utilitas 1999.1.69 → 1999.1.71

package/lib/alan.mjs

@@ -1,8 +1,8 @@
 import { checkSearch, distill, search } from './web.mjs';
 import { create as createUoid } from './uoid.mjs';
-import { createWavHeader } from './media.mjs';
 import { end, loop } from './event.mjs';
 import { fileTypeFromBuffer } from 'file-type';
+import { packPcmToWav } from './media.mjs';
 import { v4 as uuidv4 } from 'uuid';
 
 import {
@@ -62,7 +62,7 @@ const [
     JINA_CLIP, VERTEX, GEMINI_25_PRO, SILICONFLOW, SF_DEEPSEEK_V3,
 ] = [
         'OpenAI', 'Gemini', 'OPENAI_TRAINING', 'Ollama', 'gpt-4o', 'o3',
-        'gemini-2.5-flash-preview-04-17', 'nova', 'deepseek-r1', '```',
+        'gemini-2.5-flash-preview-05-20', 'nova', 'deepseek-r1', '```',
         'text-embedding-3-small', 'text-embedding-3-large',
         'claude-3-7-sonnet@20250219', 'audio', 'wav', '[ATTACHMENTS]', 'CHAT',
         'OPENAI_VOICE', 'medium', 'low', 'high', 'medium', 'think', '<think>',
@@ -465,7 +465,7 @@ const init = async (options = {}) => {
             process.env['ANTHROPIC_VERTEX_PROJECT_ID'] = options.projectId;
             var model = models[0];
             var client = new ((
-                await need('@anthropic-ai/vertex-sdk')
+                await need('@anthropic-ai/vertex-sdk', { raw: true })
             ).AnthropicVertex)({ region: options?.region || 'us-east5' });
             setupAi({ provider, model, client, prompt: promptAnthropic });
             break;
@@ -707,12 +707,8 @@ const packResp = async (resp, options) => {
         const str = simpleText.indexOf(x);
         str >= 0 && (simpleText = simpleText.slice(0, str).trim());
     });
-    audio && (audio = Buffer.isBuffer(audio) ? audio : await convert(audio, {
-        input: BASE64, expected: BUFFER,
-    })) && audio.length && (audio = Buffer.concat([
-        createWavHeader(audio.length), audio
-    ])) && (audio = await convert(audio, {
-        input: BUFFER, expected: BUFFER, ...options || {},
+    audio = await ignoreErrFunc(async () => await packPcmToWav(audio, {
+        input: Buffer.isBuffer(audio) ? BUFFER : BASE64, expected: BUFFER,
     }));
     if (images?.length) {
         for (let i in images) {
@@ -1205,12 +1201,17 @@ const promptGemini = async (aiId, content, options = {}) => {
             systemInstruction, responseModalities: options.modalities || (
                 options.imageMode ? [TEXT, IMAGE] : undefined
             ), ...options?.config || {}, ...model?.tools && !options.jsonMode
-                && options.model !== GEMINI_20_FLASH ? (options.tools ?? {
+                // @todo: Gemini 2.5 flash 05-20 refusing to use certain tool calls
+                // https://discuss.ai.google.dev/t/gemini-2-5-flash-05-20-refusing-to-use-certain-tool-calls/84086
+                && ![GEMINI_20_FLASH, GEMINI_25_FLASH].includes(options.model)
+                ? (options.tools ?? {
                     tools: [
                         // @todo: Gemini will failed when using these tools together.
                         // https://ai.google.dev/gemini-api/docs/function-calling
                         // { codeExecution: {} },
                         // { googleSearch: {} },
+                        // { urlContext: {} },
+                        // @todo: test these tools in next version 👆
                         {
                             functionDeclarations: (
                                 await toolsGemini({ provider })
@@ -1222,7 +1223,11 @@ const promptGemini = async (aiId, content, options = {}) => {
     });
     const resp = await chat.sendMessageStream({ message: prompt });
     for await (const chunk of resp) {
-        event = chunk.candidates[0];
+        assert(
+            !chunk?.promptFeedback?.blockReason,
+            chunk?.promptFeedback?.blockReason
+        );
+        event = chunk?.candidates?.[0];
         let [deltaText, deltaImages] = ['', []];
         event?.content?.parts?.map(x => {
             if (x.text) { deltaText = x.text; }
@@ -1583,6 +1588,7 @@ export {
     getSession,
     init,
     initChat,
+    k,
     listFiles,
     listGptFineTuningEvents,
     listGptFineTuningJobs,

package/lib/manifest.mjs

@@ -1,7 +1,7 @@
 const manifest = {
     "name": "utilitas",
     "description": "Just another common utility for JavaScript.",
-    "version": "1999.1.69",
+    "version": "1999.1.71",
     "private": false,
     "homepage": "https://github.com/Leask/utilitas",
     "main": "index.mjs",
@@ -20,22 +20,21 @@ const manifest = {
     },
     "dependencies": {
         "file-type": "^20.5.0",
-        "mathjs": "^14.4.0",
+        "mathjs": "^14.5.0",
         "uuid": "^11.1.0"
     },
     "devDependencies": {
-        "@anthropic-ai/sdk": "^0.40.1",
-        "@anthropic-ai/vertex-sdk": "^0.7.0",
+        "@anthropic-ai/sdk": "^0.51.0",
+        "@anthropic-ai/vertex-sdk": "^0.11.3",
         "@ffmpeg-installer/ffmpeg": "^1.1.0",
         "@ffprobe-installer/ffprobe": "^2.1.2",
-        "@google-cloud/speech": "^7.0.1",
+        "@google-cloud/speech": "^7.1.0",
         "@google-cloud/storage": "^7.16.0",
-        "@google-cloud/text-to-speech": "^6.0.1",
         "@google-cloud/vision": "^5.1.0",
-        "@google/genai": "^0.10.0",
+        "@google/genai": "^1.0.0",
         "@mozilla/readability": "github:mozilla/readability",
-        "@sentry/node": "^9.15.0",
-        "@sentry/profiling-node": "^9.15.0",
+        "@sentry/node": "^9.22.0",
+        "@sentry/profiling-node": "^9.22.0",
         "acme-client": "^5.4.0",
         "browserify-fs": "^1.0.0",
         "buffer": "^6.0.3",
@@ -47,28 +46,28 @@ const manifest = {
         "jsdom": "^26.1.0",
         "lorem-ipsum": "^2.0.8",
         "mailgun.js": "^12.0.1",
-        "mailparser": "^3.7.2",
+        "mailparser": "^3.7.3",
         "mime": "^4.0.7",
         "mysql2": "^3.14.1",
         "node-mailjet": "^6.0.8",
         "node-polyfill-webpack-plugin": "^4.1.0",
         "office-text-extractor": "^3.0.3",
-        "openai": "^4.96.2",
+        "openai": "^4.100.0",
         "pdfjs-dist": "^5.2.133",
-        "pg": "^8.15.6",
-        "pgvector": "^0.2.0",
+        "pg": "^8.16.0",
+        "pgvector": "^0.2.1",
         "ping": "^0.4.4",
         "process": "^0.11.10",
-        "puppeteer": "^24.7.2",
+        "puppeteer": "^24.9.0",
         "say": "^0.16.0",
         "telegraf": "^4.16.3",
-        "telesignsdk": "^3.0.2",
+        "telesignsdk": "^3.0.3",
         "tesseract.js": "^6.0.1",
-        "twilio": "^5.5.2",
+        "twilio": "^5.6.1",
         "url": "github:Leask/node-url",
         "webpack-cli": "^6.0.1",
         "whisper-node": "^1.1.1",
-        "wrangler": "^4.14.0",
+        "wrangler": "^4.16.0",
         "xlsx": "https://cdn.sheetjs.com/xlsx-0.20.1/xlsx-0.20.1.tgz",
         "youtube-transcript": "^1.2.1"
     }

package/lib/media.mjs

@@ -38,6 +38,17 @@ const createWavHeader = (
     return header;
 };
 
+const packPcmToWav = async (audio, options) => {
+    (audio = await convert(audio, { ...options || {}, expected: BUFFER })) // DON'T override expected
+        && audio.length
+        && (audio = Buffer.concat([createWavHeader(audio.length), audio]))
+        && (audio = await convert(audio, {
+            expected: BUFFER, ...options || {}, input: BUFFER, // DON'T override input
+        }));
+    assert(audio, 'Failed to pack PCM to WAV.', 500);
+    return audio;
+};
+
 // https://codex.so/ffmpeg-node-js
 const getFfmpeg = async (options) => {
     const ffmpeg = await need('fluent-ffmpeg');
@@ -92,4 +103,5 @@ export {
     convertAudioTo16kNanoPcmWave,
     createWavHeader,
     getFfmpeg,
+    packPcmToWav,
 };

package/lib/speech.mjs

@@ -1,6 +1,6 @@
-import { DEFAULT_MODELS, OPENAI_VOICE } from './alan.mjs';
+import { DEFAULT_MODELS, OPENAI_VOICE, countTokens, k } from './alan.mjs';
 import { getApiKeyCredentials, hash } from './encryption.mjs';
-import { getFfmpeg } from './media.mjs';
+import { getFfmpeg, packPcmToWav } from './media.mjs';
 import { get } from './web.mjs';
 import { convert, getTempPath } from './storage.mjs';
 import { ensureString } from './utilitas.mjs';
@@ -17,19 +17,32 @@ import {
 
 const _NEED = [
     '@google-cloud/speech',
-    '@google-cloud/text-to-speech',
+    '@google/genai',
     'OpenAI',
     'whisper-node',
 ];
 
 const WHISPER_DEFAULT_MODEL = 'base';
 const errorMessage = 'Invalid audio data.';
-const [BUFFER, STREAM, BASE64, FILE, clients, languageCode, audioEncoding, suffix, SPEAKER, cleanup]
-    = ['BUFFER', 'STREAM', 'BASE64', 'FILE', {}, 'en-US', 'OGG_OPUS', 'ogg', 'SPEAKER', true];
-const [GPT_4O_MIMI_TTS, GPT_4O_TRANSCRIBE, OPENAI_TTS_MAX_LENGTH]
-    = ['gpt-4o-mini-tts', 'gpt-4o-transcribe', 4096];
-const [defaultOpenAITtsModel, defaultOpenAISttModel]
-    = [GPT_4O_MIMI_TTS, GPT_4O_TRANSCRIBE];
+
+const [
+    BUFFER, STREAM, BASE64, FILE, clients, languageCode, audioEncoding, suffix,
+    SPEAKER, cleanup, wav,
+] = [
+        'BUFFER', 'STREAM', 'BASE64', 'FILE', {}, 'en-US', 'OGG_OPUS', 'ogg',
+        'SPEAKER', true, 'wav'
+    ];
+
+const [
+    GPT_4O_MIMI_TTS, GPT_4O_TRANSCRIBE, GEMINI_25_PRO_TTS, GEMINI_25_FLASH_TTS,
+    OPENAI_TTS_MAX_LENGTH,
+] = [
+        'gpt-4o-mini-tts', 'gpt-4o-transcribe', 'gemini-2.5-pro-preview-tts',
+        'gemini-2.5-flash-preview-tts', 4096
+    ];
+
+const [defaultOpenAITtsModel, defaultOpenAISttModel, defaultGeminiTtsModel]
+    = [GPT_4O_MIMI_TTS, GPT_4O_TRANSCRIBE, GEMINI_25_PRO_TTS];
 
 const WHISPER_MODELS = [
     // npx whisper-node download tiny.en
@@ -104,13 +117,14 @@ const init = async (options) => {
                 break;
             case 'GOOGLE':
                 clients._provider = provider;
-                const sslCreds = await getApiKeyCredentials(options);
                 if (options?.tts) {
-                    const tts = (await need('@google-cloud/text-to-speech')).default;
-                    clients.tts = new tts.TextToSpeechClient({ sslCreds });
+                    let { GoogleGenAI } = await need('@google/genai');
+                    let client = new GoogleGenAI(options);
+                    clients.tts = client.models.generateContent;
                 }
                 if (options?.stt) {
                     const stt = (await need('@google-cloud/speech')).default;
+                    const sslCreds = await getApiKeyCredentials(options);
                     clients.stt = new stt.SpeechClient({ sslCreds });
                 }
                 break;
@@ -159,15 +173,29 @@ const ttsOpenAI = async (input, options) => {
     return await convert(buffer, { suffix, ...options || {} });
 };
 
-const ttsGoogle = async (text, options) => {
+// https://ai.google.dev/gemini-api/docs/speech-generation#voices
+const ttsGoogle = async (contents, options) => {
     assert(clients.tts, 'Google TTS API has not been initialized.', 500);
-    assert(text, 'Text is required.', 400);
-    const [response] = await clients.tts.synthesizeSpeech({
-        input: { text, ...options?.input || {} },
-        voice: { languageCode, name: 'en-US-Wavenet-F', ...options?.voice || {} },
-        audioConfig: { audioEncoding, ...options?.audioConfig || {} },
+    assert(contents, 'Text is required.', 400);
+    assert(await countTokens(contents) <= k(32), 'Text is too long.', 400);
+    const resp = await clients.tts({
+        model: options?.model || defaultGeminiTtsModel, contents,
+        config: {
+            responseModalities: ['AUDIO'],
+            speechConfig: {
+                voiceConfig: {
+                    prebuiltVoiceConfig: {
+                        voiceName: options?.voice || 'Leda',
+                    },
+                },
+            },
+        },
+    });
+    const rawAudio = resp?.candidates?.[0]?.content?.parts?.[0]?.inlineData;
+    assert(rawAudio, 'Failed to generate audio.', 500);
+    return options?.raw ? rawAudio : await packPcmToWav(rawAudio?.data, {
+        input: BASE64, expected: 'FILE', suffix: wav, ...options || {},
     });
-    return await convert(response.audioContent, { suffix, ...options || {} });
 };
 
 const ttsSay = async (text, options) => {
@@ -275,13 +303,16 @@ const stt = async (audio, options) => {
 export default init;
 export {
     _NEED,
+    OPENAI_TTS_MAX_LENGTH,
     checkSay,
     checkWhisper,
-    init, OPENAI_TTS_MAX_LENGTH, stt, sttGoogle,
+    init,
+    stt,
+    sttGoogle,
     sttOpenAI,
     sttWhisper,
     tts,
     ttsGoogle,
     ttsOpenAI,
-    ttsSay
+    ttsSay,
 };

package/package.json

@@ -1,7 +1,7 @@
 {
     "name": "utilitas",
     "description": "Just another common utility for JavaScript.",
-    "version": "1999.1.69",
+    "version": "1999.1.71",
     "private": false,
     "homepage": "https://github.com/Leask/utilitas",
     "main": "index.mjs",
@@ -31,22 +31,21 @@
     },
     "dependencies": {
         "file-type": "^20.5.0",
-        "mathjs": "^14.4.0",
+        "mathjs": "^14.5.0",
         "uuid": "^11.1.0"
     },
     "devDependencies": {
-        "@anthropic-ai/sdk": "^0.40.1",
-        "@anthropic-ai/vertex-sdk": "^0.7.0",
+        "@anthropic-ai/sdk": "^0.51.0",
+        "@anthropic-ai/vertex-sdk": "^0.11.3",
         "@ffmpeg-installer/ffmpeg": "^1.1.0",
         "@ffprobe-installer/ffprobe": "^2.1.2",
-        "@google-cloud/speech": "^7.0.1",
+        "@google-cloud/speech": "^7.1.0",
         "@google-cloud/storage": "^7.16.0",
-        "@google-cloud/text-to-speech": "^6.0.1",
         "@google-cloud/vision": "^5.1.0",
-        "@google/genai": "^0.10.0",
+        "@google/genai": "^1.0.0",
         "@mozilla/readability": "github:mozilla/readability",
-        "@sentry/node": "^9.15.0",
-        "@sentry/profiling-node": "^9.15.0",
+        "@sentry/node": "^9.22.0",
+        "@sentry/profiling-node": "^9.22.0",
         "acme-client": "^5.4.0",
         "browserify-fs": "^1.0.0",
         "buffer": "^6.0.3",
@@ -58,28 +57,28 @@
         "jsdom": "^26.1.0",
         "lorem-ipsum": "^2.0.8",
         "mailgun.js": "^12.0.1",
-        "mailparser": "^3.7.2",
+        "mailparser": "^3.7.3",
         "mime": "^4.0.7",
         "mysql2": "^3.14.1",
         "node-mailjet": "^6.0.8",
         "node-polyfill-webpack-plugin": "^4.1.0",
         "office-text-extractor": "^3.0.3",
-        "openai": "^4.96.2",
+        "openai": "^4.100.0",
         "pdfjs-dist": "^5.2.133",
-        "pg": "^8.15.6",
-        "pgvector": "^0.2.0",
+        "pg": "^8.16.0",
+        "pgvector": "^0.2.1",
         "ping": "^0.4.4",
         "process": "^0.11.10",
-        "puppeteer": "^24.7.2",
+        "puppeteer": "^24.9.0",
         "say": "^0.16.0",
         "telegraf": "^4.16.3",
-        "telesignsdk": "^3.0.2",
+        "telesignsdk": "^3.0.3",
         "tesseract.js": "^6.0.1",
-        "twilio": "^5.5.2",
+        "twilio": "^5.6.1",
         "url": "github:Leask/node-url",
         "webpack-cli": "^6.0.1",
         "whisper-node": "^1.1.1",
-        "wrangler": "^4.14.0",
+        "wrangler": "^4.16.0",
         "xlsx": "https://cdn.sheetjs.com/xlsx-0.20.1/xlsx-0.20.1.tgz",
         "youtube-transcript": "^1.2.1"
     }