Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Release v1.3.3 #488

Merged
merged 21 commits into from
Dec 5, 2023
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
21 commits
Select commit Hold shift + click to select a range
c73b0cc
Improve error handling for outer sandbox timeouts
rhennigan Nov 30, 2023
341161e
Increment paclet version
rhennigan Nov 30, 2023
c00ed82
Add tags to `Sow` and `Reap`
rhennigan Nov 30, 2023
f14046c
Merge pull request #480 from WolframResearch/479-a-wolframlanguageeva…
rhennigan Nov 30, 2023
5370ce5
Initial implementation for WolframLanguageEvaluator in cloud
rhennigan Nov 30, 2023
153d74e
Enable evaluator
rhennigan Nov 30, 2023
edb3754
Bugfix: Add missing serialization rule for cloud chat code blocks
rhennigan Nov 30, 2023
0419565
Merge pull request #481 from WolframResearch/feature/cloud-sandbox-ev…
rhennigan Dec 1, 2023
541270a
Avoid serializing giant tokenizers into tagging rules of chat outputs
rhennigan Dec 1, 2023
02895ea
Merge pull request #482 from WolframResearch/bugfix/serialized-chat-d…
rhennigan Dec 1, 2023
01bee30
Fix `tokenizerName["gpt-4-vision-preview"]` incorrectly giving `"gpt-…
rhennigan Dec 1, 2023
e2df35d
Merge pull request #483 from WolframResearch/bugfix/gpt4-vision-using…
rhennigan Dec 1, 2023
64926ba
Make `inlineInteractiveCodeCell` portable between cloud and desktop a…
rhennigan Dec 1, 2023
f3ca2d4
Merge pull request #484 from WolframResearch/bugfix/avoid-bouncing-co…
rhennigan Dec 1, 2023
a796112
Bugfix: dynamic box was preventing editing and code buttons were failing
rhennigan Dec 4, 2023
42257ff
Merge pull request #486 from WolframResearch/485-click-to-copy-curren…
rhennigan Dec 5, 2023
ddcff10
Use different URI schemes to inform the LLM what type of interactive …
rhennigan Dec 5, 2023
22bc34f
Bugfix: fix conflicting names between a prompt component and class
rhennigan Dec 5, 2023
b3be42d
Generate markdown URIs for video and audio boxes
rhennigan Dec 5, 2023
6cd24be
Missed renaming
rhennigan Dec 5, 2023
e812774
Merge pull request #487 from WolframResearch/feature/improved-uri-pro…
rhennigan Dec 5, 2023
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion PacletInfo.wl
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
PacletObject[ <|
"Name" -> "Wolfram/Chatbook",
"PublisherID" -> "Wolfram",
"Version" -> "1.3.2",
"Version" -> "1.3.3",
"WolframVersion" -> "13.3+",
"Description" -> "Wolfram Notebooks + LLMs",
"License" -> "MIT",
Expand Down
2 changes: 1 addition & 1 deletion Scripts/Common.wl
Original file line number Diff line number Diff line change
Expand Up @@ -141,7 +141,7 @@ releaseID[ dir_ ] := FirstCase[
(* ::Subsection::Closed:: *)
(*releaseURL*)
releaseURL[ file_ ] := Enclose[
Enclose @ Module[ { pac, repo, ver },
Module[ { pac, repo, ver },
pac = PacletObject @ Flatten @ File @ file;
repo = ConfirmBy[ Environment[ "GITHUB_REPOSITORY" ], StringQ ];
ver = ConfirmBy[ pac[ "Version" ], StringQ ];
Expand Down
113 changes: 93 additions & 20 deletions Source/Chatbook/ChatMessages.wl
Original file line number Diff line number Diff line change
Expand Up @@ -17,6 +17,7 @@ Wolfram`Chatbook`CellToChatMessage;
`constructMessages;
`expandMultimodalString;
`getTokenizer;
`getTokenizerName;
`resizeMultimodalImage;

Begin[ "`Private`" ];
Expand Down Expand Up @@ -70,6 +71,10 @@ $styleRoles = <|
"ChatSystemInput" -> "System"
|>;

$cachedTokenizerNames = { "chat-bison", "claude", "gpt-2", "gpt-3.5", "gpt-4-vision", "gpt-4" };
$cachedTokenizers = <| |>;
$fallbackTokenizer = "gpt-2";

(* ::**************************************************************************************************************:: *)
(* ::Section::Closed:: *)
(*CellToChatMessage*)
Expand Down Expand Up @@ -1079,58 +1084,120 @@ argumentTokenToString // endDefinition;
(* ::**************************************************************************************************************:: *)
(* ::Section::Closed:: *)
(*Tokenization*)
$tokenizer := gpt2Tokenizer;
$tokenizer := $gpt2Tokenizer;

(* ::**************************************************************************************************************:: *)
(* ::Subsection::Closed:: *)
(*getTokenizerName*)
getTokenizerName // beginDefinition;

getTokenizerName[ KeyValuePattern[ "TokenizerName"|"Tokenizer" -> name_String ] ] :=
tokenizerName @ name;

getTokenizerName[ KeyValuePattern[ "Tokenizer" -> Except[ $$unspecified ] ] ] :=
"Custom";

getTokenizerName[ KeyValuePattern[ "Model" -> model_ ] ] :=
With[ { name = tokenizerName @ toModelName @ model },
If[ MemberQ[ $cachedTokenizerNames, name ],
name,
$fallbackTokenizer
]
];

getTokenizerName // endDefinition;

(* ::**************************************************************************************************************:: *)
(* ::Subsection::Closed:: *)
(*getTokenizer*)
getTokenizer // beginDefinition;
getTokenizer[ KeyValuePattern[ "Tokenizer" -> tokenizer: Except[ $$unspecified ] ] ] := tokenizer;
getTokenizer[ KeyValuePattern[ "Model" -> model_ ] ] := getTokenizer @ model;
getTokenizer[ model_ ] := cachedTokenizer @ toModelName @ model;
getTokenizer[ KeyValuePattern[ "TokenizerName" -> name_String ] ] := cachedTokenizer @ name;
getTokenizer[ KeyValuePattern[ "Model" -> model_ ] ] := cachedTokenizer @ toModelName @ model;
getTokenizer // endDefinition;

(* ::**************************************************************************************************************:: *)
(* ::Subsubsection::Closed:: *)
(*cachedTokenizer*)
cachedTokenizer // beginDefinition;
cachedTokenizer[ All ] := AssociationMap[ cachedTokenizer, $cachedTokenizerNames ];
cachedTokenizer[ name_String ] := cachedTokenizer0 @ tokenizerName @ toModelName @ name;

cachedTokenizer[ All ] :=
AssociationMap[ cachedTokenizer, $cachedTokenizerNames ];

cachedTokenizer[ id_String ] :=
With[ { tokenizer = $cachedTokenizers[ tokenizerName @ toModelName @ id ] },
tokenizer /; ! MatchQ[ tokenizer, $$unspecified ]
];

cachedTokenizer[ id_String ] := Enclose[
Module[ { name, tokenizer },
name = ConfirmBy[ tokenizerName @ toModelName @ id, StringQ, "Name" ];
tokenizer = findTokenizer @ name;
If[ MissingQ @ tokenizer,
(* Fallback to the GPT-2 tokenizer: *)
tokenizer = ConfirmMatch[ $gpt2Tokenizer, Except[ $$unspecified ], "GPT2Tokenizer" ];
If[ TrueQ @ Wolfram`ChatbookInternal`$BuildingMX,
tokenizer, (* Avoid caching fallback values into MX definitions *)
cacheTokenizer[ name, tokenizer ]
],
cacheTokenizer[ name, ConfirmMatch[ tokenizer, Except[ $$unspecified ], "Tokenizer" ] ]
]
],
throwInternalFailure
];

cachedTokenizer // endDefinition;

(* ::**************************************************************************************************************:: *)
(* ::Subsubsubsection::Closed:: *)
(*cacheTokenizer*)
cacheTokenizer // beginDefinition;

cachedTokenizer0 // beginDefinition;
cacheTokenizer[ name_String, tokenizer: Except[ $$unspecified ] ] := (
$cachedTokenizerNames = Union[ $cachedTokenizerNames, { name } ];
$cachedTokenizers[ name ] = tokenizer
);

cachedTokenizer0[ "chat-bison" ] = ToCharacterCode[ #, "UTF8" ] &;
cacheTokenizer // endDefinition;

cachedTokenizer0[ "gpt-4-vision" ] :=
If[ graphicsQ[ # ],
gpt4ImageTokenizer[ # ],
cachedTokenizer[ "gpt-4" ][ # ]
] &;
(* ::**************************************************************************************************************:: *)
(* ::Subsubsubsection::Closed:: *)
(*findTokenizer*)
findTokenizer // beginDefinition;

cachedTokenizer0[ model_String ] := Enclose[
findTokenizer[ model_String ] := Enclose[
Quiet @ Module[ { name, tokenizer },
initTools[ ];
Quiet @ Needs[ "Wolfram`LLMFunctions`Utilities`Tokenization`" -> None ];
name = ConfirmBy[ tokens`FindTokenizer @ model, StringQ, "Name" ];
tokenizer = ConfirmMatch[ tokens`LLMTokenizer[ Method -> name ], Except[ _tokens`LLMTokenizer ], "Tokenizer" ];
ConfirmMatch[ tokenizer[ "test" ], _List, "TokenizerTest" ];
cachedTokenizer0[ model ] = tokenizer
tokenizer
],
gpt2Tokenizer &
Missing[ "NotFound" ] &
];

cachedTokenizer0 // endDefinition;
findTokenizer // endDefinition;

(* ::**************************************************************************************************************:: *)
(* ::Subsubsubsection::Closed:: *)
(*Pre-cached small tokenizer functions*)
$cachedTokenizers[ "chat-bison" ] = ToCharacterCode[ #, "UTF8" ] &;
$cachedTokenizers[ "gpt-4-vision" ] = If[ graphicsQ[ # ], gpt4ImageTokenizer[ # ], cachedTokenizer[ "gpt-4" ][ # ] ] &;

(* ::**************************************************************************************************************:: *)
(* ::Subsubsection::Closed:: *)
(*tokenizerName*)
tokenizerName // beginDefinition;
tokenizerName[ name_String ] := SelectFirst[ $cachedTokenizerNames, StringContainsQ[ name, # ] &, name ];
tokenizerName // endDefinition;

$cachedTokenizerNames = { "gpt-4-vision", "gpt-4", "gpt-3.5", "gpt-2", "claude-2", "claude-instant-1", "chat-bison" };
tokenizerName[ name_String ] :=
SelectFirst[
ReverseSortBy[ $cachedTokenizerNames, StringLength ],
StringContainsQ[ name, #, IgnoreCase -> True ] &,
name
];

tokenizerName // endDefinition;

(* ::**************************************************************************************************************:: *)
(* ::Subsection::Closed:: *)
Expand Down Expand Up @@ -1182,13 +1249,19 @@ gpt4ImageTokenCount0 // endDefinition;
(* ::**************************************************************************************************************:: *)
(* ::Subsection::Closed:: *)
(*Fallback Tokenizer*)
gpt2Tokenizer := gpt2Tokenizer = ResourceFunction[ "GPTTokenizer" ][ ];
$gpt2Tokenizer := $gpt2Tokenizer = gpt2Tokenizer[ ];

(* https://resources.wolframcloud.com/FunctionRepository/resources/GPTTokenizer *)
importResourceFunction[ gpt2Tokenizer, "GPTTokenizer" ];

(* ::**************************************************************************************************************:: *)
(* ::Section::Closed:: *)
(*Package Footer*)
If[ Wolfram`ChatbookInternal`$BuildingMX,
cachedTokenizer[ All ];
$gpt2Tokenizer;
(* This is only needed to generate $gpt2Tokenizer once, so it can be removed to reduce MX file size: *)
Remove[ "Wolfram`Chatbook`ResourceFunctions`GPTTokenizer`GPTTokenizer" ];
];

(* :!CodeAnalysis::EndBlock:: *)
Expand Down
51 changes: 50 additions & 1 deletion Source/Chatbook/Formatting.wl
Original file line number Diff line number Diff line change
Expand Up @@ -1137,11 +1137,12 @@ inlineInteractiveCodeCell // beginDefinition;

inlineInteractiveCodeCell[ display_, string_ ] /; $dynamicText := display;

(* TODO: make this switch dynamically depending on $cloudNotebooks (likely as a TemplateBox)*)
inlineInteractiveCodeCell[ display_, string_ ] :=
inlineInteractiveCodeCell[ display, string, contentLanguage @ string ];

inlineInteractiveCodeCell[ display_, string_, lang_ ] /; $cloudNotebooks :=
Mouseover[ display, Column @ { display, floatingButtonGrid[ string, lang ] } ];
cloudInlineInteractiveCodeCell[ display, string, lang ];

inlineInteractiveCodeCell[ display_, string_, lang_ ] :=
DynamicModule[ { $CellContext`attached, $CellContext`cell },
Expand All @@ -1167,6 +1168,54 @@ inlineInteractiveCodeCell[ display_, string_, lang_ ] :=

inlineInteractiveCodeCell // endDefinition;

(* ::**************************************************************************************************************:: *)
(* ::Subsubsection::Closed:: *)
(*cloudInlineInteractiveCodeCell*)
cloudInlineInteractiveCodeCell // beginDefinition;

cloudInlineInteractiveCodeCell[ display_, string_, lang_ ] :=
Module[ { padded, buttons },

padded = Pane[ display, ImageSize -> { { 100, Automatic }, { 30, Automatic } } ];

buttons = Framed[
floatingButtonGrid[ string, lang ],
Background -> White,
FrameMargins -> { { 1, 0 }, { 0, 1 } },
FrameStyle -> White,
ImageMargins -> 1,
RoundingRadius -> 3
];

Mouseover[
buttonOverlay[ padded, Invisible @ buttons ],
buttonOverlay[ padded, buttons ],
ContentPadding -> False,
FrameMargins -> 0,
ImageMargins -> 0,
ImageSize -> All
]
];

cloudInlineInteractiveCodeCell // endDefinition;

(* ::**************************************************************************************************************:: *)
(* ::Subsubsection::Closed:: *)
(*buttonOverlay*)
buttonOverlay // beginDefinition;

buttonOverlay[ a_, b_ ] := Overlay[
{ a, b },
All,
2,
Alignment -> { Left, Bottom },
ContentPadding -> False,
FrameMargins -> 0,
ImageMargins -> 0
];

buttonOverlay // endDefinition;

(* ::**************************************************************************************************************:: *)
(* ::Subsection::Closed:: *)
(*makeInlineCodeCell*)
Expand Down
1 change: 1 addition & 0 deletions Source/Chatbook/Main.wl
Original file line number Diff line number Diff line change
Expand Up @@ -6,6 +6,7 @@ BeginPackage[ "Wolfram`Chatbook`" ];
(* ::**************************************************************************************************************:: *)
(* ::Subsection::Closed:: *)
(*Declare Symbols*)
`$AvailableTools;
`$ChatHandlerData;
`$ChatPost;
`$ChatPre;
Expand Down
22 changes: 22 additions & 0 deletions Source/Chatbook/Prompting.wl
Original file line number Diff line number Diff line change
Expand Up @@ -35,6 +35,10 @@ $basePromptOrder = {
"Checkboxes",
"CheckboxesIndeterminate",
"ConversionFormatting",
"SpecialURI",
"SpecialURIAudio",
"SpecialURIVideo",
"SpecialURIDynamic",
"VisibleUserInput",
"TrivialCode",
"Packages",
Expand All @@ -51,6 +55,7 @@ $basePromptClasses = <|
"Math" -> { "MathExpressions" },
"Formatting" -> { "CodeBlocks", "MathExpressions", "EscapedCharacters" },
"MessageConversion" -> { "ConversionLargeOutputs", "ConversionGraphics", "ConversionFormatting" },
"SpecialURIs" -> { "SpecialURIAudio", "SpecialURIVideo", "SpecialURIDynamic" },
"All" -> $basePromptOrder
|>;

Expand All @@ -71,6 +76,10 @@ $basePromptDependencies = Append[ "GeneralInstructionsHeader" ] /@ <|
"ConversionGraphics" -> { "MessageConversionHeader" },
"MarkdownImageBox" -> { "MessageConversionHeader" },
"ConversionFormatting" -> { "MessageConversionHeader" },
"SpecialURI" -> { },
"SpecialURIAudio" -> { "SpecialURI" },
"SpecialURIVideo" -> { "SpecialURI" },
"SpecialURIDynamic" -> { "SpecialURI" },
"VisibleUserInput" -> { },
"TrivialCode" -> { },
"WolframSymbolCapitalization" -> { },
Expand Down Expand Up @@ -159,6 +168,19 @@ $basePromptComponents[ "ConversionFormatting" ] = "\
``Cell[TextData[{StyleBox[\"Styled\", FontSlant -> \"Italic\"], \" message\"}], \"ChatInput\"]`` \
becomes ``Styled message``.";

$basePromptComponents[ "SpecialURI" ] = "\
* You will occasionally see markdown links with special URI schemes, e.g. ![label](scheme://content-id) that represent \
interactive interface elements. You can use these in your responses to display the same elements to the user.";

$basePromptComponents[ "SpecialURIAudio" ] = "\
* ![label](audio://content-id) represents an interactive audio player.";

$basePromptComponents[ "SpecialURIVideo" ] = "\
* ![label](video://content-id) represents an interactive video player.";

$basePromptComponents[ "SpecialURIDynamic" ] = "\
* ![label](dynamic://content-id) represents an embedded dynamic UI.";

$basePromptComponents[ "VisibleUserInput" ] = "\
* The user can still see their input, so there's no need to repeat it in your response";

Expand Down
Loading