utilitas 1998.2.7 → 1998.2.9

package/lib/alan.mjs

@@ -543,11 +543,15 @@ const packResp = async (resp, options) => {
         // DeepSeek R1 {
         let lines = (richText || txt).split('\n');
         const indexOfEnd = lines.indexOf(THINK_END);
-        if (lines[0] === THINK_STR && indexOfEnd !== -1) {
-            lines[0] = MD_CODE + THINK;
-            lines[indexOfEnd] = MD_CODE;
-            lines.slice(1, indexOfEnd).join('').trim()
-                || (lines = lines.slice(indexOfEnd + 1));
+        if (lines[0] === THINK_STR) {
+            if (indexOfEnd === -1) {
+                lines.shift();
+            } else {
+                lines[0] = MD_CODE + THINK;
+                lines[indexOfEnd] = MD_CODE;
+                lines.slice(1, indexOfEnd).join('').trim()
+                    || (lines = lines.slice(indexOfEnd + 1));
+            }
             richText = lines.join('\n').trim();
         }
         // }
@@ -1083,9 +1087,15 @@ const getMaxChatPromptLimit = (options) => {
 
 const distillFile = async (attachments, o) => {
     const strPmt = o?.prompt || (
-        'You are an intelligent document scanner.'
-        + ' Please help me scan the following document and only return the scanning results.'
-        + (o.summarize ? ' Please organize the content of this document, systematically present the information it contains, and remove any unimportant filler content.' : '')
+        'You are an intelligent document analyzer.'
+        + ' You will receive various multimedia files, including images, audio, and videos.'
+        + ' Please analyze these documents, extract the information, and organize it into an easy-to-read format.'
+        + ' For document-type files or image files primarily containing text information, act as a document scanner, return the text content, and describe any images and tables present.'
+        + ' For audio files, please provide a transcript of the spoken voices. If there are background noises or music, attempt to briefly describe the environmental sounds and music sections.'
+        + ' For images or video files that are not primarily text-based, describe the tragic scene you observe, highlight key details, convey the emotional tone of the setting, and share your impressions.'
+        + ' For video files, please describe the content, including the theme, subjects, characters, scenes, objects, storyline, and emotional tone.'
+        + ' Please RETURN ONLY your analysis results without including your thought process or other unrelated information.'
+        + (o.summarize ? ' Please organize the key content of this document, systematically present the key information it contains in a concise summary, and remove any unimportant filler content.' : '')
         + (o.toLanguage ? ` Please return the results in ${o.toLanguage}.` : '')
         + (o.keepPaging ? '' : ' If the document has multiple pages, merge them into one page. Please do not return any paging information.')
         + (o.keepDecoration ? '' : ' If the document has headers, footers, or watermarks, please ignore them.')
@@ -1103,12 +1113,16 @@ const distillFile = async (attachments, o) => {
     }
     attachments = await Promise.all(attachments);
     // print(attachments);
-    return await prompt(strPmt, { fast: true, ...o, attachments });
+    return await prompt(strPmt, {
+        fast: true, multimodal: true, ...o, attachments,
+    });
 };
 
 const prompt = async (input, options) => {
     let egn = options?.engine && unifyEngine(options);
-    for (const engine of PREFERRED_ENGINES) {
+    const engines = PREFERRED_ENGINES.slice();
+    options?.multimodal && engines.sort((x, y) => x.multimodal - y.multimodal);
+    for (const engine of engines) {
         if ((egn ? engine.client === egn : true) && clients[engine.client]) {
             const extra = {};
             if (engine.client === OPENAI) {
@@ -1171,9 +1185,10 @@ const analyzeSessions = async (sessionIds, options) => {
 };
 
 const PREFERRED_ENGINES = [
-    { client: OPENAI, func: promptChatGPT },
-    { client: GEMINI, func: promptGemini },
-    { client: CLAUDE, func: promptClaude },
+    { client: OPENAI, func: promptChatGPT, multimodal: 1 },
+    { client: GEMINI, func: promptGemini, multimodal: 0 },
+    { client: CLAUDE, func: promptClaude, multimodal: 2 },
+    { client: OLLAMA, func: promptOllama, multimodal: 99 },
 ]; // keep gpt first to avoid gemini grounding by default
 
 export default init;

package/lib/manifest.mjs

@@ -1,7 +1,7 @@
 const manifest = {
     "name": "utilitas",
     "description": "Just another common utility for JavaScript.",
-    "version": "1998.2.7",
+    "version": "1998.2.9",
     "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": "1998.2.7",
+    "version": "1998.2.9",
     "private": false,
     "homepage": "https://github.com/Leask/utilitas",
     "main": "index.mjs",