halbot 1993.2.90 → 1993.2.91

package/lib/hal.mjs

@@ -1,8 +1,5 @@
-import {
-    alan, bot, callosum, dbio, storage, uoid, utilitas, web,
-} from 'utilitas';
-
 import { basename, join } from 'path';
+import { bot, callosum, dbio, storage, uoid, utilitas, web } from 'utilitas';
 import { parseArgs as _parseArgs } from 'node:util';
 import { readdirSync } from 'fs';
 
@@ -612,7 +609,11 @@ const subconscious = [{
                 const analyze = async () => {
                     const resp = await utilitas.ignoreErrFunc(async () => {
                         [
-                            alan.mp3, alan.mpega, alan.mp4, alan.mpeg, alan.mpga, alan.m4a, alan.wav, alan.webm, alan.ogg
+                            storage.MIME_MP3, storage.MIME_MPEGA,
+                            storage.MIME_MP4, storage.MIME_MPEG,
+                            storage.MIME_MPGA, storage.MIME_M4A,
+                            storage.MIME_WAV, storage.MIME_WEBM,
+                            storage.MIME_OGG,
                         ].includes(audio.mime_type) || (
                                 file = await media.convertAudioTo16kNanoPcmWave(
                                     file, { input: storage.BUFFER, expected: storage.BUFFER }
@@ -623,9 +624,9 @@ const subconscious = [{
                     log(`STT: '${resp}'`);
                     ctx.collect(resp);
                 };
-                if (hal._.supportedMimeTypes.has(alan.wav)) {
+                if (hal._.supportedMimeTypes.has(MIME_WAV)) {
                     ctx.collect({
-                        mime_type: alan.wav, url, analyze,
+                        mime_type: MIME_WAV, url, analyze,
                         data: await media.convertAudioTo16kNanoPcmWave(file, {
                             input: storage.BUFFER, expected: storage.BASE64,
                         }),
@@ -736,9 +737,10 @@ const subconscious = [{
             if (m.photo?.[m.photo?.length - 1]) {
                 const p = m.photo[m.photo.length - 1];
                 files.push({
-                    asPrompt: hal._.supportedMimeTypes.has(alan.jpeg),
+                    asPrompt: hal._.supportedMimeTypes.has(storage.MIME_JPEG),
                     file_name: `${p.file_id}.jpg`, fileId: p.file_id,
-                    mime_type: alan.jpeg, type: 'PHOTO', ocrFunc: ctx._.vision?.see,
+                    mime_type: storage.MIME_JPEG, type: 'PHOTO',
+                    ocrFunc: ctx._.vision?.see,
                 });
             }
             if (m.video_note) {

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "halbot",
   "description": "Just another `ChatGPT` / `Gemini` / `Claude` / `Azure` / `Jina` / `Ollama` Telegram bob, which is simple design, easy to use, extendable and fun.",
-  "version": "1993.2.90",
+  "version": "1993.2.91",
   "private": false,
   "homepage": "https://github.com/Leask/halbot",
   "type": "module",
@@ -46,12 +46,12 @@
     "mime": "^4.0.7",
     "mysql2": "^3.14.1",
     "office-text-extractor": "^3.0.3",
-    "openai": "^4.96.2",
+    "openai": "^4.97.0",
     "pg": "^8.15.6",
     "pgvector": "^0.2.0",
     "telegraf": "^4.16.3",
     "tesseract.js": "^6.0.1",
-    "utilitas": "^1999.1.60",
+    "utilitas": "^1999.1.61",
     "youtube-transcript": "^1.2.1"
   }
 }

package/skills/40_dream.mjs

@@ -1,13 +1,19 @@
-import { bot } from '../index.mjs';
+import { bot, storage } from '../index.mjs';
 
 const GEMINI = 'GEMINI';
 const types = { image: 'photo', video: 'video' };
 
 const action = async (ctx, next) => {
-    let [provider, func] = [GEMINI, 'image'];
+    let [provider, func, reference] = [GEMINI, 'image', null];
     switch (ctx.cmd.cmd) {
-        case 'gptimage': provider = 'OPENAI'; break;
-        case 'fantasy': func = 'video';
+        case 'fantasy': func = 'video'; break;
+        case 'gptimage':
+            provider = 'OPENAI';
+            reference = ctx.collected.filter(x => [
+                storage.MIME_JPEG, storage.MIME_PNG, storage.MIME_WEBP
+            ].includes(x?.content?.mime_type)).slice(0, 16).map(
+                x => x?.content?.data
+            );
     }
     if (!ctx.cmd.args) {
         return await ctx.ok('Please input your prompt.');
@@ -15,7 +21,8 @@ const action = async (ctx, next) => {
     let [objMsg, output] = [(await ctx.ok('💭'))[0], null]; //tts = null
     try {
         output = (await ctx._.gen[func](ctx.cmd.args, {
-            provider, expected: 'FILE'
+            provider, expected: 'FILE',
+            ...reference?.length ? { reference, input: 'BASE64' } : {},
         })) || [];
     } catch (err) {
         return await ctx.er(err.message || `Error generating ${func}.`,