfram Alpha pods by @rhennigan in #716
  • Bugfix: Avoid some messages that can leak outside of Quiet[...] by @rhennigan in #717
  • Bugfix: Handle timeouts in the session evaluator by @rhennigan in #719
  • Improved support for claude-3.5 by @rhennigan in #720
  • Bugfix: Display markdown images for personal cloud objects with private permissions by @rhennigan in #721
  • Bugfix: Add missing argument pattern to importImage0 by @rhennigan in #723
  • Bugfix: Anthropic tool response format was being used for all models by @rhennigan in #724
  • Bugfix: Fix internal error that occurs when encountering a cell with no style by @rhennigan in #727
  • MistralAI fixes and better bug reports by @rhennigan in #729
  • More workspace chat implementation by @rhennigan in #731
  • Initial prototype for inline chat by @rhennigan in #733
  • Fix conflicting keybinding by @rhennigan in #734
  • Swap modifier key on MacOSX to avoid conflict with "OpenCloseGroup" token by @rhennigan in #735
  • Selection context improvements by @rhennigan in #736
  • Workaround: Avoid unpacking raster boxes in cloud by @rhennigan in #737
  • Bugfix: Fix send chat button in cloud by @rhennigan in #738
  • Workarounds for send/stop button in cloud notebooks by @rhennigan in #739
  • Bugfix: Handle escaped markdown characters correctly by @rhennigan in #741
  • Minor GPT-4o and GPT-4o-Mini updates by @rhennigan in #742
  • Increment paclet version to 1.4.7 since an earlier 1.4.6 is included in 14.1 by @rhennigan in #743
  • Disable code assistance menu items since they aren't ready for 14.1 yet by @rhennigan in #744
  • Release v1.4.7 by @rhennigan in #745
  • Revert back to submit-paclet action instead of script to avoid segfault by @rhennigan in #746
  • Release v1.4.7 by @rhennigan in #747

Full Changelog: v1.4.5...v1.4.7

Release v1.4.5

11 Apr 16:23
463c800
Compare
Choose a tag to compare

What's Changed

  • Major update to support additional persona customizations by @rhennigan in #608
  • Support natural language input for the AI in WL evaluator tool by @rhennigan in #610
  • Bugfix: Fix mouseover error on notebooks cloud deployed from desktop by @rhennigan in #611
  • Bugfix: Only do full cell rasterization for non-chat cells at top level by @rhennigan in #613
  • Add syntax to let the AI insert images inline in the evaluator and markdown code blocks by @rhennigan in #616
  • A few minor cloud improvements by @rhennigan in #617
  • Append URI instructions to evaluator outputs by @rhennigan in #618
  • Bugfix: Add missing pattern for logUsage by @rhennigan in #619
  • Bugfix: Handle server errors returned as XML by @rhennigan in #620
  • 621 an uncaught throw in the evaluator tool gives an internal error by @rhennigan in #622
  • Bugfix: Fixed internal error that occurs if the AI writes an unbalanced markdown table by @rhennigan in #623
  • Bugfix: Markdown tables with less than 3 columns were not formatting by @rhennigan in #624
  • Minor improvements and fixes by @rhennigan in #626
  • Automatically remove the string "<|image_sentinel|>" from chat outputs by @rhennigan in #627
  • Additional minor fixes by @rhennigan in #628
  • Add Wolfram Alpha sources to tool outputs by @rhennigan in #629
  • Improved Documentation to Markdown Conversion by @rhennigan in #631
  • Added BasePrompt by @rhennigan in #632
  • Avoid error messages if code block has been pulled out of its surrounding dynamic module by @rhennigan in #633
  • Bugfix: Show steps in pods to match raw text by @rhennigan in #635
  • Bugfix: Fixed minor inconsistency in system prompt tool call example by @rhennigan in #637
  • Improve LLM behavior by consistently escaping certain markdown characters during serialization by @rhennigan in #638
  • Add support for Claude 3 by @rhennigan in #640
  • Bugfix: Use multimodal tokenizer for updated gpt-4-turbo by @rhennigan in #642
  • Release v1.4.5 by @rhennigan in #641

Full Changelog: v1.4.4...v1.4.5

Release v1.4.4

16 Feb 17:26
ce7fb3b
Compare
Choose a tag to compare

What's Changed

  • Bugfix: Handle service API errors better when retrieving model lists for the preferences dialog by @rhennigan in #597
  • Bugfix: Fix an internal failure that can occur when a dynamic chat output hasn't been cleaned up properly by @rhennigan in #598
  • Bugfix: Use correct settings for "VisiblePersonas" and "PersonaFavorites" in menus by @rhennigan in #599
  • Bugfix: Fixed pink boxing that occurs in cloud when chat outputs contain inline section cells by @rhennigan in #601
  • Bugfix: Make cloud tagging rules inheritence fix version independent by @rhennigan in #602
  • Release v1.4.4 by @rhennigan in #603

Full Changelog: v1.4.3...v1.4.4

Release v1.4.3

08 Feb 18:15
4abdb41
Compare
Choose a tag to compare

What's Changed

Full Changelog: v1.4.2...v1.4.3

Release v1.4.2

04 Feb 16:42
6cd862a
Compare
Choose a tag to compare

What's Changed

  • Bugfix: Fix for some settings not being inherited at the chat block level by @rhennigan in #564
  • Improve formatting for outputs that contain $ to denote currency by @rhennigan in #567
  • Prevent unnecessary autoloading when reading persona files by @rhennigan in #568
  • Enable a tool when installing by @rhennigan in #569
  • Exclude WolframAlpha pods that have missing content from text sent to the LLM by @rhennigan in #571
  • Support AppearanceRules for tools by @rhennigan in #573
  • Bugfix: Fixed an internal failure that can occur when a previous chat output cell is removed early during a chat evaluation by @rhennigan in #574
  • Base prompt improvements by @rhennigan in #577
  • Feature: Added $ChatNotebookEvaluation and fixed some other minor misc issues by @rhennigan in #578
  • Bugfix: Fixed some symbols being created in the global context when using the chat stylesheet by @rhennigan in #581
  • Suppress benign OptionValue messages by @rhennigan in #583
  • Bugfix: Fixed an issue where FormatChatOutput could be parsed in the wrong context in dynamic output by @rhennigan in #586
  • Bugfix: Disable SideChat cell dingbat menu in cloud by @rhennigan in #588
  • Release v1.4.2 by @rhennigan in #590

Full Changelog: v1.4.1...v1.4.2

Release v1.4.1

22 Jan 21:43
99ef68c
Compare
Choose a tag to compare

What's Changed

  • Stop chat if there's a user abort by @rhennigan in #526
  • Bugfix: Do not include generated cells that appear after the chat input in the chat context by @rhennigan in #532
  • Preserve order of cells when inserting via code block button tray by @rhennigan in #533
  • Initial version of a chat preferences dialog by @rhennigan in #536
  • Fix a recursion error that occurs when there's no icon available for a registered service by @rhennigan in #537
  • Fix error message that occurs during build due to recent LLMFunctions paclet update by @rhennigan in #538
  • Fix unnecessary loading of package files when Chatbook subcontexts are loaded via Needs by @rhennigan in #540
  • Add "Services" tab to cloud preferences by @rhennigan in #541
  • Bugfix: Fix an internal failure that could occur in tool manager when model name was Automatic by @rhennigan in #544
  • Bugfix: Make the "Reset to Defaults" button work in cloud by @rhennigan in #545
  • Add "Personas" tab to cloud preferences by @rhennigan in #546
  • Split Tools.wl into a subpackage by @rhennigan in #549
  • Add "Tools" tab to cloud preferences by @rhennigan in #553
  • Add an "Insert Chat Cell" menu to the cloud toolbar by @rhennigan in #554
  • Bugfix: Use Import as a fallback for the WebFetcher tool when StartWebSession fails by @rhennigan in #556
  • Bugfix: FIx StopChat button failures in cloud by @rhennigan in #558
  • Bugfix: ChatOutput cells will now inherit cell tags from the corresponding chat input by @rhennigan in #559
  • Revert some temporary changes made to CodeAssistant on main branch during development by @rhennigan in #561
  • Release v1.4.1 by @rhennigan in #562

Full Changelog: v1.4.0...v1.4.1