[wip] Update transformers.js

2025-06-05 21:59:27 +02:00 · 2025-04-15 21:14:09 +03:00
25 changed files with 736 additions and 1202 deletions
--- a/.github/workflows/pr-auto-manager.yml
+++ b/.github/workflows/pr-auto-manager.yml
@@ -30,8 +30,7 @@ jobs:
        # https://github.com/marketplace/actions/checkout
        uses: actions/checkout@v4.2.2
        with:
-            ref: ${{ github.event.pull_request.head.sha }}
-            repository: ${{ github.event.pull_request.head.repo.full_name }}
+            ref: ${{ github.head_ref }}

      - name: Setup Node.js
        # Setup Node.js environment
--- a/default/config.yaml
+++ b/default/config.yaml
@@ -155,7 +155,6 @@ whitelistImportDomains:
  - cdn.discordapp.com
  - files.catbox.moe
  - raw.githubusercontent.com
-  - char-archive.evulid.cc
 # API request overrides (for KoboldAI and Text Completion APIs)
 ## Note: host includes the port number if it's not the default (80 or 443)
 ## Format is an array of objects:
--- a/package-lock.json
+++ b/package-lock.json
--- a/package.json
+++ b/package.json
@@ -3,6 +3,7 @@
        "@adobe/css-tools": "^4.4.2",
        "@agnai/sentencepiece-js": "^1.1.1",
        "@agnai/web-tokenizers": "^0.1.3",
+        "@huggingface/transformers": "^3.4.2",
        "@iconfu/svg-inject": "^1.2.3",
        "@jimp/core": "^1.6.0",
        "@jimp/js-bmp": "^1.6.0",
@@ -74,7 +75,6 @@
        "sanitize-filename": "^1.6.3",
        "seedrandom": "^3.0.5",
        "showdown": "^2.1.0",
-        "sillytavern-transformers": "2.14.6",
        "simple-git": "^3.27.0",
        "slidetoggle": "^4.0.0",
        "tiktoken": "^1.0.20",
@@ -100,6 +100,9 @@
        },
        "node-fetch": {
            "whatwg-url": "^14.0.0"
+        },
+        "@huggingface/transformers": {
+            "onnxruntime-node": "https://github.com/Cohee1207/onnxruntime/releases/download/1.20.1/onnxruntime-node-1.20.1.tgz"
        }
    },
    "name": "sillytavern",
@@ -109,7 +112,7 @@
        "type": "git",
        "url": "https://github.com/SillyTavern/SillyTavern.git"
    },
-    "version": "1.12.14",
+    "version": "1.12.13",
    "scripts": {
        "start": "node server.js",
        "debug": "node --inspect server.js",
--- a/post-install.js
+++ b/post-install.js
@@ -332,7 +332,7 @@ try {
    // 1. Create default config files
    createDefaultFiles();
    // 2. Copy transformers WASM binaries from node_modules
-    copyWasmFiles();
+    // copyWasmFiles();
    // 3. Add missing config values
    addMissingConfigValues();
 } catch (error) {
--- a/public/index.html
+++ b/public/index.html
@@ -1419,7 +1419,7 @@
                                        </div>
                                    </div>

-                                    <div data-tg-type="aphrodite, ooba, koboldcpp, tabby, llamacpp, dreamgen" id="dryBlock" class="wide100p">
+                                    <div data-tg-type="aphrodite, ooba, koboldcpp, tabby, llamacpp" id="dryBlock" class="wide100p">
                                        <h4 class="wide100p textAlignCenter" title="DRY penalizes tokens that would extend the end of the input into a sequence that has previously occurred in the input. Set multiplier to 0 to disable." data-i18n="[title]DRY_Repetition_Penalty_desc">
                                            <label data-i18n="DRY Repetition Penalty">DRY Repetition Penalty</label>
                                            <a href="https://github.com/oobabooga/text-generation-webui/pull/5677" target="_blank">
@@ -1574,7 +1574,7 @@
                                                    <div class="fa-solid fa-circle-info opacity50p " data-i18n="[title]Add the bos_token to the beginning of prompts. Disabling this can make the replies more creative" title="Add the bos_token to the beginning of prompts. Disabling this can make the replies more creative."></div>
                                                </label>
                                            </label>
-                                            <label data-tg-type="ooba, llamacpp, tabby, koboldcpp, dreamgen" class="checkbox_label  flexGrow flexShrink" for="ban_eos_token_textgenerationwebui">
+                                            <label data-tg-type="ooba, llamacpp, tabby, koboldcpp" class="checkbox_label  flexGrow flexShrink" for="ban_eos_token_textgenerationwebui">
                                                <input type="checkbox" id="ban_eos_token_textgenerationwebui" />
                                                <label>
                                                    <small data-i18n="Ban EOS Token">Ban EOS Token</small>
@@ -2762,7 +2762,7 @@
                                <option value="xai">xAI (Grok)</option>
                            </optgroup>
                        </select>
-                        <div class="inline-drawer wide100p" data-source="openai,claude,mistralai,makersuite,deepseek,xai">
+                        <div class="inline-drawer wide100p" data-source="openai,claude,mistralai,makersuite,deepseek">
                            <div class="inline-drawer-toggle inline-drawer-header">
                                <b data-i18n="Reverse Proxy">Reverse Proxy</b>
                                <div class="fa-solid fa-circle-chevron-down inline-drawer-icon down"></div>
@@ -2825,7 +2825,7 @@
                                </div>
                            </div>
                        </div>
-                        <div id="ReverseProxyWarningMessage" data-source="openai,claude,mistralai,makersuite,deepseek,xai">
+                        <div id="ReverseProxyWarningMessage" data-source="openai,claude,mistralai,makersuite,deepseek">
                            <div class="reverse_proxy_warning">
                                <b>
                                    <div data-i18n="Using a proxy that you're not running yourself is a risk to your data privacy.">
@@ -2897,7 +2897,7 @@
                                        <option value="gpt-4.1-nano">gpt-4.1-nano</option>
                                        <option value="gpt-4.1-nano-2025-04-14">gpt-4.1-nano-2025-04-14</option>
                                    </optgroup>
-                                    <optgroup label="o1">
+                                    <optgroup label="o1 and o1-mini">
                                        <option value="o1">o1</option>
                                        <option value="o1-2024-12-17">o1-2024-12-17</option>
                                        <option value="o1-mini">o1-mini</option>
@@ -2906,15 +2906,9 @@
                                        <option value="o1-preview-2024-09-12">o1-preview-2024-09-12</option>
                                    </optgroup>
                                    <optgroup label="o3">
-                                        <option value="o3">o3</option>
-                                        <option value="o3-2025-04-16">o3-2025-04-16</option>
                                        <option value="o3-mini">o3-mini</option>
                                        <option value="o3-mini-2025-01-31">o3-mini-2025-01-31</option>
                                    </optgroup>
-                                    <optgroup label="o4">
-                                        <option value="o4-mini">o4-mini</option>
-                                        <option value="o4-mini-2025-04-16">o4-mini-2025-04-16</option>
-                                    </optgroup>
                                    <optgroup label="GPT-4.5">
                                        <option value="gpt-4.5-preview">gpt-4.5-preview</option>
                                        <option value="gpt-4.5-preview-2025-02-27">gpt-4.5-preview-2025-02-27</option>
@@ -3162,7 +3156,6 @@
                                        <option value="gemini-2.5-pro-exp-03-25">Gemini 2.5 Pro Experimental 2025-03-25</option>
                                        <option value="gemini-2.0-pro-exp">Gemini 2.0 Pro Experimental</option>
                                        <option value="gemini-2.0-pro-exp-02-05">Gemini 2.0 Pro Experimental 2025-02-05</option>
-                                        <option value="gemini-2.5-flash-preview-04-17">Gemini 2.5 Flash Preview 2025-04-17</option>
                                        <option value="gemini-2.0-flash-lite-preview">Gemini 2.0 Flash-Lite Preview</option>
                                        <option value="gemini-2.0-flash-lite-preview-02-05">Gemini 2.0 Flash-Lite Preview 2025-02-05</option>
                                        <option value="gemini-2.0-flash-001">Gemini 2.0 Flash [001]</option>
--- a/public/script.js
+++ b/public/script.js
@@ -11847,8 +11847,8 @@ jQuery(async function () {
            return;
        }
        const drawer = $(this).closest('.inline-drawer');
-        const icon = drawer.find('>.inline-drawer-header .inline-drawer-icon');
-        const drawerContent = drawer.find('>.inline-drawer-content');
+        const icon = drawer.find('.inline-drawer-icon');
+        const drawerContent = drawer.find('.inline-drawer-content');
        icon.toggleClass('down up');
        icon.toggleClass('fa-circle-chevron-down fa-circle-chevron-up');
        drawerContent.stop().slideToggle({
--- a/public/scripts/extensions/assets/window.html
+++ b/public/scripts/extensions/assets/window.html
@@ -39,7 +39,7 @@ To install a single 3rd party extension, use the &quot;Install Extensions&quot;
                    <span data-i18n="Characters">Characters</span>
                </div>
            </div>
-            <div id="assets_menu">
+            <div class="inline-drawer-content" id="assets_menu">
            </div>
        </div>
    </div>
--- a/public/scripts/extensions/caption/settings.html
+++ b/public/scripts/extensions/caption/settings.html
@@ -60,10 +60,6 @@
                        <option data-type="openai" value="chatgpt-4o-latest">chatgpt-4o-latest</option>
                        <option data-type="openai" value="o1">o1</option>
                        <option data-type="openai" value="o1-2024-12-17">o1-2024-12-17</option>
-                        <option data-type="openai" value="o3">o3</option>
-                        <option data-type="openai" value="o3-2025-04-16">o3-2025-04-16</option>
-                        <option data-type="openai" value="o4-mini">o4-mini</option>
-                        <option data-type="openai" value="o4-mini-2025-04-16">o4-mini-2025-04-16</option>
                        <option data-type="openai" value="gpt-4.5-preview">gpt-4.5-preview</option>
                        <option data-type="openai" value="gpt-4.5-preview-2025-02-27">gpt-4.5-preview-2025-02-27</option>
                        <option data-type="anthropic" value="claude-3-7-sonnet-latest">claude-3-7-sonnet-latest</option>
@@ -80,7 +76,6 @@
                        <option data-type="google" value="gemini-2.5-pro-exp-03-25">gemini-2.5-pro-exp-03-25</option>
                        <option data-type="google" value="gemini-2.0-pro-exp">gemini-2.0-pro-exp</option>
                        <option data-type="google" value="gemini-2.0-pro-exp-02-05">gemini-2.0-pro-exp-02-05</option>
-                        <option data-type="google" value="gemini-2.5-flash-preview-04-17">gemini-2.5-flash-preview-04-17</option>
                        <option data-type="google" value="gemini-2.0-flash-lite-preview">gemini-2.0-flash-lite-preview</option>
                        <option data-type="google" value="gemini-2.0-flash-lite-preview-02-05">gemini-2.0-flash-lite-preview-02-05</option>
                        <option data-type="google" value="gemini-2.0-flash">gemini-2.0-flash</option>
--- a/public/scripts/extensions/tts/settings.html
+++ b/public/scripts/extensions/tts/settings.html
@@ -76,7 +76,7 @@
            <div id="tts_voicemap_block">
            </div>
            <hr>
-            <form id="tts_provider_settings">
+            <form id="tts_provider_settings" class="inline-drawer-content">
            </form>
            <div class="tts_buttons">
                <input id="tts_voices" class="menu_button" type="submit" value="Available voices" />
--- a/public/scripts/extensions/tts/system.js
+++ b/public/scripts/extensions/tts/system.js
@@ -79,10 +79,6 @@ class SystemTtsProvider {
    // Config //
    //########//

-    // Static constants for the simulated default voice
-    static BROWSER_DEFAULT_VOICE_ID = '__browser_default__';
-    static BROWSER_DEFAULT_VOICE_NAME = 'System Default Voice';
-
    settings;
    ready = false;
    voices = [];
@@ -172,97 +168,51 @@ class SystemTtsProvider {
    //#################//
    fetchTtsVoiceObjects() {
        if (!('speechSynthesis' in window)) {
-            return Promise.resolve([]);
+            return [];
        }

        return new Promise((resolve) => {
            setTimeout(() => {
-                let voices = speechSynthesis.getVoices();
+                const voices = speechSynthesis
+                    .getVoices()
+                    .sort((a, b) => a.lang.localeCompare(b.lang) || a.name.localeCompare(b.name))
+                    .map(x => ({ name: x.name, voice_id: x.voiceURI, preview_url: false, lang: x.lang }));

-                if (voices.length === 0) {
-                    // Edge compat: Provide default when voices empty
-                    console.warn('SystemTTS: getVoices() returned empty list. Providing browser default option.');
-                    const defaultVoice = {
-                        name: SystemTtsProvider.BROWSER_DEFAULT_VOICE_NAME,
-                        voice_id: SystemTtsProvider.BROWSER_DEFAULT_VOICE_ID,
-                        preview_url: false,
-                        lang: navigator.language || 'en-US',
-                    };
-                    resolve([defaultVoice]);
-                } else {
-                    const mappedVoices = voices
-                        .sort((a, b) => a.lang.localeCompare(b.lang) || a.name.localeCompare(b.name))
-                        .map(x => ({ name: x.name, voice_id: x.voiceURI, preview_url: false, lang: x.lang }));
-                    resolve(mappedVoices);
-                }
-            }, 50);
+                resolve(voices);
+            }, 1);
        });
    }

    previewTtsVoice(voiceId) {
        if (!('speechSynthesis' in window)) {
-            throw new Error('Speech synthesis API is not supported');
+            throw 'Speech synthesis API is not supported';
        }

-        let voice = null;
-        if (voiceId !== SystemTtsProvider.BROWSER_DEFAULT_VOICE_ID) {
-            const voices = speechSynthesis.getVoices();
-            voice = voices.find(x => x.voiceURI === voiceId);
+        const voice = speechSynthesis.getVoices().find(x => x.voiceURI === voiceId);

-            if (!voice && voices.length > 0) {
-                console.warn(`SystemTTS Preview: Voice ID "${voiceId}" not found among available voices. Using browser default.`);
-            } else if (!voice && voices.length === 0) {
-                console.warn('SystemTTS Preview: Voice list is empty. Using browser default.');
-            }
-        } else {
-            console.log('SystemTTS Preview: Using browser default voice as requested.');
+        if (!voice) {
+            throw `TTS Voice id ${voiceId} not found`;
        }

        speechSynthesis.cancel();
-        const langForPreview = voice ? voice.lang : (navigator.language || 'en-US');
-        const text = getPreviewString(langForPreview);
+        const text = getPreviewString(voice.lang);
        const utterance = new SpeechSynthesisUtterance(text);
-
-        if (voice) {
-            utterance.voice = voice;
-        }
-
+        utterance.voice = voice;
        utterance.rate = this.settings.rate || 1;
        utterance.pitch = this.settings.pitch || 1;
-
-        utterance.onerror = (event) => {
-            console.error(`SystemTTS Preview Error: ${event.error}`, event);
-        };
-
        speechSynthesis.speak(utterance);
    }

    async getVoice(voiceName) {
        if (!('speechSynthesis' in window)) {
-            return { voice_id: null, name: 'API Not Supported' };
-        }
-
-        if (voiceName === SystemTtsProvider.BROWSER_DEFAULT_VOICE_NAME) {
-            return {
-                voice_id: SystemTtsProvider.BROWSER_DEFAULT_VOICE_ID,
-                name: SystemTtsProvider.BROWSER_DEFAULT_VOICE_NAME,
-            };
+            return { voice_id: null };
        }

        const voices = speechSynthesis.getVoices();
-
-        if (voices.length === 0) {
-            console.warn('SystemTTS: Empty voice list, using default fallback');
-            return {
-                voice_id: SystemTtsProvider.BROWSER_DEFAULT_VOICE_ID,
-                name: SystemTtsProvider.BROWSER_DEFAULT_VOICE_NAME,
-            };
-        }
-
        const match = voices.find(x => x.name == voiceName);

        if (!match) {
-            throw new Error(`SystemTTS getVoice: TTS Voice name "${voiceName}" not found`);
+            throw `TTS Voice name ${voiceName} not found`;
        }

        return { voice_id: match.voiceURI, name: match.name };
@@ -287,6 +237,7 @@ class SystemTtsProvider {
            speechUtteranceChunker(utterance, {
                chunkLength: 200,
            }, function () {
+                //some code to execute when done
                resolve(silence);
                console.log('System TTS done');
            });
--- a/public/scripts/extensions/vectors/index.js
+++ b/public/scripts/extensions/vectors/index.js
@@ -55,8 +55,6 @@ const getBatchSize = () => ['transformers', 'palm', 'ollama'].includes(settings.
 const settings = {
    // For both
    source: 'transformers',
-    alt_endpoint_url: '',
-    use_alt_endpoint: false,
    include_wi: false,
    togetherai_model: 'togethercomputer/m2-bert-80M-32k-retrieval',
    openai_model: 'text-embedding-ada-002',
@@ -111,7 +109,6 @@ const settings = {
 const moduleWorker = new ModuleWorkerWrapper(synchronizeChat);
 const webllmProvider = new WebLlmVectorProvider();
 const cachedSummaries = new Map();
-const vectorApiRequiresUrl = ['llamacpp', 'vllm', 'ollama', 'koboldcpp'];

 /**
 * Gets the Collection ID for a file embedded in the chat.
@@ -780,14 +777,14 @@ function getVectorsRequestBody(args = {}) {
            break;
        case 'ollama':
            body.model = extension_settings.vectors.ollama_model;
-            body.apiUrl = settings.use_alt_endpoint ? settings.alt_endpoint_url : textgenerationwebui_settings.server_urls[textgen_types.OLLAMA];
+            body.apiUrl = textgenerationwebui_settings.server_urls[textgen_types.OLLAMA];
            body.keep = !!extension_settings.vectors.ollama_keep;
            break;
        case 'llamacpp':
-            body.apiUrl = settings.use_alt_endpoint ? settings.alt_endpoint_url : textgenerationwebui_settings.server_urls[textgen_types.LLAMACPP];
+            body.apiUrl = textgenerationwebui_settings.server_urls[textgen_types.LLAMACPP];
            break;
        case 'vllm':
-            body.apiUrl = settings.use_alt_endpoint ? settings.alt_endpoint_url : textgenerationwebui_settings.server_urls[textgen_types.VLLM];
+            body.apiUrl = textgenerationwebui_settings.server_urls[textgen_types.VLLM];
            body.model = extension_settings.vectors.vllm_model;
            break;
        case 'webllm':
@@ -886,18 +883,11 @@ function throwIfSourceInvalid() {
        throw new Error('Vectors: API key missing', { cause: 'api_key_missing' });
    }

-    if (vectorApiRequiresUrl.includes(settings.source) && settings.use_alt_endpoint) {
-        if (!settings.alt_endpoint_url) {
-            throw new Error('Vectors: API URL missing', { cause: 'api_url_missing' });
-        }
-    }
-    else {
-        if (settings.source === 'ollama' && !textgenerationwebui_settings.server_urls[textgen_types.OLLAMA] ||
-            settings.source === 'vllm' && !textgenerationwebui_settings.server_urls[textgen_types.VLLM] ||
-            settings.source === 'koboldcpp' && !textgenerationwebui_settings.server_urls[textgen_types.KOBOLDCPP] ||
-            settings.source === 'llamacpp' && !textgenerationwebui_settings.server_urls[textgen_types.LLAMACPP]) {
-            throw new Error('Vectors: API URL missing', { cause: 'api_url_missing' });
-        }
+    if (settings.source === 'ollama' && !textgenerationwebui_settings.server_urls[textgen_types.OLLAMA] ||
+        settings.source === 'vllm' && !textgenerationwebui_settings.server_urls[textgen_types.VLLM] ||
+        settings.source === 'koboldcpp' && !textgenerationwebui_settings.server_urls[textgen_types.KOBOLDCPP] ||
+        settings.source === 'llamacpp' && !textgenerationwebui_settings.server_urls[textgen_types.LLAMACPP]) {
+        throw new Error('Vectors: API URL missing', { cause: 'api_url_missing' });
    }

    if (settings.source === 'ollama' && !settings.ollama_model || settings.source === 'vllm' && !settings.vllm_model) {
@@ -1097,7 +1087,6 @@ function toggleSettings() {
    $('#webllm_vectorsModel').toggle(settings.source === 'webllm');
    $('#koboldcpp_vectorsModel').toggle(settings.source === 'koboldcpp');
    $('#google_vectorsModel').toggle(settings.source === 'palm');
-    $('#vector_altEndpointUrl').toggle(vectorApiRequiresUrl.includes(settings.source));
    if (settings.source === 'webllm') {
        loadWebLlmModels();
    }
@@ -1176,7 +1165,7 @@ async function createKoboldCppEmbeddings(items) {
        headers: getRequestHeaders(),
        body: JSON.stringify({
            items: items,
-            server: settings.use_alt_endpoint ? settings.alt_endpoint_url : textgenerationwebui_settings.server_urls[textgen_types.KOBOLDCPP],
+            server: textgenerationwebui_settings.server_urls[textgen_types.KOBOLDCPP],
        }),
    });

@@ -1478,16 +1467,6 @@ jQuery(async () => {
        saveSettingsDebounced();
        toggleSettings();
    });
-    $('#vector_altEndpointUrl_enabled').prop('checked', settings.use_alt_endpoint).on('input', () => {
-        settings.use_alt_endpoint = $('#vector_altEndpointUrl_enabled').prop('checked');
-        Object.assign(extension_settings.vectors, settings);
-        saveSettingsDebounced();
-    });
-    $('#vector_altEndpoint_address').val(settings.alt_endpoint_url).on('change', () => {
-        settings.alt_endpoint_url = String($('#vector_altEndpoint_address').val());
-        Object.assign(extension_settings.vectors, settings);
-        saveSettingsDebounced();
-    });
    $('#api_key_nomicai').on('click', async () => {
        const popupText = 'NomicAI API Key:';
        const key = await callGenericPopup(popupText, POPUP_TYPE.INPUT, '', {
--- a/public/scripts/extensions/vectors/settings.html
+++ b/public/scripts/extensions/vectors/settings.html
@@ -25,16 +25,6 @@
                    <option value="webllm" data-i18n="WebLLM Extension">WebLLM Extension</option>
                </select>
            </div>
-            <div class="flex-container flexFlowColumn" id="vector_altEndpointUrl">
-                <label class="checkbox_label" for="vector_altEndpointUrl_enabled" title="Enable secondary endpoint URL usage, instead of the main one.">
-                    <input id="vector_altEndpointUrl_enabled" type="checkbox" class="checkbox">
-                    <span data-i18n="Use secondary URL">Use secondary URL</span>
-                </label>
-                <label for="vector_altEndpoint_address" data-i18n="Secondary Embedding endpoint URL">
-                    Secondary Embedding endpoint URL
-                </label>
-                <input id="vector_altEndpoint_address" class="text_pole" type="text" placeholder="e.g. http://localhost:5001" />
-            </div>
            <div class="flex-container flexFlowColumn" id="webllm_vectorsModel">
                <label for="vectors_webllm_model" data-i18n="Vectorization Model">
                    Vectorization Model
--- a/public/scripts/openai.js
+++ b/public/scripts/openai.js
@@ -1412,9 +1412,9 @@ export async function prepareOpenAIMessages({
        await populateChatCompletion(prompts, chatCompletion, { bias, quietPrompt, quietImage, type, cyclePrompt, messages, messageExamples });
    } catch (error) {
        if (error instanceof TokenBudgetExceededError) {
-            toastr.error(t`Mandatory prompts exceed the context size.`);
-            chatCompletion.log('Mandatory prompts exceed the context size.');
-            promptManager.error = t`Not enough free tokens for mandatory prompts. Raise your token limit or disable custom prompts.`;
+            toastr.error(t`An error occurred while counting tokens: Token budget exceeded.`);
+            chatCompletion.log('Token budget exceeded.');
+            promptManager.error = t`Not enough free tokens for mandatory prompts. Raise your token Limit or disable custom prompts.`;
        } else if (error instanceof InvalidCharacterNameError) {
            toastr.warning(t`An error occurred while counting tokens: Invalid character name`);
            chatCompletion.log('Invalid character name');
@@ -2039,7 +2039,7 @@ async function sendOpenAIRequest(type, messages, signal) {
    }

    // Proxy is only supported for Claude, OpenAI, Mistral, and Google MakerSuite
-    if (oai_settings.reverse_proxy && [chat_completion_sources.CLAUDE, chat_completion_sources.OPENAI, chat_completion_sources.MISTRALAI, chat_completion_sources.MAKERSUITE, chat_completion_sources.DEEPSEEK, chat_completion_sources.XAI].includes(oai_settings.chat_completion_source)) {
+    if (oai_settings.reverse_proxy && [chat_completion_sources.CLAUDE, chat_completion_sources.OPENAI, chat_completion_sources.MISTRALAI, chat_completion_sources.MAKERSUITE, chat_completion_sources.DEEPSEEK].includes(oai_settings.chat_completion_source)) {
        await validateReverseProxy();
        generate_data['reverse_proxy'] = oai_settings.reverse_proxy;
        generate_data['proxy_password'] = oai_settings.proxy_password;
@@ -2180,27 +2180,25 @@ async function sendOpenAIRequest(type, messages, signal) {
        generate_data['seed'] = oai_settings.seed;
    }

-    if (isOAI && /^(o1|o3|o4)/.test(oai_settings.openai_model)) {
+    if (isOAI && (oai_settings.openai_model.startsWith('o1') || oai_settings.openai_model.startsWith('o3'))) {
+        generate_data.messages.forEach((msg) => {
+            if (msg.role === 'system') {
+                msg.role = 'user';
+            }
+        });
        generate_data.max_completion_tokens = generate_data.max_tokens;
        delete generate_data.max_tokens;
        delete generate_data.logprobs;
        delete generate_data.top_logprobs;
-        delete generate_data.stop;
-        delete generate_data.logit_bias;
+        delete generate_data.n;
        delete generate_data.temperature;
        delete generate_data.top_p;
        delete generate_data.frequency_penalty;
        delete generate_data.presence_penalty;
-        if (oai_settings.openai_model.startsWith('o1')) {
-            generate_data.messages.forEach((msg) => {
-                if (msg.role === 'system') {
-                    msg.role = 'user';
-                }
-            });
-            delete generate_data.n;
-            delete generate_data.tools;
-            delete generate_data.tool_choice;
-        }
+        delete generate_data.tools;
+        delete generate_data.tool_choice;
+        delete generate_data.stop;
+        delete generate_data.logit_bias;
    }

    await eventSource.emit(event_types.CHAT_COMPLETION_SETTINGS_READY, generate_data);
@@ -2236,8 +2234,7 @@ async function sendOpenAIRequest(type, messages, signal) {

                if (Array.isArray(parsed?.choices) && parsed?.choices?.[0]?.index > 0) {
                    const swipeIndex = parsed.choices[0].index - 1;
-                    // FIXME: state.reasoning should be an array to support multi-swipe
-                    swipes[swipeIndex] = (swipes[swipeIndex] || '') + getStreamingReply(parsed, state, { overrideShowThoughts: false });
+                    swipes[swipeIndex] = (swipes[swipeIndex] || '') + getStreamingReply(parsed, state);
                } else {
                    text += getStreamingReply(parsed, state);
                }
@@ -3548,7 +3545,7 @@ async function getStatusOpen() {
        chat_completion_source: oai_settings.chat_completion_source,
    };

-    if (oai_settings.reverse_proxy && [chat_completion_sources.CLAUDE, chat_completion_sources.OPENAI, chat_completion_sources.MISTRALAI, chat_completion_sources.MAKERSUITE, chat_completion_sources.DEEPSEEK, chat_completion_sources.XAI].includes(oai_settings.chat_completion_source)) {
+    if (oai_settings.reverse_proxy && [chat_completion_sources.CLAUDE, chat_completion_sources.OPENAI, chat_completion_sources.MISTRALAI, chat_completion_sources.MAKERSUITE, chat_completion_sources.DEEPSEEK].includes(oai_settings.chat_completion_source)) {
        await validateReverseProxy();
    }

@@ -4131,12 +4128,9 @@ function getMaxContextOpenAI(value) {
    else if (value.includes('gpt-4.1')) {
        return max_1mil;
    }
-    else if (value.startsWith('o1')) {
+    else if (value.startsWith('o1') || value.startsWith('o3')) {
        return max_128k;
    }
-    else if (value.startsWith('o4') || value.startsWith('o3')) {
-        return max_200k;
-    }
    else if (value.includes('chatgpt-4o-latest') || value.includes('gpt-4-turbo') || value.includes('gpt-4o') || value.includes('gpt-4-1106') || value.includes('gpt-4-0125') || value.includes('gpt-4-vision')) {
        return max_128k;
    }
@@ -4451,7 +4445,7 @@ async function onModelChange() {
            $('#openai_max_context').attr('max', max_32k);
        } else if (value.includes('gemini-1.5-pro') || value.includes('gemini-exp-1206') || value.includes('gemini-2.0-pro')) {
            $('#openai_max_context').attr('max', max_2mil);
-        } else if (value.includes('gemini-1.5-flash') || value.includes('gemini-2.0-flash') || value.includes('gemini-2.5-flash-preview-04-17') || value.includes('gemini-2.5-pro-exp-03-25') || value.includes('gemini-2.5-pro-preview-03-25')) {
+        } else if (value.includes('gemini-1.5-flash') || value.includes('gemini-2.0-flash') || value.includes('gemini-2.5-pro-exp-03-25') || value.includes('gemini-2.5-pro-preview-03-25')) {
            $('#openai_max_context').attr('max', max_1mil);
        } else if (value.includes('gemini-1.0-pro') || value === 'gemini-pro') {
            $('#openai_max_context').attr('max', max_32k);
@@ -4949,7 +4943,7 @@ async function onConnectButtonClick(e) {
            await writeSecret(SECRET_KEYS.XAI, api_key_xai);
        }

-        if (!secret_state[SECRET_KEYS.XAI] && !oai_settings.reverse_proxy) {
+        if (!secret_state[SECRET_KEYS.XAI]) {
            console.log('No secret key saved for XAI');
            return;
        }
@@ -5103,7 +5097,6 @@ export function isImageInliningSupported() {
        'gemini-2.5-pro-preview-03-25',
        'gemini-2.0-pro-exp',
        'gemini-2.0-pro-exp-02-05',
-        'gemini-2.5-flash-preview-04-17',
        'gemini-2.0-flash-lite-preview',
        'gemini-2.0-flash-lite-preview-02-05',
        'gemini-2.0-flash',
@@ -5156,15 +5149,11 @@ export function isImageInliningSupported() {
        'grok-2-vision',
        'grok-vision',
        'gpt-4.1',
-        'o3',
-        'o3-2025-04-16',
-        'o4-mini',
-        'o4-mini-2025-04-16',
    ];

    switch (oai_settings.chat_completion_source) {
        case chat_completion_sources.OPENAI:
-            return visionSupportedModels.some(model => oai_settings.openai_model.includes(model) && !oai_settings.openai_model.includes('gpt-4-turbo-preview') && !oai_settings.openai_model.includes('o3-mini'));
+            return visionSupportedModels.some(model => oai_settings.openai_model.includes(model) && !oai_settings.openai_model.includes('gpt-4-turbo-preview'));
        case chat_completion_sources.MAKERSUITE:
            return visionSupportedModels.some(model => oai_settings.google_model.includes(model));
        case chat_completion_sources.CLAUDE:
--- a/public/scripts/textgen-models.js
+++ b/public/scripts/textgen-models.js
@@ -928,10 +928,6 @@ export function getCurrentDreamGenModelTokenizer() {
        return tokenizers.YI;
    } else if (model.id.startsWith('opus-v1-xl')) {
        return tokenizers.LLAMA;
-    } else if (model.id.startsWith('lucid-v1-medium')) {
-        return tokenizers.NEMO;
-    } else if (model.id.startsWith('lucid-v1-extra-large')) {
-        return tokenizers.LLAMA3;
    } else {
        return tokenizers.MISTRAL;
    }
--- a/public/scripts/world-info.js
+++ b/public/scripts/world-info.js
@@ -1338,18 +1338,6 @@ function registerWorldInfoSlashCommands() {
        }
    }

-    async function getGlobalBooksCallback() {
-        if (!selected_world_info?.length) {
-            return JSON.stringify([]);
-        }
-
-        let entries = selected_world_info.slice();
-
-        console.debug(`[WI] Selected global world info has ${entries.length} entries`, selected_world_info);
-
-        return JSON.stringify(entries);
-    }
-
    SlashCommandParser.addCommandObject(SlashCommand.fromProps({
        name: 'world',
        callback: onWorldInfoChange,
@@ -1391,13 +1379,6 @@ function registerWorldInfoSlashCommands() {
        ],
        aliases: ['getchatlore', 'getchatwi'],
    }));
-    SlashCommandParser.addCommandObject(SlashCommand.fromProps({
-        name: 'getglobalbooks',
-        callback: getGlobalBooksCallback,
-        returns: 'list of selected lorebook names',
-        helpString: 'Get a list of names of the selected global lorebooks and pass it down the pipe.',
-        aliases: ['getgloballore', 'getglobalwi'],
-    }));
    SlashCommandParser.addCommandObject(SlashCommand.fromProps({
        name: 'getpersonabook',
        callback: getPersonaBookCallback,
--- a/src/constants.js
+++ b/src/constants.js
@@ -268,6 +268,23 @@ export const FEATHERLESS_KEYS = [
    'guided_whitespace_pattern',
 ];

+// https://dreamgen.com/docs/api#openai-text
+export const DREAMGEN_KEYS = [
+    'model',
+    'prompt',
+    'max_tokens',
+    'temperature',
+    'top_p',
+    'top_k',
+    'min_p',
+    'repetition_penalty',
+    'frequency_penalty',
+    'presence_penalty',
+    'stop',
+    'stream',
+    'minimum_message_content_tokens',
+];
+
 // https://docs.together.ai/reference/completions
 export const TOGETHERAI_KEYS = [
    'model',
--- a/src/endpoints/backends/chat-completions.js
+++ b/src/endpoints/backends/chat-completions.js
@@ -250,7 +250,7 @@ async function sendClaudeRequest(request, response) {
            if (!generateResponse.ok) {
                const generateResponseText = await generateResponse.text();
                console.warn(color.red(`Claude API returned error: ${generateResponse.status} ${generateResponse.statusText}\n${generateResponseText}\n${divider}`));
-                return response.status(500).send({ error: true });
+                return response.status(generateResponse.status).send({ error: true });
            }

            /** @type {any} */
@@ -366,7 +366,6 @@ async function sendMakerSuiteRequest(request, response) {

        const useSystemPrompt = !useMultiModal && (
            model.includes('gemini-2.5-pro') ||
-            model.includes('gemini-2.5-flash') ||
            model.includes('gemini-2.0-pro') ||
            model.includes('gemini-2.0-flash') ||
            model.includes('gemini-2.0-flash-thinking-exp') ||
@@ -464,7 +463,7 @@ async function sendMakerSuiteRequest(request, response) {
        } else {
            if (!generateResponse.ok) {
                console.warn(`Google AI Studio API returned error: ${generateResponse.status} ${generateResponse.statusText} ${await generateResponse.text()}`);
-                return response.status(500).send({ error: true });
+                return response.status(generateResponse.status).send({ error: true });
            }

            /** @type {any} */
@@ -831,100 +830,6 @@ async function sendDeepSeekRequest(request, response) {
    }
 }

-/**
- * Sends a request to XAI API.
- * @param {express.Request} request Express request
- * @param {express.Response} response Express response
- */
-async function sendXaiRequest(request, response) {
-    const apiUrl = new URL(request.body.reverse_proxy || API_XAI).toString();
-    const apiKey = request.body.reverse_proxy ? request.body.proxy_password : readSecret(request.user.directories, SECRET_KEYS.XAI);
-
-    if (!apiKey && !request.body.reverse_proxy) {
-        console.warn('xAI API key is missing.');
-        return response.status(400).send({ error: true });
-    }
-
-    const controller = new AbortController();
-    request.socket.removeAllListeners('close');
-    request.socket.on('close', function () {
-        controller.abort();
-    });
-
-    try {
-        let bodyParams = {};
-
-        if (request.body.logprobs > 0) {
-            bodyParams['top_logprobs'] = request.body.logprobs;
-            bodyParams['logprobs'] = true;
-        }
-
-        if (Array.isArray(request.body.tools) && request.body.tools.length > 0) {
-            bodyParams['tools'] = request.body.tools;
-            bodyParams['tool_choice'] = request.body.tool_choice;
-        }
-
-        if (Array.isArray(request.body.stop) && request.body.stop.length > 0) {
-            bodyParams['stop'] = request.body.stop;
-        }
-
-        if (['grok-3-mini-beta', 'grok-3-mini-fast-beta'].includes(request.body.model)) {
-            bodyParams['reasoning_effort'] = request.body.reasoning_effort === 'high' ? 'high' : 'low';
-        }
-
-        const processedMessages = request.body.messages = convertXAIMessages(request.body.messages, getPromptNames(request));
-
-        const requestBody = {
-            'messages': processedMessages,
-            'model': request.body.model,
-            'temperature': request.body.temperature,
-            'max_tokens': request.body.max_tokens,
-            'max_completion_tokens': request.body.max_completion_tokens,
-            'stream': request.body.stream,
-            'presence_penalty': request.body.presence_penalty,
-            'frequency_penalty': request.body.frequency_penalty,
-            'top_p': request.body.top_p,
-            'seed': request.body.seed,
-            'n': request.body.n,
-            ...bodyParams,
-        };
-
-        const config = {
-            method: 'POST',
-            headers: {
-                'Content-Type': 'application/json',
-                'Authorization': 'Bearer ' + apiKey,
-            },
-            body: JSON.stringify(requestBody),
-            signal: controller.signal,
-        };
-
-        console.debug('xAI request:', requestBody);
-
-        const generateResponse = await fetch(apiUrl + '/chat/completions', config);
-
-        if (request.body.stream) {
-            forwardFetchResponse(generateResponse, response);
-        } else {
-            if (!generateResponse.ok) {
-                const errorText = await generateResponse.text();
-                console.warn(`xAI API returned error: ${generateResponse.status} ${generateResponse.statusText} ${errorText}`);
-                const errorJson = tryParse(errorText) ?? { error: true };
-                return response.status(500).send(errorJson);
-            }
-            const generateResponseJson = await generateResponse.json();
-            console.debug('xAI response:', generateResponseJson);
-            return response.send(generateResponseJson);
-        }
-    } catch (error) {
-        console.error('Error communicating with xAI API: ', error);
-        if (!response.headersSent) {
-            response.send({ error: true });
-        } else {
-            response.end();
-        }
-    }
-}

 export const router = express.Router();

@@ -970,9 +875,8 @@ router.post('/status', async function (request, response_getstatus_openai) {
        api_key_openai = request.body.reverse_proxy ? request.body.proxy_password : readSecret(request.user.directories, SECRET_KEYS.DEEPSEEK);
        headers = {};
    } else if (request.body.chat_completion_source === CHAT_COMPLETION_SOURCES.XAI) {
-        api_url = new URL(request.body.reverse_proxy || API_XAI);
-        api_key_openai = request.body.reverse_proxy ? request.body.proxy_password : readSecret(request.user.directories, SECRET_KEYS.XAI);
-        headers = {};
+        api_url = API_XAI;
+        api_key_openai = readSecret(request.user.directories, SECRET_KEYS.XAI);
    } else {
        console.warn('This chat completion source is not supported yet.');
        return response_getstatus_openai.status(400).send({ error: true });
@@ -1140,7 +1044,6 @@ router.post('/generate', function (request, response) {
        case CHAT_COMPLETION_SOURCES.MISTRALAI: return sendMistralAIRequest(request, response);
        case CHAT_COMPLETION_SOURCES.COHERE: return sendCohereRequest(request, response);
        case CHAT_COMPLETION_SOURCES.DEEPSEEK: return sendDeepSeekRequest(request, response);
-        case CHAT_COMPLETION_SOURCES.XAI: return sendXaiRequest(request, response);
    }

    let apiUrl;
@@ -1252,6 +1155,12 @@ router.post('/generate', function (request, response) {
        apiKey = readSecret(request.user.directories, SECRET_KEYS.ZEROONEAI);
        headers = {};
        bodyParams = {};
+    } else if (request.body.chat_completion_source === CHAT_COMPLETION_SOURCES.XAI) {
+        apiUrl = API_XAI;
+        apiKey = readSecret(request.user.directories, SECRET_KEYS.XAI);
+        headers = {};
+        bodyParams = {};
+        request.body.messages = convertXAIMessages(request.body.messages, getPromptNames(request));
    } else {
        console.warn('This chat completion source is not supported yet.');
        return response.status(400).send({ error: true });
@@ -1259,11 +1168,17 @@ router.post('/generate', function (request, response) {

    // A few of OpenAIs reasoning models support reasoning effort
    if ([CHAT_COMPLETION_SOURCES.CUSTOM, CHAT_COMPLETION_SOURCES.OPENAI].includes(request.body.chat_completion_source)) {
-        if (['o1', 'o3-mini', 'o3-mini-2025-01-31', 'o4-mini', 'o4-mini-2025-04-16', 'o3', 'o3-2025-04-16'].includes(request.body.model)) {
+        if (['o1', 'o3-mini', 'o3-mini-2025-01-31'].includes(request.body.model)) {
            bodyParams['reasoning_effort'] = request.body.reasoning_effort;
        }
    }

+    if ([CHAT_COMPLETION_SOURCES.XAI].includes(request.body.chat_completion_source)) {
+        if (['grok-3-mini-beta', 'grok-3-mini-fast-beta'].includes(request.body.model)) {
+            bodyParams['reasoning_effort'] = request.body.reasoning_effort === 'high' ? 'high' : 'low';
+        }
+    }
+
    if (!apiKey && !request.body.reverse_proxy && request.body.chat_completion_source !== CHAT_COMPLETION_SOURCES.CUSTOM) {
        console.warn('OpenAI API key is missing.');
        return response.status(400).send({ error: true });
--- a/src/endpoints/backends/text-completions.js
+++ b/src/endpoints/backends/text-completions.js
@@ -10,6 +10,7 @@ import {
    INFERMATICAI_KEYS,
    OPENROUTER_KEYS,
    VLLM_KEYS,
+    DREAMGEN_KEYS,
    FEATHERLESS_KEYS,
    OPENAI_KEYS,
 } from '../../constants.js';
@@ -339,6 +340,9 @@ router.post('/generate', async function (request, response) {
        }

        if (request.body.api_type === TEXTGEN_TYPES.DREAMGEN) {
+            request.body = _.pickBy(request.body, (_, key) => DREAMGEN_KEYS.includes(key));
+            // NOTE: DreamGen sometimes get confused by the unusual formatting in the character cards.
+            request.body.stop?.push('### User', '## User');
            args.body = JSON.stringify(request.body);
        }

@@ -446,7 +450,7 @@ ollama.post('/download', async function (request, response) {

        if (!fetchResponse.ok) {
            console.error('Download error:', fetchResponse.status, fetchResponse.statusText);
-            return response.status(500).send({ error: true });
+            return response.status(fetchResponse.status).send({ error: true });
        }

        console.debug('Ollama pull response:', await fetchResponse.json());
@@ -655,14 +659,14 @@ tabby.post('/download', async function (request, response) {
            }
        } else {
            console.error('API Permission error:', permissionResponse.status, permissionResponse.statusText);
-            return response.status(500).send({ error: true });
+            return response.status(permissionResponse.status).send({ error: true });
        }

        const fetchResponse = await fetch(`${baseUrl}/v1/download`, args);

        if (!fetchResponse.ok) {
            console.error('Download error:', fetchResponse.status, fetchResponse.statusText);
-            return response.status(500).send({ error: true });
+            return response.status(fetchResponse.status).send({ error: true });
        }

        return response.send({ ok: true });
--- a/src/endpoints/content-manager.js
+++ b/src/endpoints/content-manager.js
@@ -540,21 +540,9 @@ async function downloadGenericPng(url) {

        if (result.ok) {
            const buffer = Buffer.from(await result.arrayBuffer());
-            let fileName = sanitize(result.url.split('?')[0].split('/').reverse()[0]);
+            const fileName = sanitize(result.url.split('?')[0].split('/').reverse()[0]);
            const contentType = result.headers.get('content-type') || 'image/png'; //yoink it from AICC function lol

-            // The `importCharacter()` function detects the MIME (content-type) of the file
-            // using its file extension. The problem is that not all third-party APIs serve
-            // their cards with a `.png` extension. To support more third-party sites,
-            // dynamically append the `.png` extension to the filename if it doesn't
-            // already have a file extension.
-            if (contentType === 'image/png') {
-                const ext = fileName.match(/\.(\w+)$/); // Same regex used by `importCharacter()`
-                if (!ext) {
-                    fileName += '.png';
-                }
-            }
-
            return {
                buffer: buffer,
                fileName: fileName,
@@ -706,11 +694,10 @@ router.post('/importURL', async (request, response) => {
            type = 'character';
            result = await downloadRisuCharacter(uuid);
        } else if (isGeneric) {
-            console.info('Downloading from generic url:', url);
+            console.info('Downloading from generic url.');
            type = 'character';
            result = await downloadGenericPng(url);
        } else {
-            console.error(`Received an import for "${getHostFromUrl(url)}", but site is not whitelisted. This domain must be added to the config key "whitelistImportDomains" to allow import from this source.`);
            return response.sendStatus(404);
        }

--- a/src/endpoints/google.js
+++ b/src/endpoints/google.js
@@ -45,7 +45,7 @@ router.post('/caption-image', async (request, response) => {
        if (!result.ok) {
            const error = await result.json();
            console.error(`Google AI Studio API returned error: ${result.status} ${result.statusText}`, error);
-            return response.status(500).send({ error: true });
+            return response.status(result.status).send({ error: true });
        }

        /** @type {any} */
--- a/src/endpoints/novelai.js
+++ b/src/endpoints/novelai.js
@@ -270,7 +270,7 @@ router.post('/generate', async function (req, res) {
                    // ignore
                }

-                return res.status(500).send({ error: { message } });
+                return res.status(response.status).send({ error: { message } });
            }

            /** @type {any} */
--- a/src/endpoints/tokenizers.js
+++ b/src/endpoints/tokenizers.js
@@ -407,10 +407,6 @@ export function getTokenizerModel(requestModel) {
        return 'o1';
    }

-    if (requestModel.includes('o3') || requestModel.includes('o4-mini')) {
-        return 'o1';
-    }
-
    if (requestModel.includes('gpt-4o') || requestModel.includes('chatgpt-4o-latest')) {
        return 'gpt-4o';
    }
--- a/src/prompt-converters.js
+++ b/src/prompt-converters.js
@@ -364,7 +364,6 @@ export function convertGooglePrompt(messages, model, useSysPrompt, names) {
        'gemini-2.5-pro-exp-03-25',
        'gemini-2.0-pro-exp',
        'gemini-2.0-pro-exp-02-05',
-        'gemini-2.5-flash-preview-04-17',
        'gemini-2.0-flash-lite-preview',
        'gemini-2.0-flash-lite-preview-02-05',
        'gemini-2.0-flash',
@@ -509,12 +508,7 @@ export function convertGooglePrompt(messages, model, useSysPrompt, names) {
        if (index > 0 && message.role === contents[contents.length - 1].role) {
            parts.forEach((part) => {
                if (part.text) {
-                    const textPart = contents[contents.length - 1].parts.find(p => typeof p.text === 'string');
-                    if (textPart) {
-                        textPart.text += '\n\n' + part.text;
-                    } else {
-                        contents[contents.length - 1].parts.push(part);
-                    }
+                    contents[contents.length - 1].parts[0].text += '\n\n' + part.text;
                }
                if (part.inlineData || part.functionCall || part.functionResponse) {
                    contents[contents.length - 1].parts.push(part);
--- a/src/transformers.js
+++ b/src/transformers.js
@@ -3,16 +3,16 @@ import fs from 'node:fs';
 import process from 'node:process';
 import { Buffer } from 'node:buffer';

-import { pipeline, env, RawImage } from 'sillytavern-transformers';
+import { pipeline, env, RawImage } from '@huggingface/transformers';
 import { getConfigValue } from './util.js';

 configureTransformers();

 function configureTransformers() {
    // Limit the number of threads to 1 to avoid issues on Android
-    env.backends.onnx.wasm.numThreads = 1;
+    // env.backends.onnx.wasm.numThreads = 1;
    // Use WASM from a local folder to avoid CDN connections
-    env.backends.onnx.wasm.wasmPaths = path.join(process.cwd(), 'dist') + path.sep;
+    // env.backends.onnx.wasm.wasmPaths = path.join(process.cwd(), 'dist') + path.sep;
 }

 const tasks = {
@@ -115,9 +115,9 @@ async function migrateCacheToDataDir() {

 /**
 * Gets the transformers.js pipeline for a given task.
- * @param {import('sillytavern-transformers').PipelineType} task The task to get the pipeline for
+ * @param {import('@huggingface/transformers').PipelineType} task The task to get the pipeline for
 * @param {string} forceModel The model to use for the pipeline, if any
- * @returns {Promise<import('sillytavern-transformers').Pipeline>} The transformers.js pipeline
+ * @returns {Promise<import('@huggingface/transformers').Pipeline>} The transformers.js pipeline
 */
 export async function getPipeline(task, forceModel = '') {
    await migrateCacheToDataDir();