utilitas 1999.1.85 → 1999.1.87

package/lib/alan.mjs

@@ -10,7 +10,7 @@ import {
     MIME_JPEG, MIME_MOV, MIME_MPEG, MIME_MP4, MIME_MPG, MIME_AVI, MIME_WMV,
     MIME_MPEGPS, MIME_FLV, MIME_GIF, MIME_WEBP, MIME_PDF, MIME_AAC, MIME_FLAC,
     MIME_MP3, MIME_MPEGA, MIME_M4A, MIME_MPGA, MIME_OPUS, MIME_PCM, MIME_WAV,
-    MIME_WEBM, MIME_TGPP, MIME_PCM16, STREAM, convert, formatDataURL
+    MIME_WEBM, MIME_TGPP, MIME_PCM16, MIME_OGG, STREAM, convert, formatDataURL
 } from './storage.mjs';
 
 import {
@@ -123,7 +123,10 @@ const GEMINI_RULES = {
         MIME_WMV, MIME_MPEGPS, MIME_FLV, MIME_PDF, MIME_AAC, MIME_FLAC,
         MIME_MP3, MIME_MPEGA, MIME_M4A, MIME_MPGA, MIME_OPUS, MIME_PCM,
         MIME_WAV, MIME_WEBM, MIME_TGPP,
-    ], defaultProvider: GEMINI,
+    ], supportedAudioTypes: [MIME_WAV, MIME_OGG, MIME_OPUS],
+    audio: 'gemini-2.5-flash-exp-native-audio-thinking-dialog',
+    defaultProvider: GEMINI,
+    // gemini-2.5-flash-preview-native-audio-dialog
 };
 
 const DEEPSEEK_R1_RULES = {
@@ -203,7 +206,7 @@ for (const n in MODELS) {
 const DEFAULT_MODELS = {
     [OPENAI]: GPT_5,
     [SILICONFLOW]: SF_DEEPSEEK_R1,
-    [GEMINI]: GEMINI_25_FLASH,
+    [GEMINI]: GEMINI_25_PRO,
     [ANTHROPIC]: CLOUD_41_OPUS,
     [VERTEX_ANTHROPIC]: CLOUD_41_OPUS,
     [JINA]: JINA_DEEPSEARCH,
@@ -557,7 +560,7 @@ const countTokens = async (input, options) => {
     );
 };
 
-const selectGptVisionModel = options => {
+const selectVisionModel = options => {
     assert(
         MODELS[options.model]?.vision,
         `Vision modality is not supported by model: ${options.model}`
@@ -566,7 +569,7 @@ const selectGptVisionModel = options => {
         ? MODELS[options.model].vision : null;
 };
 
-const selectGptAudioModel = options => {
+const selectAudioModel = options => {
     assert(
         MODELS[options.model]?.audio,
         `Audio modality is not supported by model: ${options.model}`
@@ -577,17 +580,17 @@ const selectGptAudioModel = options => {
 
 const buildGptMessage = (content, options) => {
     content = content || '';
-    let alterModel = options?.audioMode && selectGptAudioModel(options);
+    let alterModel = options?.audioMode && selectAudioModel(options);
     const attachments = (options?.attachments || []).map(x => {
         assert(MODELS[options?.model], 'Model is required.');
         if (MODELS[options.model]?.supportedMimeTypes?.includes?.(x.mime_type)) {
-            alterModel = selectGptVisionModel(options); // URL or Base64URL
+            alterModel = selectVisionModel(options); // URL or Base64URL
             return {
                 type: 'image_url',
                 image_url: { url: x.url, detail: 'high' },
             };
         } else if (MODELS[options.model]?.supportedDocTypes?.includes?.(x.mime_type)) {
-            alterModel = selectGptVisionModel(options);
+            alterModel = selectVisionModel(options);
             return {
                 type: 'file',
                 file: {
@@ -597,7 +600,7 @@ const buildGptMessage = (content, options) => {
                 },
             };
         } else if (MODELS[options.model]?.supportedAudioTypes?.includes?.(x.mime_type)) {
-            alterModel = selectGptAudioModel(options);
+            alterModel = selectAudioModel(options);
             return {
                 type: 'input_audio',
                 input_audio: { data: x.data, format: WAV },
@@ -627,9 +630,24 @@ const buildGeminiParts = (text, attachments) => {
 
 const buildGeminiMessage = (content, options) => {
     content = content || '';
-    const attachments = (options?.attachments?.length ? options.attachments : []).map(x => ({
-        inlineData: { mimeType: x.mime_type, data: x.data }
-    }));
+    // @todo: for future audio mode support
+    // let alterModel = options?.audioMode && selectAudioModel(options);
+    const attachments = (
+        options?.attachments?.length ? options.attachments : []
+    ).map(x => {
+        // assert(MODELS[options?.model], 'Model is required.');
+        // if (MODELS[options.model]?.supportedAudioTypes?.includes?.(x.mime_type)
+        //     && !options.imageMode) {
+        //     alterModel = selectAudioModel(options);
+        // }
+        return {
+            inlineData: { mimeType: x.mime_type, data: x.data }
+        };
+    });
+    // if (alterModel) {
+    //     options.model = alterModel;
+    //     options.audioMode = true;
+    // }
     return String.isString(content) ? (options?.history ? {
         role: options?.role || user,
         parts: buildGeminiParts(content, attachments),
@@ -1190,17 +1208,19 @@ const promptGemini = async (aiId, content, options = {}) => {
         options.imageMode = true;
         model = MODELS[options.model];
     }
+    options.flavor = GEMINI;
     const { systemPrompt: systemInstruction, history, prompt }
-        = await buildPrompts(model, content, { ...options, flavor: GEMINI });
+        = await buildPrompts(model, content, options);
+    const responseModalities = options.modalities
+        || (options.imageMode ? [TEXT, IMAGE] : undefined)
+        || (options.audioMode ? [TEXT, AUDIO] : undefined);
     const chat = client.chats.create({
         model: options.model, history, config: {
             responseMimeType: options.jsonMode ? MIME_JSON : MIME_TEXT,
             ...model.reasoning ? {
                 thinkingConfig: { includeThoughts: true },
-            } : {}, systemInstruction,
-            responseModalities: options.modalities || (
-                options.imageMode ? [TEXT, IMAGE] : undefined
-            ), ...options?.config || {}, ...model?.tools && !options.jsonMode
+            } : {}, systemInstruction, responseModalities,
+            ...options?.config || {}, ...model?.tools && !options.jsonMode
                 && ![GEMINI_20_FLASH].includes(options.model)
                 ? (options.tools ?? {
                     tools: [

package/lib/manifest.mjs

@@ -1,7 +1,7 @@
 const manifest = {
     "name": "utilitas",
     "description": "Just another common utility for JavaScript.",
-    "version": "1999.1.85",
+    "version": "1999.1.87",
     "private": false,
     "homepage": "https://github.com/Leask/utilitas",
     "main": "index.mjs",

package/package.json

@@ -1,7 +1,7 @@
 {
     "name": "utilitas",
     "description": "Just another common utility for JavaScript.",
-    "version": "1999.1.85",
+    "version": "1999.1.87",
     "private": false,
     "homepage": "https://github.com/Leask/utilitas",
     "main": "index.mjs",