utilitas 1993.3.21 → 1993.3.22

package/index.mjs

@@ -26,6 +26,7 @@ import * as storage from './lib/storage.mjs';
 import * as tape from './lib/tape.mjs';
 import * as uoid from './lib/uoid.mjs';
 import * as utilitas from './lib/utilitas.mjs';
+import * as vision from './lib/vision.mjs';
 import * as web from './lib/web.mjs';
 import color from './lib/color.mjs';
 import manifest from './lib/manifest.mjs';
@@ -38,7 +39,7 @@ export {
     // features
     bot, boxes, cache, callosum, color, dbio, email, encryption, event, hal,
     manifest, memory, network, sentinel, shekel, shell, shot, sms, speech, ssl,
-    storage, tape, uoid, utilitas, web
+    storage, tape, uoid, utilitas, vision, web
 };
 
 if (utilitas.inBrowser() && !globalThis.utilitas) {

package/lib/bot.mjs

@@ -42,6 +42,8 @@ const [ // https://limits.tginfo.me/en
         256, 'bot_command', '🗣️', '👀', '🤖',
     ];
 
+const [BUFFER_ENCODE] = [{ encode: BUFFER }];
+
 const KNOWN_UPDATE_TYPES = [
     'message', 'callback_query', 'channel_post', 'edited_message',
     'my_chat_member'
@@ -230,8 +232,10 @@ const subconscious = [{
             }
             return target === ctx.botInfo.username;
         })) { ctx.chatType = MENTION; }
-        if ((ctx.text || ctx.msg.voice || ctx.msg.poll || ctx.msg.data)
-            && ctx.messageId) { await next(); }
+        if ((ctx.text || ctx.msg.voice || ctx.msg.poll || ctx.msg.data
+            || ctx.msg.document || ctx.msg.photo) && ctx.messageId) {
+            await next();
+        }
         await sessionSet(ctx.chatId);
     },
 }, {
@@ -281,9 +285,7 @@ const subconscious = [{
         if (ctx._.speech?.stt && ctx.msg.voice?.mime_type === 'audio/ogg') {
             await ctx.ok(EMOJI_SPEECH);
             try {
-                const file = await getFile(
-                    ctx.msg.voice.file_id, { encode: BUFFER }
-                );
+                const file = await getFile(ctx.msg.voice.file_id, BUFFER_ENCODE);
                 const resp = await ignoreErrFunc(
                     async () => await ctx._.speech?.stt(
                         file, { config: sampleRateHertz }
@@ -334,7 +336,31 @@ const subconscious = [{
         await next();
     },
 }, {
-    run: true, priority: -8880, name: 'commands', func: async (ctx, next) => {
+    run: true, priority: -8880, name: 'ocr', func: async (ctx, next) => {
+        if (!ctx._.vision?.ocrImage) { return await next(); }
+        let fileId;
+        if (/^image\/.*$/ig.test(ctx.msg?.document?.mime_type)) {
+            fileId = ctx.msg.document.file_id;
+        } else if (ctx.msg?.photo) {
+            fileId = ctx.msg.photo[ctx.msg.photo.length - 1]?.file_id;
+        }
+        if (fileId) {
+            await ctx.ok(EMOJI_LOOK);
+            try {
+                const file = await getFile(fileId, BUFFER_ENCODE);
+                const content = await ignoreErrFunc(
+                    async () => await ctx._.vision?.ocrImage(
+                        file, BUFFER_ENCODE
+                    ), { log: true }
+                );
+                console.log(content);
+                ctx.collect(content, 'OCR');
+            } catch (err) { return await ctx.er(err); }
+        }
+        await next();
+    },
+}, {
+    run: true, priority: -8870, name: 'commands', func: async (ctx, next) => {
         for (let ent of ctx.msg?.entities || []) {
             if (ent.type !== bot_command) { continue; }
             const rawCmd = ctx.text.substr(ent.offset, ent.length);
@@ -349,7 +375,7 @@ const subconscious = [{
         await next();
     },
 }, {
-    run: true, priority: -8870, name: 'help', func: async (ctx, next) => {
+    run: true, priority: -8860, name: 'help', func: async (ctx, next) => {
         const help = ctx._.info ? [ctx._.info] : [];
         for (let i in ctx._.skills) {
             const _help = [];
@@ -387,7 +413,7 @@ const subconscious = [{
         echo: 'Show debug message.',
     },
 }, {
-    run: true, priority: -8870, name: 'configuration', func: async (ctx, next) => {
+    run: true, priority: -8850, name: 'configuration', func: async (ctx, next) => {
         switch (ctx.cmd.cmd) {
             case 'set':
                 try {
@@ -506,6 +532,7 @@ const init = async (options) => {
                 private: options?.private && new Set(options.private),
                 session: { get: options?.session?.get, set: options?.session?.set },
                 speech: options?.speech,
+                vision: options?.vision,
                 skills: { ...options?.skills || {} },
             };
             (!options?.session?.get || !options?.session?.set)

package/lib/encryption.mjs

@@ -1,6 +1,7 @@
 import { createHash, randomBytes as random } from 'crypto';
 import { createReadStream } from 'fs';
 import { ensureString } from './utilitas.mjs';
+import { need } from './utilitas.mjs';
 import { networkInterfaces } from 'os';
 
 const defaultAlgorithm = 'sha256';
@@ -42,9 +43,20 @@ const hexToBigInt = (hex) => {
     return BigInt(hex, 16).toString(10);
 };
 
+const getApiKeyCredentials = async (options) => {
+    // Included in @google-cloud/vision, @google-cloud/speech and @google-cloud/text-to-speech
+    const { GoogleAuth, grpc } = await need('google-gax');
+    const authClient = new GoogleAuth().fromAPIKey(options?.apiKey);
+    return grpc.credentials.combineChannelCredentials(
+        grpc.credentials.createSsl(),
+        grpc.credentials.createFromGoogleCredential(authClient)
+    );
+};
+
 export {
     defaultAlgorithm,
     digestObject,
+    getApiKeyCredentials,
     getSortedQueryString,
     hash as sha256,
     hash,

package/lib/manifest.mjs

@@ -1,7 +1,7 @@
 const manifest = {
     "name": "utilitas",
     "description": "Just another common utility for JavaScript.",
-    "version": "1993.3.21",
+    "version": "1993.3.22",
     "private": false,
     "homepage": "https://github.com/Leask/utilitas",
     "main": "index.mjs",
@@ -23,6 +23,7 @@ const manifest = {
     "devDependencies": {
         "@google-cloud/speech": "^5.4.0",
         "@google-cloud/text-to-speech": "^4.2.1",
+        "@google-cloud/vision": "^3.1.2",
         "@mozilla/readability": "^0.4.4",
         "@sentry/node": "^7.47.0",
         "@waylaidwanderer/chatgpt-api": "^1.34.0",

package/lib/speech.mjs

@@ -2,6 +2,7 @@ import {
     base64Encode, log as _log, countKeys, ensureString, need, throwError
 } from './utilitas.mjs';
 
+import { getApiKeyCredentials } from './encryption.mjs';
 import { readFile, writeFile, writeTempFile } from './storage.mjs';
 
 const _NEED = ['@google-cloud/speech', '@google-cloud/text-to-speech'];
@@ -9,16 +10,6 @@ const log = (content) => _log(content, import.meta.url);
 const [BUFFER, BASE64, FILE, clients, languageCode, audioEncoding, suffix, encoding]
     = ['BUFFER', 'BASE64', 'FILE', {}, 'en-US', 'OGG_OPUS', 'OGG', { encoding: 'binary' }];
 
-const getApiKeyCredentials = async (options) => {
-    // Included in @google-cloud/speech or @google-cloud/text-to-speech
-    const { GoogleAuth, grpc } = await need('google-gax');
-    const authClient = new GoogleAuth().fromAPIKey(options?.apiKey);
-    return grpc.credentials.combineChannelCredentials(
-        grpc.credentials.createSsl(),
-        grpc.credentials.createFromGoogleCredential(authClient)
-    );
-};
-
 const init = async (options) => {
     if (options) {
         assert(options?.apiKey, 'Google Cloud API Key is required.', 500);

package/lib/vision.mjs

@@ -0,0 +1,42 @@
+import { base64Decode, ensureString, need, throwError } from './utilitas.mjs';
+import { getApiKeyCredentials } from './encryption.mjs';
+import { writeTempFile } from './storage.mjs';
+
+const _NEED = ['@google-cloud/vision'];
+const [BUFFER, BASE64, FILE, encoding]
+    = ['BUFFER', 'BASE64', 'FILE', { encoding: 'binary' }];
+
+let client;
+
+const init = async (options) => {
+    if (options) {
+        assert(options?.apiKey, 'Google Cloud API Key is required.', 500);
+        const sslCreds = await getApiKeyCredentials(options);
+        const vision = (await need('@google-cloud/vision')).default;
+        client = new vision.ImageAnnotatorClient({ sslCreds });
+    }
+    assert(client, 'Vision API client has not been initialized.', 501);
+    return client;
+};
+
+const ocrImage = async (image, options) => {
+    assert(client, 'Vision API has not been initialized.', 500);
+    assert(image, 'Image data is required.', 400);
+    let content;
+    switch (ensureString(options?.input, { case: 'UP' })) {
+        case BASE64: image = base64Decode(image, true);
+        case BUFFER: case '': content = await writeTempFile(image, encoding); break;
+        case FILE: content = image; break;
+        default: throwError('Invalid input format.', 400);
+    }
+    const [response] = await client.textDetection(content);
+    let detections = response.textAnnotations;
+    options?.raw || (detections = detections.map(t => t.description).join(' '));
+    return detections;
+};
+
+export {
+    _NEED,
+    init,
+    ocrImage,
+};

package/package.json

@@ -1,7 +1,7 @@
 {
     "name": "utilitas",
     "description": "Just another common utility for JavaScript.",
-    "version": "1993.3.21",
+    "version": "1993.3.22",
     "private": false,
     "homepage": "https://github.com/Leask/utilitas",
     "main": "index.mjs",
@@ -34,6 +34,7 @@
     "devDependencies": {
         "@google-cloud/speech": "^5.4.0",
         "@google-cloud/text-to-speech": "^4.2.1",
+        "@google-cloud/vision": "^3.1.2",
         "@mozilla/readability": "^0.4.4",
         "@sentry/node": "^7.47.0",
         "@waylaidwanderer/chatgpt-api": "^1.34.0",