Merge branches 'staging' and 'staging' of https://github.com/SillyTavern/SillyTavern into staging

public/css/toggle-dependent.css

@@ -328,9 +328,7 @@ body.movingUI .drawer-content,
 body.movingUI #expression-holder,
 body.movingUI .zoomed_avatar,
 body.movingUI .draggable,
-body.movingUI #floatingPrompt,
+body.movingUI #floatingPrompt {
-body.movingUI #groupMemberListPopout,
-body.movingUI #summaryExtensionPopout {
     resize: both;
 }
 

public/index.html

@@ -122,11 +122,11 @@
                     <div class="flex-container" id="ai_response_configuration">
                         <div id="respective-presets-block" class="width100p">
                             <div id="kobold_api-presets">
-                                <h3><span data-i18n="kobldpresets">Kobold Presets</span>
+                                <h4 class="margin0"><span data-i18n="kobldpresets">Kobold Presets</span>
                                     <a href="https://docs.sillytavern.app/usage/api-connections/koboldai/" class="notes-link" target="_blank">
                                         <span class="note-link-span">?</span>
                                     </a>
-                                </h3>
+                                </h4>
                                 <div class="preset_buttons">
                                     <select id="settings_perset" data-preset-manager-for="kobold">
                                         <option value="gui" data-i18n="guikoboldaisettings">GUI KoboldAI Settings</option>
@@ -140,12 +140,12 @@
                                 </div>
                             </div>
                             <div id="novel_api-presets">
-                                <h3>
+                                <h4 class="margin0">
                                     <span data-i18n="novelaipreserts">NovelAI Presets</span>
                                     <a href="https://docs.sillytavern.app/usage/api-connections/novelai/" class="notes-link" target="_blank">
                                         <span class="note-link-span">?</span>
                                     </a>
-                                </h3>
+                                </h4>
                                 <div class="preset_buttons">
                                     <select id="settings_perset_novel" data-preset-manager-for="novel">
                                         <option value="gui" data-i18n="default">Default</option>
@@ -160,7 +160,7 @@
                             </div>
                             <div id="openai_api-presets">
                                 <div>
-                                    <h4><span data-i18n="openaipresets">Chat Completion Presets</span></h4>
+                                    <h4 class="margin0"><span data-i18n="openaipresets">Chat Completion Presets</span></h4>
                                     <div class="openai_preset_buttons">
                                         <select id="settings_perset_openai">
                                             <option value="gui" data-i18n="default">Default</option>
@@ -175,8 +175,7 @@
                                 </div>
                             </div>
                             <div id="textgenerationwebui_api-presets">
-                                <h3><span data-i18n="Text Gen WebUI (ooba/Mancer) presets">Text Gen WebUI (ooba/Mancer) presets</span>
+                                <h4 class="margin0"><span data-i18n="Text Gen WebUI (ooba/Mancer) presets">Text Gen WebUI (ooba/Mancer) presets</span></h4>
-                                </h3>
                                 <div class="preset_buttons">
                                     <select id="settings_preset_textgenerationwebui" data-preset-manager-for="textgenerationwebui">
                                     </select>
@@ -1755,7 +1754,7 @@
                 <div id="API-status-top" class="drawer-icon fa-solid fa-plug-circle-exclamation  closedIcon" title="API Connections" data-i18n="[title]API Connections"></div>
             </div>
             <div id="rm_api_block" class="drawer-content closedDrawer">
-                <h3 id="title_api">API</h3>
+                <h3 class="margin0" id="title_api">API</h3>
                 <div class="flex-container flexFlowColumn">
                     <div id="main-API-selector-block">
                         <select id="main_api">
@@ -1967,9 +1966,9 @@
                         </div>
                     </div>
                     <div id="openai_api" style="display: none;position: relative;">
-                        <h3 data-i18n="Chat Completion Source">
+                        <h4 class="margin0" data-i18n="Chat Completion Source">
                             Chat Completion Source
-                        </h3>
+                        </h4>
                         <select id="chat_completion_source">
                             <option value="openai">OpenAI</option>
                             <option value="windowai">Window AI</option>
@@ -2237,7 +2236,7 @@
                 <div class="drawer-icon fa-solid fa-font  closedIcon" title="AI Response Formatting" data-i18n="[title]AI Response Formatting"></div>
             </div>
             <div class="drawer-content">
-                <h3 data-i18n="Advanced Formatting">Advanced Formatting
+                <h3 class="margin0" data-i18n="Advanced Formatting">Advanced Formatting
                     <a href="https://docs.sillytavern.app/usage/core-concepts/advancedformatting/" class="notes-link" target="_blank">
                         <span class="note-link-span">?</span>
                     </a>
@@ -2620,7 +2619,7 @@
                         <div class="unchecked fa-solid fa-unlock "></div>
                         <div class="checked fa-solid fa-lock "></div>
                     </label>
-                    <h3>
+                    <h3 class="margin0">
                         <span data-i18n="Worlds/Lorebooks">Worlds/Lorebooks</span>
                         <a href="https://docs.sillytavern.app/usage/core-concepts/worldinfo/" class="notes-link" target="_blank">
                             <span class="note-link-span">?</span>
@@ -2797,7 +2796,7 @@
                     <div name="userSettingsRowOne" class="flex-container flexFlowRow alignitemscenter spaceBetween">
                         <div class="flex-container">
                             <div class="flex-container flexnowrap alignitemscenter">
-                                <h3><span data-i18n="User Settings">User Settings</span></h3>
+                                <h3 class="margin0"><span data-i18n="User Settings">User Settings</span></h3>
                                 <select id="ui_mode_select" class="margin0 widthNatural">
                                     <option value="0" data-i18n="Simple">Simple</option>
                                     <option value="1" data-i18n="Advanced">Advanced</option>
@@ -3273,7 +3272,7 @@
             </div>
             <div class="drawer-content closedDrawer">
                 <div class="flex-container wide100p alignitemscenter spaceBetween">
-                    <h3><span data-i18n="Persona Management">Persona Management</span></h3>
+                    <h3 class="margin0"><span data-i18n="Persona Management">Persona Management</span></h3>
                     <a href="https://docs.sillytavern.app/usage/core-concepts/personas/" target="_blank" data-i18n="How do I use this?">How do I use this?</a>
                     <div id="persona-management-block" class="flex-container wide100p">
                         <div class="flex1">
@@ -3674,7 +3673,7 @@
         <div id="dialogue_popup">
             <div id="dialogue_popup_holder">
                 <div id="dialogue_popup_text">
-                    <h3>text</h3>
+                    <h3 class="margin0">text</h3>
                 </div>
                 <textarea id="dialogue_popup_input" class="text_pole" rows="1"></textarea>
                 <div id="dialogue_popup_controls">
@@ -3687,7 +3686,7 @@
     <div id="character_popup" class="flex-container flexFlowColumn flexNoGap">
 
         <div id="character_popup_text">
-            <h3 id="character_popup_text_h3"></h3> <span data-i18n="Advanced Defininitions">- Advanced
+            <h3 id="character_popup_text_h3" class="margin0"></h3> <span data-i18n="Advanced Defininitions">- Advanced
                 Definitions</span>
         </div>
         <hr class="margin-bot-10px">
@@ -3868,7 +3867,7 @@
     <div id="scenario_override_template" class="template_element">
         <div class="scenario_override range-block flexFlowColumn flex-container">
             <div class="range-block-title title_restorable">
-                <h3><span data-i18n="Chat Scenario Override">Chat Scenario Override</span></h3>
+                <h3><span data-i18n="Chat Scenario Override" class="margin0">Chat Scenario Override</span></h3>
                 <div title="Remove" data-i18n="[title]Remove" class="menu_button fa-solid fa-trash-can remove_scenario_override"></div>
             </div>
             <div class="range-block-counter justifyLeft flex-container flexFlowColumn">
@@ -4580,6 +4579,11 @@
                             Tokens: <span id="extension_floating_prompt_token_counter">0</span>
                         </div>
 
+                        <label class="checkbox_label" for="extension_floating_allow_wi_scan">
+                            <input id="extension_floating_allow_wi_scan" type="checkbox" />
+                            <span data-i18n="Include in World Info Scanning">Include in World Info Scanning</span>
+                        </label>
+
                         <div class="floating_prompt_radio_group">
                             <label>
                                 <input type="radio" name="extension_floating_position" value="2" />
@@ -4776,6 +4780,7 @@
 
     <template id="generic_draggable_template">
         <div class="draggable">
+            <div class="dragTitle"></div>
             <div class="panelControlBar flex-container">
                 <div class="fa-solid fa-grip drag-grabber"></div>
                 <div class="fa-solid fa-circle-xmark dragClose"></div>
@@ -4833,4 +4838,4 @@
     </script>
 </body>
 
-</html>
+</html>

public/script.js

@@ -212,6 +212,7 @@ export {
     saveChat,
     messageFormatting,
     getExtensionPrompt,
+    getExtensionPromptByName,
     showSwipeButtons,
     hideSwipeButtons,
     changeMainAPI,
@@ -2069,6 +2070,15 @@ function getAllExtensionPrompts() {
     return value.length ? substituteParams(value) : '';
 }
 
+// Wrapper to fetch extension prompts by module name
+function getExtensionPromptByName(moduleName) {
+    if (moduleName) {
+        return substituteParams(extension_prompts[moduleName]?.value);
+    } else {
+        return;
+    }
+}
+
 function getExtensionPrompt(position = 0, depth = undefined, separator = "\n") {
     let extension_prompt = Object.keys(extension_prompts)
         .sort()

public/scripts/authors-note.js

@@ -235,6 +235,7 @@ function loadSettings() {
     chat_metadata[metadata_keys.depth] = chat_metadata[metadata_keys.depth] ?? extension_settings.note.defaultDepth ?? DEFAULT_DEPTH;
     $('#extension_floating_prompt').val(chat_metadata[metadata_keys.prompt]);
     $('#extension_floating_interval').val(chat_metadata[metadata_keys.interval]);
+    $('#extension_floating_allow_wi_scan').prop('checked', extension_settings.note.allowWIScan ?? false);
     $('#extension_floating_depth').val(chat_metadata[metadata_keys.depth]);
     $(`input[name="extension_floating_position"][value="${chat_metadata[metadata_keys.position]}"]`).prop('checked', true);
 
@@ -389,6 +390,11 @@ function onChatChanged() {
     $('#extension_floating_default_token_counter').text(tokenCounter3);
 }
 
+function onAllowWIScanCheckboxChanged() {
+    extension_settings.note.allowWIScan = !!$(this).prop('checked');
+    updateSettings();
+}
+
 /**
  * Inject author's note options and setup event listeners.
  */
@@ -402,6 +408,7 @@ export function initAuthorsNote() {
     $('#extension_floating_default').on('input', onExtensionFloatingDefaultInput);
     $('#extension_default_depth').on('input', onDefaultDepthInput);
     $('#extension_default_interval').on('input', onDefaultIntervalInput);
+    $('#extension_floating_allow_wi_scan').on('input', onAllowWIScanCheckboxChanged);
     $('input[name="extension_floating_position"]').on('change', onExtensionFloatingPositionInput);
     $('input[name="extension_default_position"]').on('change', onDefaultPositionInput);
     $('input[name="extension_floating_char_position"]').on('change', onExtensionFloatingCharPositionInput);

public/scripts/extensions/assets/index.js

@@ -123,8 +123,7 @@ function downloadAssetsList(url) {
 
                         $(`<i></i>`)
                             .append(element)
-                            .append(`<span>${displayName}</span>`)
+                            .append(`<div class="flex-container flexFlowColumn"><span>${displayName}</span><span>${description}</span></div>`)
-                            .append(`<span>${description}</span>`)
                             .appendTo(assetTypeMenu);
                     }
                     assetTypeMenu.appendTo("#assets_menu");

public/scripts/extensions/gallery/index.js

@@ -5,7 +5,7 @@ import {
     getRequestHeaders,
 } from "../../../script.js";
 import { selected_group } from "../../group-chats.js";
-import { loadFileToDocument } from "../../utils.js";
+import { loadFileToDocument, delay } from "../../utils.js";
 import { loadMovingUIState } from '../../power-user.js';
 import { dragElement } from '../../RossAscends-mods.js';
 import { registerSlashCommand } from "../../slash-commands.js";
@@ -109,6 +109,13 @@ async function initGallery(items, url) {
         let file = e.originalEvent.dataTransfer.files[0];
         uploadFile(file, url);  // Added url parameter to know where to upload
     });
+
+    //let images populate first
+    await delay(100)
+    //unset the height (which must be getting set by the gallery library at some point)
+    $("#dragGallery").css('height', 'unset');
+    //force a resize to make images display correctly
+    jQuery("#dragGallery").nanogallery2('resize');
 }
 
 /**
@@ -247,14 +254,16 @@ $(document).ready(function () {
  * The cloned element has its attributes set, a new child div appended, and is made visible on the body.
  * Additionally, it sets up the element to prevent dragging on its images.
  */
-function makeMovable(id="gallery"){
+function makeMovable(id = "gallery") {
 
     console.debug('making new container from template')
     const template = $('#generic_draggable_template').html();
     const newElement = $(template);
+    newElement.css('background-color', 'var(--SmartThemeBlurTintColor)');
     newElement.attr('forChar', id);
     newElement.attr('id', `${id}`);
     newElement.find('.drag-grabber').attr('id', `${id}header`);
+    newElement.find('.dragTitle').text('Image Gallery')
     //add a div for the gallery
     newElement.append(`<div id="dragGallery"></div>`);
     // add no-scrollbar class to this element
@@ -326,6 +335,8 @@ function makeDragImg(id, url) {
 
         // Ensure that the newly added element is displayed as block
         draggableElem.style.display = 'block';
+        //and has no padding unlike other non-zoomed-avatar draggables
+        draggableElem.style.padding = '0';
 
         // Add an id to the close button
         // If the close button exists, set related-id
@@ -375,11 +386,11 @@ function makeDragImg(id, url) {
  * @param {string} id - The ID to be sanitized.
  * @returns {string} - The sanitized ID.
  */
-function sanitizeHTMLId(id){
+function sanitizeHTMLId(id) {
     // Replace spaces and non-word characters
     id = id.replace(/\s+/g, '-')
-           .replace(/[^\x00-\x7F]/g, '-')
+        .replace(/[^\x00-\x7F]/g, '-')
-           .replace(/\W/g, '');
+        .replace(/\W/g, '');
 
     return id;
 }

public/scripts/extensions/memory/index.js

@@ -571,12 +571,13 @@ function doPopout(e) {
         const template = $('#zoomed_avatar_template').html();
         const controlBarHtml = `<div class="panelControlBar flex-container">
         <div id="summaryExtensionPopoutheader" class="fa-solid fa-grip drag-grabber hoverglow"></div>
-        <div id="summaryExtensionPopoutClose" class="fa-solid fa-circle-xmark hoverglow"></div>
+        <div id="summaryExtensionPopoutClose" class="fa-solid fa-circle-xmark hoverglow dragClose"></div>
     </div>`
         const newElement = $(template);
-        newElement.attr('id', 'summaryExtensionPopout');
+        newElement.attr('id', 'summaryExtensionPopout')
-        newElement.removeClass('zoomed_avatar')
+            .removeClass('zoomed_avatar')
-        newElement.empty()
+            .addClass('draggable')
+            .empty()
         const prevSummaryBoxContents = $('#memory_contents').val(); //copy summary box before emptying
         originalElement.empty();
         originalElement.html(`<div class="flex-container alignitemscenter justifyCenter wide100p"><small>Currently popped out</small></div>`)

public/scripts/extensions/speech-recognition/browser.js

@@ -1,233 +0,0 @@
-// Borrowed from Agnai (AGPLv3)
-// https://github.com/agnaistic/agnai/blob/dev/web/pages/Chat/components/SpeechRecognitionRecorder.tsx
-// First version by Cohee#1207
-// Adapted by Tony-sama
-
-export { BrowserSttProvider }
-
-const DEBUG_PREFIX = "<Speech Recognition module (Browser)> "
-
-class BrowserSttProvider {
-    //########//
-    // Config //
-    //########//
-
-    settings = {
-        language: ""
-    }
-
-    defaultSettings = {
-        language: "en-US",
-    }
-
-    processTranscriptFunction = null;
-
-    get settingsHtml() {
-        let html = ' \
-        <span>Language</span> </br> \
-        <select id="speech_recognition_browser_provider_language"> \
-            <option value="ar-SA">ar-SA: Arabic (Saudi Arabia)</option> \
-            <option value="bn-BD">bn-BD: Bangla (Bangladesh)</option> \
-            <option value="bn-IN">bn-IN: Bangla (India)</option> \
-            <option value="cs-CZ">cs-CZ: Czech (Czech Republic)</option> \
-            <option value="da-DK">da-DK: Danish (Denmark)</option> \
-            <option value="de-AT">de-AT: German (Austria)</option> \
-            <option value="de-CH">de-CH: German (Switzerland)</option> \
-            <option value="de-DE">de-DE: German (Germany)</option> \
-            <option value="el-GR">el-GR: Greek (Greece)</option> \
-            <option value="en-AU">en-AU: English (Australia)</option> \
-            <option value="en-CA">en-CA: English (Canada)</option> \
-            <option value="en-GB">en-GB: English (United Kingdom)</option> \
-            <option value="en-IE">en-IE: English (Ireland)</option> \
-            <option value="en-IN">en-IN: English (India)</option> \
-            <option value="en-NZ">en-NZ: English (New Zealand)</option> \
-            <option value="en-US">en-US: English (United States)</option> \
-            <option value="en-ZA">en-ZA: English (South Africa)</option> \
-            <option value="es-AR">es-AR: Spanish (Argentina)</option> \
-            <option value="es-CL">es-CL: Spanish (Chile)</option> \
-            <option value="es-CO">es-CO: Spanish (Columbia)</option> \
-            <option value="es-ES">es-ES: Spanish (Spain)</option> \
-            <option value="es-MX">es-MX: Spanish (Mexico)</option> \
-            <option value="es-US">es-US: Spanish (United States)</option> \
-            <option value="fi-FI">fi-FI: Finnish (Finland)</option> \
-            <option value="fr-BE">fr-BE: French (Belgium)</option> \
-            <option value="fr-CA">fr-CA: French (Canada)</option> \
-            <option value="fr-CH">fr-CH: French (Switzerland)</option> \
-            <option value="fr-FR">fr-FR: French (France)</option> \
-            <option value="he-IL">he-IL: Hebrew (Israel)</option> \
-            <option value="hi-IN">hi-IN: Hindi (India)</option> \
-            <option value="hu-HU">hu-HU: Hungarian (Hungary)</option> \
-            <option value="id-ID">id-ID: Indonesian (Indonesia)</option> \
-            <option value="it-CH">it-CH: Italian (Switzerland)</option> \
-            <option value="it-IT">it-IT: Italian (Italy)</option> \
-            <option value="ja-JP">ja-JP: Japanese (Japan)</option> \
-            <option value="ko-KR">ko-KR: Korean (Republic of Korea)</option> \
-            <option value="nl-BE">nl-BE: Dutch (Belgium)</option> \
-            <option value="nl-NL">nl-NL: Dutch (The Netherlands)</option> \
-            <option value="no-NO">no-NO: Norwegian (Norway)</option> \
-            <option value="pl-PL">pl-PL: Polish (Poland)</option> \
-            <option value="pt-BR">pt-BR: Portugese (Brazil)</option> \
-            <option value="pt-PT">pt-PT: Portugese (Portugal)</option> \
-            <option value="ro-RO">ro-RO: Romanian (Romania)</option> \
-            <option value="ru-RU">ru-RU: Russian (Russian Federation)</option> \
-            <option value="sk-SK">sk-SK: Slovak (Slovakia)</option> \
-            <option value="sv-SE">sv-SE: Swedish (Sweden)</option> \
-            <option value="ta-IN">ta-IN: Tamil (India)</option> \
-            <option value="ta-LK">ta-LK: Tamil (Sri Lanka)</option> \
-            <option value="th-TH">th-TH: Thai (Thailand)</option> \
-            <option value="tr-TR">tr-TR: Turkish (Turkey)</option> \
-            <option value="zh-CN">zh-CN: Chinese (China)</option> \
-            <option value="zh-HK">zh-HK: Chinese (Hond Kong)</option> \
-            <option value="zh-TW">zh-TW: Chinese (Taiwan)</option> \
-        </select> \
-        '
-        return html
-    }
-
-    onSettingsChange() {
-        // Used when provider settings are updated from UI
-        this.settings.language = $("#speech_recognition_browser_provider_language").val();
-        console.debug(DEBUG_PREFIX+"Change language to",this.settings.language);
-        this.loadSettings(this.settings);
-    }
-
-    static capitalizeInterim(interimTranscript) {
-        let capitalizeIndex = -1;
-        if (interimTranscript.length > 2 && interimTranscript[0] === ' ') capitalizeIndex = 1;
-        else if (interimTranscript.length > 1) capitalizeIndex = 0;
-        if (capitalizeIndex > -1) {
-            const spacing = capitalizeIndex > 0 ? ' '.repeat(capitalizeIndex - 1) : '';
-            const capitalized = interimTranscript[capitalizeIndex].toLocaleUpperCase();
-            const rest = interimTranscript.substring(capitalizeIndex + 1);
-            interimTranscript = spacing + capitalized + rest;
-        }
-        return interimTranscript;
-    }
-    
-    static composeValues(previous, interim) {
-        let spacing = '';
-        if (previous.endsWith('.')) spacing = ' ';
-        return previous + spacing + interim;
-    }
-
-    loadSettings(settings) {
-        const processTranscript = this.processTranscriptFunction;
-        
-        // Populate Provider UI given input settings
-        if (Object.keys(settings).length == 0) {
-            console.debug(DEBUG_PREFIX+"Using default browser STT settings")
-        }
-
-        // Initialise as defaultSettings
-        this.settings = this.defaultSettings;
-
-        for (const key in settings){
-            if (key in this.settings){
-                this.settings[key] = settings[key]
-            } else {
-                throw `Invalid setting passed to Speech recogniton extension (browser): ${key}`
-            }
-        }
-
-        $("#speech_recognition_browser_provider_language").val(this.settings.language);
-
-        const speechRecognitionSettings = $.extend({
-            grammar: '' // Custom grammar
-        }, options);
-
-        const speechRecognition = window.SpeechRecognition || window.webkitSpeechRecognition;
-        const speechRecognitionList = window.SpeechGrammarList || window.webkitSpeechGrammarList;
-
-        if (!speechRecognition) {
-            console.warn(DEBUG_PREFIX+'Speech recognition is not supported in this browser.');
-            $("#microphone_button").hide();
-            toastr.error("Speech recognition is not supported in this browser, use another browser or another provider of SillyTavern-extras Speech recognition extension.", "Speech recognition activation Failed (Browser)", { timeOut: 10000, extendedTimeOut: 20000, preventDuplicates: true });
-            return;
-        }
-
-        const recognition = new speechRecognition();
-
-        if (speechRecognitionSettings.grammar && speechRecognitionList) {
-            speechRecognitionList.addFromString(speechRecognitionSettings.grammar, 1);
-            recognition.grammars = speechRecognitionList;
-        }
-
-        recognition.continuous = true;
-        recognition.interimResults = true;
-        recognition.lang = this.settings.language;
-
-        const textarea = $('#send_textarea');
-        const button = $('#microphone_button');
-
-        let listening = false;
-        button.off('click').on("click", function () {
-            if (listening) {
-                recognition.stop();
-            } else {
-                recognition.start();
-            }
-            listening = !listening;
-        });
-
-        let initialText = '';
-
-        recognition.onresult = function (speechEvent) {
-            let finalTranscript = '';
-            let interimTranscript = ''
-
-            for (let i = speechEvent.resultIndex; i < speechEvent.results.length; ++i) {
-            const transcript = speechEvent.results[i][0].transcript;
-
-            if (speechEvent.results[i].isFinal) {
-                let interim = BrowserSttProvider.capitalizeInterim(transcript);
-                if (interim != '') {
-                let final = finalTranscript;
-                final = BrowserSttProvider.composeValues(final, interim);
-                if (final.slice(-1) != '.' & final.slice(-1) != '?') final += '.';
-                finalTranscript = final;
-                recognition.abort();
-                listening = false;
-                }
-                interimTranscript = ' ';
-            } else {
-                interimTranscript += transcript;
-            }
-            }
-
-            interimTranscript = BrowserSttProvider.capitalizeInterim(interimTranscript);
-                
-            textarea.val(initialText + finalTranscript + interimTranscript);
-        };
-
-        recognition.onerror = function (event) {
-            console.error('Error occurred in recognition:', event.error);
-            //if ($('#speech_recognition_debug').is(':checked'))
-            //    toastr.error('Error occurred in recognition:'+ event.error, 'STT Generation error (Browser)', { timeOut: 10000, extendedTimeOut: 20000, preventDuplicates: true });
-        };
-
-        recognition.onend = function () {
-            listening = false;
-            button.toggleClass('fa-microphone fa-microphone-slash');
-            const newText = textarea.val().substring(initialText.length);
-            textarea.val(textarea.val().substring(0,initialText.length));
-            processTranscript(newText);
-
-        };
-
-        recognition.onstart = function () {
-            initialText = textarea.val();
-            button.toggleClass('fa-microphone fa-microphone-slash');
-            
-            if ($("#speech_recognition_message_mode").val() == "replace") {
-                textarea.val("");
-                initialText = ""
-            }
-        };
-        
-        $("#microphone_button").show();
-        
-        console.debug(DEBUG_PREFIX+"Browser STT settings loaded")
-    }
-
-
-}

public/scripts/extensions/speech-recognition/index.js

@@ -1,452 +0,0 @@
-/*
-TODO:
- - try pseudo streaming audio by just sending chunk every X seconds and asking VOSK if it is full text.
-*/
-
-import { saveSettingsDebounced } from "../../../script.js";
-import { getContext, extension_settings, ModuleWorkerWrapper } from "../../extensions.js";
-import { VoskSttProvider } from './vosk.js'
-import { WhisperSttProvider } from './whisper.js'
-import { BrowserSttProvider } from './browser.js'
-import { StreamingSttProvider } from './streaming.js'
-import { getMessageTimeStamp } from "../../RossAscends-mods.js";
-export { MODULE_NAME };
-
-const MODULE_NAME = 'Speech Recognition';
-const DEBUG_PREFIX = "<Speech Recognition module> "
-const UPDATE_INTERVAL = 100;
-
-let inApiCall = false;
-
-let sttProviders = {
-    None: null,
-    Browser: BrowserSttProvider,
-    Whisper: WhisperSttProvider,
-    Vosk: VoskSttProvider,
-    Streaming: StreamingSttProvider,
-}
-
-let sttProvider = null
-let sttProviderName = "None"
-
-let audioRecording = false
-const constraints = { audio: { sampleSize: 16, channelCount: 1, sampleRate: 16000 } };
-let audioChunks = [];
-
-async function moduleWorker() {
-    if (sttProviderName != "Streaming") {
-        return;
-    }
-
-    // API is busy
-    if (inApiCall) {
-        return;
-    }
-
-    try {
-        inApiCall = true;
-        const userMessageOriginal =  await sttProvider.getUserMessage();
-        let userMessageFormatted = userMessageOriginal.trim();
-
-        if (userMessageFormatted.length > 0)
-        {
-            console.debug(DEBUG_PREFIX+"recorded transcript: \""+userMessageFormatted+"\"");
-
-            let userMessageLower = userMessageFormatted.toLowerCase();
-            // remove punctuation
-            let userMessageRaw = userMessageLower.replace(/[^\w\s\']|_/g, "").replace(/\s+/g, " ");
-
-            console.debug(DEBUG_PREFIX+"raw transcript:",userMessageRaw);
-
-            // Detect trigger words
-            let messageStart = -1;
-
-            if (extension_settings.speech_recognition.Streaming.triggerWordsEnabled) {
-
-                for (const triggerWord of extension_settings.speech_recognition.Streaming.triggerWords) {
-                    const triggerPos = userMessageRaw.indexOf(triggerWord.toLowerCase());
-
-                    // Trigger word not found or not starting message and just a substring
-                    if (triggerPos == -1){ // | (triggerPos > 0 & userMessageFormatted[triggerPos-1] != " ")) {
-                        console.debug(DEBUG_PREFIX+"trigger word not found: ", triggerWord);
-                    }
-                    else {
-                        console.debug(DEBUG_PREFIX+"Found trigger word: ", triggerWord, " at index ", triggerPos);
-                        if (triggerPos < messageStart || messageStart == -1) { // & (triggerPos + triggerWord.length) < userMessageFormatted.length)) {
-                            messageStart = triggerPos; // + triggerWord.length + 1;
-
-                            if (!extension_settings.speech_recognition.Streaming.triggerWordsIncluded)
-                                messageStart = triggerPos + triggerWord.length + 1;
-                        }
-                    }
-                }
-            } else {
-                messageStart = 0;
-            }
-
-            if (messageStart == -1) {
-                console.debug(DEBUG_PREFIX+"message ignored, no trigger word preceding a message. Voice transcript: \""+ userMessageOriginal +"\"");
-                if (extension_settings.speech_recognition.Streaming.debug) {
-                    toastr.info(
-                        "No trigger word preceding a message. Voice transcript: \""+ userMessageOriginal +"\"",
-                        DEBUG_PREFIX+"message ignored.",
-                        { timeOut: 10000, extendedTimeOut: 20000, preventDuplicates: true },
-                    );
-                }
-            }
-            else{
-                userMessageFormatted = userMessageFormatted.substring(messageStart);
-                // Trim non alphanumeric character from the start
-                messageStart = 0;
-                for(const i of userMessageFormatted) {
-                    if(/^[a-z]$/i.test(i)) {
-                        break;
-                    }
-                    messageStart += 1;
-                }
-                userMessageFormatted = userMessageFormatted.substring(messageStart);
-                userMessageFormatted = userMessageFormatted.charAt(0).toUpperCase() + userMessageFormatted.substring(1);
-                processTranscript(userMessageFormatted);
-            }
-        }
-        else
-        {
-            console.debug(DEBUG_PREFIX+"Received empty transcript, ignored");
-        }
-    }
-    catch (error) {
-        console.debug(error);
-    }
-    finally {
-        inApiCall = false;
-    }
-}
-
-async function processTranscript(transcript) {
-    try {
-        const transcriptOriginal =  transcript;
-        let transcriptFormatted = transcriptOriginal.trim();
-
-        if (transcriptFormatted.length > 0)
-        {
-            console.debug(DEBUG_PREFIX+"recorded transcript: \""+transcriptFormatted+"\"");
-            const messageMode = extension_settings.speech_recognition.messageMode;
-            console.debug(DEBUG_PREFIX+"mode: "+messageMode);
-
-            let transcriptLower = transcriptFormatted.toLowerCase()
-            // remove punctuation
-            let transcriptRaw = transcriptLower.replace(/[^\w\s\']|_/g, "").replace(/\s+/g, " ");
-
-            // Check message mapping
-            if (extension_settings.speech_recognition.messageMappingEnabled) {
-                console.debug(DEBUG_PREFIX+"Start searching message mapping into:",transcriptRaw)
-                for (const key in extension_settings.speech_recognition.messageMapping) {
-                    console.debug(DEBUG_PREFIX+"message mapping searching: ", key,"=>",extension_settings.speech_recognition.messageMapping[key]);
-                    if (transcriptRaw.includes(key)) {
-                        var message = extension_settings.speech_recognition.messageMapping[key];
-                        console.debug(DEBUG_PREFIX+"message mapping found: ", key,"=>",extension_settings.speech_recognition.messageMapping[key]);
-                        $("#send_textarea").val(message);
-
-                        if (messageMode == "auto_send") await getContext().generate();
-                        return;
-                    }
-                }
-            }
-
-            console.debug(DEBUG_PREFIX+"no message mapping found, processing transcript as normal message");
-
-            switch (messageMode) {
-                case "auto_send":
-                    $('#send_textarea').val("") // clear message area to avoid double message
-
-                    console.debug(DEBUG_PREFIX+"Sending message")
-                    const context = getContext();
-                    const messageText = transcriptFormatted;
-                    const message = {
-                        name: context.name1,
-                        is_user: true,
-                        send_date: getMessageTimeStamp(),
-                        mes: messageText,
-                    };
-                    context.chat.push(message);
-                    context.addOneMessage(message);
-
-                    await context.generate();
-
-                    $('#debug_output').text("<SST-module DEBUG>: message sent: \""+ transcriptFormatted +"\"");
-                    break;
-
-                case "replace":
-                    console.debug(DEBUG_PREFIX+"Replacing message")
-                    $('#send_textarea').val(transcriptFormatted);
-                    break;
-
-                case "append":
-                    console.debug(DEBUG_PREFIX+"Appending message")
-                    $('#send_textarea').val($('#send_textarea').val()+" "+transcriptFormatted);
-                    break;
-
-                default:
-                    console.debug(DEBUG_PREFIX+"Not supported stt message mode: "+messageMode)
-
-            }
-        }
-        else
-        {
-            console.debug(DEBUG_PREFIX+"Empty transcript, do nothing");
-        }
-    }
-    catch (error) {
-        console.debug(error);
-    }
-}
-
-function loadNavigatorAudioRecording() {
-    if (navigator.mediaDevices.getUserMedia) {
-        console.debug(DEBUG_PREFIX+' getUserMedia supported by browser.');
-
-        let onSuccess = function(stream) {
-          const mediaRecorder = new MediaRecorder(stream);
-
-          $("#microphone_button").off('click').on("click", function() {
-            if (!audioRecording) {
-                mediaRecorder.start();
-                console.debug(mediaRecorder.state);
-                console.debug("recorder started");
-                audioRecording = true;
-                $("#microphone_button").toggleClass('fa-microphone fa-microphone-slash');
-            }
-            else {
-                mediaRecorder.stop();
-                console.debug(mediaRecorder.state);
-                console.debug("recorder stopped");
-                audioRecording = false;
-                $("#microphone_button").toggleClass('fa-microphone fa-microphone-slash');
-            }
-          });
-
-          mediaRecorder.onstop = async function() {
-            console.debug(DEBUG_PREFIX+"data available after MediaRecorder.stop() called: ", audioChunks.length, " chunks");
-            const audioBlob = new Blob(audioChunks, { type: "audio/wav; codecs=0" });
-            audioChunks = [];
-
-            const transcript = await sttProvider.processAudio(audioBlob);
-
-            // TODO: lock and release recording while processing?
-            console.debug(DEBUG_PREFIX+"received transcript:", transcript);
-            processTranscript(transcript);
-          }
-
-          mediaRecorder.ondataavailable = function(e) {
-            audioChunks.push(e.data);
-          }
-        }
-
-        let onError = function(err) {
-          console.debug(DEBUG_PREFIX+"The following error occured: " + err);
-        }
-
-        navigator.mediaDevices.getUserMedia(constraints).then(onSuccess, onError);
-
-      } else {
-         console.debug(DEBUG_PREFIX+"getUserMedia not supported on your browser!");
-         toastr.error("getUserMedia not supported", DEBUG_PREFIX+"not supported for your browser.", { timeOut: 10000, extendedTimeOut: 20000, preventDuplicates: true });
-      }
-}
-
-//##############//
-// STT Provider //
-//##############//
-
-function loadSttProvider(provider) {
-    //Clear the current config and add new config
-    $("#speech_recognition_provider_settings").html("");
-
-    // Init provider references
-    extension_settings.speech_recognition.currentProvider = provider;
-    sttProviderName = provider;
-
-    if (!(sttProviderName in extension_settings.speech_recognition)) {
-        console.warn(`Provider ${sttProviderName} not in Extension Settings, initiatilizing provider in settings`);
-        extension_settings.speech_recognition[sttProviderName] = {};
-    }
-
-    $('#speech_recognition_provider').val(sttProviderName);
-
-    if (sttProviderName == "None") {
-        $("#microphone_button").hide();
-        $("#speech_recognition_message_mode_div").hide();
-        $("#speech_recognition_message_mapping_div").hide();
-        return;
-    }
-
-    $("#speech_recognition_message_mode_div").show();
-    $("#speech_recognition_message_mapping_div").show();
-
-    sttProvider = new sttProviders[sttProviderName]
-
-    // Init provider settings
-    $('#speech_recognition_provider_settings').append(sttProvider.settingsHtml);
-
-    // Use microphone button as push to talk
-    if (sttProviderName == "Browser") {
-        sttProvider.processTranscriptFunction = processTranscript;
-        sttProvider.loadSettings(extension_settings.speech_recognition[sttProviderName]);
-        $("#microphone_button").show();
-    }
-
-    if (sttProviderName == "Vosk" | sttProviderName == "Whisper") {
-        sttProvider.loadSettings(extension_settings.speech_recognition[sttProviderName]);
-        loadNavigatorAudioRecording();
-        $("#microphone_button").show();
-    }
-
-    if (sttProviderName == "Streaming") {
-        sttProvider.loadSettings(extension_settings.speech_recognition[sttProviderName]);
-        $("#microphone_button").off('click');
-        $("#microphone_button").hide();
-    }
-
-}
-
-function onSttProviderChange() {
-    const sttProviderSelection = $('#speech_recognition_provider').val();
-    loadSttProvider(sttProviderSelection);
-    saveSettingsDebounced();
-}
-
-function onSttProviderSettingsInput() {
-    sttProvider.onSettingsChange();
-
-    // Persist changes to SillyTavern stt extension settings
-    extension_settings.speech_recognition[sttProviderName] = sttProvider.settings;
-    saveSettingsDebounced();
-    console.info(`Saved settings ${sttProviderName} ${JSON.stringify(sttProvider.settings)}`);
-}
-
-//#############################//
-//  Extension UI and Settings  //
-//#############################//
-
-const defaultSettings = {
-    currentProvider: "None",
-    messageMode: "append",
-    messageMappingText: "",
-    messageMapping: [],
-    messageMappingEnabled: false,
-}
-
-function loadSettings() {
-    if (Object.keys(extension_settings.speech_recognition).length === 0) {
-        Object.assign(extension_settings.speech_recognition, defaultSettings)
-    }
-    $('#speech_recognition_enabled').prop('checked',extension_settings.speech_recognition.enabled);
-    $('#speech_recognition_message_mode').val(extension_settings.speech_recognition.messageMode);
-
-    if (extension_settings.speech_recognition.messageMappingText.length > 0) {
-        $('#speech_recognition_message_mapping').val(extension_settings.speech_recognition.messageMappingText);
-    }
-
-    $('#speech_recognition_message_mapping_enabled').prop('checked',extension_settings.speech_recognition.messageMappingEnabled);
-}
-
-async function onMessageModeChange() {
-    extension_settings.speech_recognition.messageMode = $('#speech_recognition_message_mode').val();
-
-    if(sttProviderName != "Browser" & extension_settings.speech_recognition.messageMode == "auto_send") {
-        $("#speech_recognition_wait_response_div").show()
-    }
-    else {
-        $("#speech_recognition_wait_response_div").hide()
-    }
-
-    saveSettingsDebounced();
-}
-
-async function onMessageMappingChange() {
-    let array = $('#speech_recognition_message_mapping').val().split(",");
-    array = array.map(element => {return element.trim();});
-    array = array.filter((str) => str !== '');
-    extension_settings.speech_recognition.messageMapping = {};
-    for (const text of array) {
-        if (text.includes("=")) {
-            const pair = text.toLowerCase().split("=")
-            extension_settings.speech_recognition.messageMapping[pair[0].trim()] = pair[1].trim()
-            console.debug(DEBUG_PREFIX+"Added mapping", pair[0],"=>", extension_settings.speech_recognition.messageMapping[pair[0]]);
-        }
-        else {
-            console.debug(DEBUG_PREFIX+"Wrong syntax for message mapping, no '=' found in:", text);
-        }
-    }
-
-    $("#speech_recognition_message_mapping_status").text("Message mapping updated to: "+JSON.stringify(extension_settings.speech_recognition.messageMapping))
-    console.debug(DEBUG_PREFIX+"Updated message mapping", extension_settings.speech_recognition.messageMapping);
-    extension_settings.speech_recognition.messageMappingText = $('#speech_recognition_message_mapping').val()
-    saveSettingsDebounced();
-}
-
-async function onMessageMappingEnabledClick() {
-    extension_settings.speech_recognition.messageMappingEnabled = $('#speech_recognition_message_mapping_enabled').is(':checked');
-    saveSettingsDebounced()
-}
-
-$(document).ready(function () {
-    function addExtensionControls() {
-        const settingsHtml = `
-        <div id="speech_recognition_settings">
-            <div class="inline-drawer">
-                <div class="inline-drawer-toggle inline-drawer-header">
-                    <b>Speech Recognition</b>
-                    <div class="inline-drawer-icon fa-solid fa-circle-chevron-down down"></div>
-                </div>
-                <div class="inline-drawer-content">
-                    <div>
-                        <span>Select Speech-to-text Provider</span> </br>
-                        <select id="speech_recognition_provider">
-                        </select>
-                    </div>
-                    <div id="speech_recognition_message_mode_div">
-                        <span>Message Mode</span> </br>
-                        <select id="speech_recognition_message_mode">
-                            <option value="append">Append</option>
-                            <option value="replace">Replace</option>
-                            <option value="auto_send">Auto send</option>
-                        </select>
-                    </div>
-                    <div id="speech_recognition_message_mapping_div">
-                        <span>Message Mapping</span>
-                        <textarea id="speech_recognition_message_mapping" class="text_pole textarea_compact" type="text" rows="4" placeholder="Enter comma separated phrases mapping, example:\ncommand delete = /del 2,\nslash delete = /del 2,\nsystem roll = /roll 2d6,\nhey continue = /continue"></textarea>
-                        <span id="speech_recognition_message_mapping_status"></span>
-                        <label class="checkbox_label" for="speech_recognition_message_mapping_enabled">
-                            <input type="checkbox" id="speech_recognition_message_mapping_enabled" name="speech_recognition_message_mapping_enabled">
-                            <small>Enable messages mapping</small>
-                        </label>
-                    </div>
-                    <form id="speech_recognition_provider_settings" class="inline-drawer-content">
-                    </form>
-                </div>
-            </div>
-        </div>
-        `;
-        $('#extensions_settings').append(settingsHtml);
-        $('#speech_recognition_provider_settings').on('input', onSttProviderSettingsInput);
-        for (const provider in sttProviders) {
-            $('#speech_recognition_provider').append($("<option />").val(provider).text(provider));
-            console.debug(DEBUG_PREFIX+"added option "+provider);
-        }
-        $('#speech_recognition_provider').on('change', onSttProviderChange);
-        $('#speech_recognition_message_mode').on('change', onMessageModeChange);
-        $('#speech_recognition_message_mapping').on('change', onMessageMappingChange);
-        $('#speech_recognition_message_mapping_enabled').on('click', onMessageMappingEnabledClick);
-
-        const $button = $('<div id="microphone_button" class="fa-solid fa-microphone speech-toggle" title="Click to speak"></div>');
-        $('#send_but_sheld').prepend($button);
-
-    }
-    addExtensionControls(); // No init dependencies
-    loadSettings(); // Depends on Extension Controls and loadTtsProvider
-    loadSttProvider(extension_settings.speech_recognition.currentProvider); // No dependencies
-    const wrapper = new ModuleWorkerWrapper(moduleWorker);
-    setInterval(wrapper.update.bind(wrapper), UPDATE_INTERVAL); // Init depends on all the things
-    moduleWorker();
-})

public/scripts/extensions/speech-recognition/manifest.json

@@ -1,14 +0,0 @@
-{
-    "display_name": "Speech Recognition",
-    "loading_order": 13,
-    "requires": [],
-    "optional": [
-        "vosk-speech-recognition",
-        "whisper-speech-recognition"
-    ],
-    "js": "index.js",
-    "css": "style.css",
-    "author": "Cohee#1207 and Keij#6799",
-    "version": "1.1.0",
-    "homePage": "https://github.com/SillyTavern/SillyTavern"
-}

public/scripts/extensions/speech-recognition/streaming.js

@@ -1,109 +0,0 @@
-import { getApiUrl, doExtrasFetch, modules } from "../../extensions.js";
-export { StreamingSttProvider }
-
-const DEBUG_PREFIX = "<Speech Recognition module (streaming)> "
-
-class StreamingSttProvider {
-    //########//
-    // Config //
-    //########//
-
-    settings
-
-    defaultSettings = {
-        triggerWordsText: "",
-        triggerWords : [],
-        triggerWordsEnabled : false,
-        triggerWordsIncluded: false,
-        debug : false,
-    }
-
-    get settingsHtml() {
-        let html = '\
-        <div id="speech_recognition_streaming_trigger_words_div">\
-        <span>Trigger words</span>\
-        <textarea id="speech_recognition_streaming_trigger_words" class="text_pole textarea_compact" type="text" rows="4" placeholder="Enter comma separated words that triggers new message, example:\nhey, hey aqua, record, listen"></textarea>\
-        <label class="checkbox_label" for="speech_recognition_streaming_trigger_words_enabled">\
-            <input type="checkbox" id="speech_recognition_streaming_trigger_words_enabled" name="speech_recognition_trigger_words_enabled">\
-            <small>Enable trigger words</small>\
-        </label>\
-        <label class="checkbox_label" for="speech_recognition_trigger_words_included">\
-            <input type="checkbox" id="speech_recognition_trigger_words_included" name="speech_recognition_trigger_words_included">\
-            <small>Include trigger words in message</small>\
-        </label>\
-        <label class="checkbox_label" for="speech_recognition_streaming_debug">\
-            <input type="checkbox" id="speech_recognition_streaming_debug" name="speech_recognition_streaming_debug">\
-            <small>Enable debug pop ups</small>\
-        </label>\
-        </div>\
-        '
-        return html
-    }
-
-    onSettingsChange() {
-        this.settings.triggerWordsText = $('#speech_recognition_streaming_trigger_words').val();
-        let array = $('#speech_recognition_streaming_trigger_words').val().split(",");
-        array = array.map(element => {return element.trim().toLowerCase();});
-        array = array.filter((str) => str !== '');
-        this.settings.triggerWords = array;
-        this.settings.triggerWordsEnabled = $("#speech_recognition_streaming_trigger_words_enabled").is(':checked');
-        this.settings.triggerWordsIncluded = $("#speech_recognition_trigger_words_included").is(':checked');
-        this.settings.debug = $("#speech_recognition_streaming_debug").is(':checked');
-        console.debug(DEBUG_PREFIX+" Updated settings: ", this.settings);
-        this.loadSettings(this.settings);
-    }
-
-    loadSettings(settings) {
-        // Populate Provider UI given input settings
-        if (Object.keys(settings).length == 0) {
-            console.debug(DEBUG_PREFIX+"Using default Whisper STT extension settings")
-        }
-
-        // Only accept keys defined in defaultSettings
-        this.settings = this.defaultSettings
-
-        for (const key in settings){
-            if (key in this.settings){
-                this.settings[key] = settings[key]
-            } else {
-                throw `Invalid setting passed to STT extension: ${key}`
-            }
-        }
-
-        $("#speech_recognition_streaming_trigger_words").val(this.settings.triggerWordsText);
-        $("#speech_recognition_streaming_trigger_words_enabled").prop('checked',this.settings.triggerWordsEnabled);
-        $("#speech_recognition_trigger_words_included").prop('checked',this.settings.triggerWordsIncluded);
-        $("#speech_recognition_streaming_debug").prop('checked',this.settings.debug);
-
-        console.debug(DEBUG_PREFIX+"streaming STT settings loaded")
-    }
-
-    async getUserMessage() {
-        // Return if module is not loaded
-        if (!modules.includes('streaming-stt')) {
-            console.debug(DEBUG_PREFIX+"Module streaming-stt must be activated in Sillytavern Extras for streaming user voice.")
-            return "";
-        }
-
-        const url = new URL(getApiUrl());
-        url.pathname = '/api/speech-recognition/streaming/record-and-transcript';
-
-        const apiResult = await doExtrasFetch(url, {
-            method: 'POST',
-            headers: {
-                'Content-Type': 'application/json',
-                'Bypass-Tunnel-Reminder': 'bypass',
-            },
-            body: JSON.stringify({ text: "" }),
-        });
-
-        if (!apiResult.ok) {
-            toastr.error(apiResult.statusText, DEBUG_PREFIX+'STT Generation Failed  (streaming)', { timeOut: 10000, extendedTimeOut: 20000, preventDuplicates: true });
-            throw new Error(`HTTP ${apiResult.status}: ${await apiResult.text()}`);
-        }
-
-        const data = await apiResult.json();
-        return data.transcript;
-    }
-
-}

public/scripts/extensions/speech-recognition/style.css

@@ -1,3 +0,0 @@
-.speech-toggle {
-    display: flex;
-}

public/scripts/extensions/speech-recognition/vosk.js

@@ -1,65 +0,0 @@
-import { getApiUrl, doExtrasFetch } from "../../extensions.js";
-export { VoskSttProvider }
-
-const DEBUG_PREFIX = "<Speech Recognition module (Vosk)> "
-
-class VoskSttProvider {
-    //########//
-    // Config //
-    //########//
-
-    settings
-
-    defaultSettings = {
-    }
-
-    get settingsHtml() {
-        let html = ""
-        return html
-    }
-
-    onSettingsChange() {
-        // Used when provider settings are updated from UI
-    }
-
-    loadSettings(settings) {
-        // Populate Provider UI given input settings
-        if (Object.keys(settings).length == 0) {
-            console.debug(DEBUG_PREFIX+"Using default vosk STT extension settings")
-        }
-
-        // Only accept keys defined in defaultSettings
-        this.settings = this.defaultSettings
-
-        for (const key in settings){
-            if (key in this.settings){
-                this.settings[key] = settings[key]
-            } else {
-                throw `Invalid setting passed to STT extension: ${key}`
-            }
-        }
-
-        console.debug(DEBUG_PREFIX+"Vosk STT settings loaded")
-    }
-
-    async processAudio(audioblob) {
-        var requestData = new FormData();
-        requestData.append('AudioFile', audioblob, 'record.wav');
-        
-        const url = new URL(getApiUrl());
-        url.pathname = '/api/speech-recognition/vosk/process-audio';
-
-        const apiResult = await doExtrasFetch(url, {
-            method: 'POST',
-            body: requestData,
-        });
-
-        if (!apiResult.ok) {
-            toastr.error(apiResult.statusText, 'STT Generation Failed  (Vosk)', { timeOut: 10000, extendedTimeOut: 20000, preventDuplicates: true });
-            throw new Error(`HTTP ${apiResult.status}: ${await apiResult.text()}`);
-        }
-        
-        const result = await apiResult.json();
-        return  result.transcript;
-    }
-}

public/scripts/extensions/speech-recognition/whisper.js

@@ -1,67 +0,0 @@
-import { getApiUrl, doExtrasFetch } from "../../extensions.js";
-export { WhisperSttProvider }
-
-const DEBUG_PREFIX = "<Speech Recognition module (Vosk)> "
-
-class WhisperSttProvider {
-    //########//
-    // Config //
-    //########//
-
-    settings
-
-    defaultSettings = {
-        //model_path: "",
-    }
-
-    get settingsHtml() {
-        let html = ""
-        return html
-    }
-
-    onSettingsChange() {
-        // Used when provider settings are updated from UI
-    }
-
-    loadSettings(settings) {
-        // Populate Provider UI given input settings
-        if (Object.keys(settings).length == 0) {
-            console.debug(DEBUG_PREFIX+"Using default Whisper STT extension settings")
-        }
-
-        // Only accept keys defined in defaultSettings
-        this.settings = this.defaultSettings
-
-        for (const key in settings){
-            if (key in this.settings){
-                this.settings[key] = settings[key]
-            } else {
-                throw `Invalid setting passed to STT extension: ${key}`
-            }
-        }
-
-        console.debug(DEBUG_PREFIX+"Whisper STT settings loaded")
-    }
-
-    async processAudio(audioblob) {
-        var requestData = new FormData();
-        requestData.append('AudioFile', audioblob, 'record.wav');
-        
-        const url = new URL(getApiUrl());
-        url.pathname = '/api/speech-recognition/whisper/process-audio';
-
-        const apiResult = await doExtrasFetch(url, {
-            method: 'POST',
-            body: requestData,
-        });
-
-        if (!apiResult.ok) {
-            toastr.error(apiResult.statusText, 'STT Generation Failed (Whisper)', { timeOut: 10000, extendedTimeOut: 20000, preventDuplicates: true });
-            throw new Error(`HTTP ${apiResult.status}: ${await apiResult.text()}`);
-        }
-        
-        const result = await apiResult.json();
-        return  result.transcript;
-    }
-
-}

public/scripts/group-chats.js

@@ -1555,11 +1555,14 @@ function doCurMemberListPopout() {
         <div id="groupMemberListPopoutClose" class="fa-solid fa-circle-xmark hoverglow"></div>
     </div>`
         const newElement = $(template);
-        newElement.attr('id', 'groupMemberListPopout');
-        newElement.removeClass('zoomed_avatar')
-        newElement.empty()
 
-        newElement.append(controlBarHtml).append(memberListClone)
+        newElement.attr('id', 'groupMemberListPopout')
+            .removeClass('zoomed_avatar')
+            .addClass('draggable')
+            .empty()
+            .append(controlBarHtml)
+            .append(memberListClone)
+
         $('body').append(newElement);
         loadMovingUIState();
         $("#groupMemberListPopout").fadeIn(250)

public/scripts/power-user.js

@@ -1413,12 +1413,14 @@ async function resetMovablePanels(type) {
         'floatingPrompt',
         'expression-holder',
         'groupMemberListPopout',
-        'summaryExtensionPopout'
+        'summaryExtensionPopout',
+        'gallery'
     ];
 
     const panelStyles = ['top', 'left', 'right', 'bottom', 'height', 'width', 'margin',];
 
     panelIds.forEach((id) => {
+        console.log(id)
         const panel = document.getElementById(id);
 
         if (panel) {

public/scripts/world-info.js

@@ -1,6 +1,6 @@
-import { saveSettings, callPopup, substituteParams, getRequestHeaders, chat_metadata, this_chid, characters, saveCharacterDebounced, menu_type, eventSource, event_types } from "../script.js";
+import { saveSettings, callPopup, substituteParams, getRequestHeaders, chat_metadata, this_chid, characters, saveCharacterDebounced, menu_type, eventSource, event_types, getExtensionPrompt, MAX_INJECTION_DEPTH, extension_prompt_types, getExtensionPromptByName } from "../script.js";
 import { download, debounce, initScrollHeight, resetScrollHeight, parseJsonFile, extractDataFromPng, getFileBuffer, getCharaFilename, getSortableDelay, escapeRegex, PAGINATION_TEMPLATE, navigation_option, waitUntilCondition } from "./utils.js";
-import { getContext } from "./extensions.js";
+import { extension_settings, getContext } from "./extensions.js";
 import { NOTE_MODULE_NAME, metadata_keys, shouldWIAddPrompt } from "./authors-note.js";
 import { registerSlashCommand } from "./slash-commands.js";
 import { getDeviceInfo } from "./RossAscends-mods.js";
@@ -581,8 +581,8 @@ function getWorldEntry(name, data, entry) {
         data.entries[uid].selectiveLogic = !isNaN(value) ? value : 0;
         setOriginalDataValue(data, uid, "selectiveLogic", data.entries[uid].selectiveLogic);
         saveWorldInfo(name, data);
-
     });
+
     template
         .find(`select[name="entryLogicType"] option[value=${entry.selectiveLogic}]`)
         .prop("selected", true)
@@ -1341,7 +1341,27 @@ async function getSortedEntries() {
 async function checkWorldInfo(chat, maxContext) {
     const context = getContext();
     const messagesToLookBack = world_info_depth * 2 || 1;
-    let textToScan = transformString(chat.slice(0, messagesToLookBack).join(""));
+
+    // Combine the chat
+    let textToScan = chat.slice(0, messagesToLookBack).join("");
+
+    // Add the depth or AN if enabled
+    // Put this code here since otherwise, the chat reference is modified
+    if (extension_settings.note.allowWIScan) {
+        let depthPrompt = getExtensionPromptByName("DEPTH_PROMPT")
+        if (depthPrompt) {
+            textToScan = `${depthPrompt}\n${textToScan}`
+        }
+
+        let anPrompt = getExtensionPromptByName(NOTE_MODULE_NAME);
+        if (anPrompt) {
+            textToScan = `${anPrompt}\n${textToScan}`
+        }
+    }
+
+    // Transform the resulting string
+    textToScan = transformString(textToScan);
+
     let needsToScan = true;
     let count = 0;
     let allActivatedEntries = new Set();

public/style.css

@@ -346,11 +346,12 @@ code {
 
 
 hr {
-    background-image: linear-gradient(90deg, var(--transparent), var(--SmartThemeBorderColor), var(--transparent));
+    background-image: linear-gradient(90deg, var(--transparent), var(--SmartThemeBodyColor), var(--transparent));
     margin: 5px 0;
     height: 1px;
     min-height: 1px;
     border: 0;
+    opacity: 0.2;
 }
 
 #bg1,
@@ -389,7 +390,7 @@ hr {
     display: inline-block;
     height: var(--bottomFormBlockSize);
     position: absolute;
-    border-bottom: 1px solid var(--SmartThemeBorderColor);
+    /* border-bottom: 1px solid var(--SmartThemeBorderColor); */
     box-shadow: 0 2px 20px 0 var(--black70a);
     backdrop-filter: blur(var(--SmartThemeBlurStrength));
     background-color: var(--SmartThemeBlurTintColor);
@@ -421,23 +422,25 @@ hr {
 
 .drag-grabber {
     position: absolute;
-    width: 20px !important;
+    /*     width: 20px !important;
-    height: 20px !important;
+    height: 20px !important; */
+    margin: 0px 5px;
     color: var(--SmartThemeBodyColor);
     z-index: 2000;
     text-align: center;
-    border-radius: 5px;
+    /* border-radius: 5px; */
     vertical-align: middle;
     right: 0px;
     top: 0px;
     opacity: 0.5;
     cursor: grab;
-    border: 1px solid var(--SmartThemeBorderColor);
+    /* border: 1px solid var(--SmartThemeBorderColor); */
     cursor: -moz-grab;
     cursor: -webkit-grab;
     display: none;
     filter: drop-shadow(0px 0px 0px black);
     transition: all 250ms;
+    font-size: calc(var(--mainFontSize)*1.3);
 }
 
 .drag-grabber:hover {
@@ -480,9 +483,9 @@ hr {
     overflow-y: scroll;
     display: flex;
     bottom: 10px;
-    border-bottom: 1px solid var(--SmartThemeBorderColor);
+    /*     border-bottom: 1px solid var(--SmartThemeBorderColor);
     border-left: 1px solid var(--SmartThemeBorderColor);
-    border-right: 1px solid var(--SmartThemeBorderColor);
+    border-right: 1px solid var(--SmartThemeBorderColor); */
     backdrop-filter: blur(var(--SmartThemeBlurStrength));
     background-color: var(--SmartThemeChatTintColor);
     -webkit-backdrop-filter: blur(var(--SmartThemeBlurStrength));
@@ -959,10 +962,10 @@ select {
 #character_cross,
 #select_chat_cross {
     position: absolute;
-    right: 15px;
+    right: 5px;
-    top: 15px;
+    top: 5px;
-    width: 20px;
+    /* width: 20px;
-    height: 20px;
+    height: 20px; */
     cursor: pointer;
     opacity: 0.6;
     font-size: 24px;
@@ -1251,9 +1254,11 @@ input[type="file"] {
 .dragClose {
     height: 15px;
     aspect-ratio: 1 / 1;
-    font-size: 20px;
+    font-size: calc(var(--mainFontSize)*1.3);
     opacity: 0.5;
     transition: all 250ms;
+    filter: drop-shadow(0px 0px 2px black);
+    text-shadow: none;
 }
 
 .dragClose:hover {
@@ -1604,11 +1609,6 @@ input[type=search]:focus::-webkit-search-cancel-button {
     align-items: center;
 }
 
-#user-settings-block h4,
-#user-settings-block h3 {
-    margin: 5px 0;
-}
-
 #avatar-and-name-block {
     justify-content: space-between;
     display: flex;
@@ -3409,51 +3409,38 @@ a {
     color: var(--SmartThemeBodyColor);
 }
 
-.zoomed_avatar,
-#groupMemberListPopout,
-#summaryExtensionPopout {
-    min-width: 100px;
-    min-height: 100px;
-    max-height: 90vh;
-    max-width: 90vh;
-    width: clamp(100px, 400px, calc((100vw - var(--sheldWidth)) /2));
-    /* width: calc((100vw - var(--sheldWidth)) /2); */
-    position: absolute;
-    padding: 0;
-    /* filter: drop-shadow(2px 2px 2px var(--grey7070a)); */
-    box-shadow: 1px 2px 2px var(--black30a);
-    z-index: 29;
-    overflow: hidden;
-    display: none;
-    bottom: 0;
-    border-radius: 10px;
-}
-
-.zoomed_avatar {
-    aspect-ratio: 2 / 3;
-}
-
-#groupMemberListPopout,
-#summaryExtensionPopout {
-    border: 1px solid var(--SmartThemeBorderColor);
-    padding: 5px;
-}
-
 .draggable {
     min-width: 100px;
     min-height: 100px;
     max-height: 90vh;
-    max-width: 90vh;
+    max-width: 90vw;
-    width: clamp(100px, 400px, calc((100vw - var(--sheldWidth)) /2));
+    width: calc((100vw - var(--sheldWidth)) /2);
     position: absolute;
     padding: 0;
-    filter: drop-shadow(2px 2px 2px var(--grey7070a));
+    filter: drop-shadow(1px 1px 2px var(--black50a));
     z-index: 29;
     overflow: hidden;
     display: none;
     bottom: 0;
-    aspect-ratio: 2 / 3;
     border-radius: 10px;
+    border: 1px solid var(--SmartThemeBorderColor);
+    aspect-ratio: unset;
+    backdrop-filter: blur(var(--SmartThemeBlurStrength));
+    background-color: var(--SmartThemeBlurTintColor);
+    padding: 5px;
+}
+
+.zoomed_avatar {
+    aspect-ratio: 2 / 3;
+    padding: 0;
+    border: 0;
+}
+
+.zoomed_avatar img {
+    height: 100%;
+    width: 100%;
+    vertical-align: bottom;
+    object-fit: cover;
 }
 
 /* Hide scrollbar for Chrome, Safari and Opera */
@@ -3477,13 +3464,13 @@ a {
     transition: all 250ms;
 }
 
-#groupMemberListPopout,
+/* #groupMemberListPopout,
 #summaryExtensionPopout {
     aspect-ratio: unset;
     backdrop-filter: blur(var(--SmartThemeBlurStrength));
     background-color: var(--SmartThemeBlurTintColor);
     overflow: auto;
-}
+} */
 
 #groupMemberListPopout {
     padding: 0;
@@ -3497,13 +3484,7 @@ a {
 
 }
 
-.zoomed_avatar img {
+
-    border-radius: 20px;
-    height: 100%;
-    width: 100%;
-    vertical-align: bottom;
-    object-fit: cover;
-}
 
 .timestamp {
     font-size: calc(var(--mainFontSize) * 0.7);

server.js

@@ -627,6 +627,9 @@ app.post("/generate_textgenerationwebui", jsonParser, async function (request, r
                         yield message.text;
                         break;
                     case 'stream_end':
+                        if (message.error) {
+                            yield `\n[API Error] ${message.error}\n`
+                        }
                         websocket.close();
                         return;
                 }