@adminforth/agent 1.42.1 → 1.43.0

package/build.log

@@ -58,5 +58,5 @@ custom/speech_recognition_frontend/voiceActivityDetection.ts
 custom/speech_recognition_frontend/types/
 custom/speech_recognition_frontend/types/voice-activity-detection.d.ts
 
-sent 1,661,051 bytes  received 860 bytes  3,323,822.00 bytes/sec
-total size is 1,657,135  speedup is 1.00
+sent 1,661,103 bytes  received 860 bytes  3,323,926.00 bytes/sec
+total size is 1,657,187  speedup is 1.00

package/custom/speech_recognition_frontend/MicrophoneButon.vue

@@ -1,7 +1,7 @@
 <template>
   <div 
     class="absolute bottom-2 flex items-center justify-center z-10 gap-4"
-    :class="[!agentStore.isAudioChatMode ? 'right-16': 'right-1/2 translate-x-1/2']"
+    :class="[!agentStore.isAudioChatMode ? 'right-16': agentStore.isMobile ? 'right-1/2 translate-x-1/2' : 'right-1/2 translate-x-3/4']"
   > 
     <button 
       v-if="isAudioChatMode && microphoneButtonMode === 'generating'"

package/dist/custom/speech_recognition_frontend/MicrophoneButon.vue

@@ -1,7 +1,7 @@
 <template>
   <div 
     class="absolute bottom-2 flex items-center justify-center z-10 gap-4"
-    :class="[!agentStore.isAudioChatMode ? 'right-16': 'right-1/2 translate-x-1/2']"
+    :class="[!agentStore.isAudioChatMode ? 'right-16': agentStore.isMobile ? 'right-1/2 translate-x-1/2' : 'right-1/2 translate-x-3/4']"
   > 
     <button 
       v-if="isAudioChatMode && microphoneButtonMode === 'generating'"

package/dist/index.js

@@ -26,6 +26,7 @@ import { detectUserLanguage } from "./agent/languageDetect.js";
 import { prepareApiBasedTools as buildApiBasedTools } from './apiBasedTools.js';
 import { createAgentEventStream } from "./agentResponseEvents.js";
 import { appendCustomSystemPrompt, buildAgentSystemPrompt, buildAgentTurnSystemPrompt, DEFAULT_AGENT_SYSTEM_PROMPT } from "./agent/systemPrompt.js";
+import { sanitizeSpeechText } from "./sanitizeSpeechText.js";
 const agentResponseBodySchema = z.object({
     message: z.string(),
     sessionId: z.string(),
@@ -474,7 +475,7 @@ export default class AdminForthAgentPlugin extends AdminForthPlugin {
                         text: agentResponse.text,
                     }, sessionId, agentResponse.turnId);
                     const speech = yield audioAdapter.synthesize({
-                        text: agentResponse.text,
+                        text: sanitizeSpeechText(agentResponse.text),
                         stream: true,
                         streamFormat: "audio",
                         format: "mp3",

package/dist/sanitizeSpeechText.js

@@ -0,0 +1,31 @@
+const URL_RE = /\b(?:https?:\/\/|www\.)[^\s<>()]+/gi;
+const MARKDOWN_LINK_RE = /\[([^\]]+)]\(([^)]+)\)/g;
+const MARKDOWN_IMAGE_RE = /!\[([^\]]*)]\(([^)]+)\)/g;
+const HTML_TAG_RE = /<[^>]+>/g;
+const CODE_FENCE_RE = /```[\s\S]*?```/g;
+const INLINE_CODE_RE = /`[^`]*`/g;
+const BRACKETED_META_RE = /\[(?:ref|link|url|source|sources?|debug|meta|data):[^\]]*]/gi;
+const MD_HEADING_LINE_RE = /^\s{0,3}#{1,6}\s+/gm;
+const MD_BLOCKQUOTE_RE = /^\s{0,3}>\s?/gm;
+const MD_HR_RE = /^\s{0,3}(-{3,}|\*{3,}|_{3,})\s*$/gm;
+const MD_LIST_MARKER_RE = /^\s{0,3}(?:[-*+]|\d+\.)\s+/gm;
+const MD_EMPHASIS_RE = /(\*\*|__)(.*?)\1/g;
+const MD_ITALIC_RE = /(^|[^\\])(?:\*|_)([^\s].*?)(?:\*|_)/g;
+export function sanitizeSpeechText(input) {
+    return input
+        .replace(CODE_FENCE_RE, " ")
+        .replace(MARKDOWN_IMAGE_RE, "$1")
+        .replace(MARKDOWN_LINK_RE, "$1")
+        .replace(URL_RE, " ")
+        .replace(MD_HEADING_LINE_RE, "")
+        .replace(MD_BLOCKQUOTE_RE, "")
+        .replace(MD_HR_RE, " ")
+        .replace(MD_LIST_MARKER_RE, "")
+        .replace(MD_EMPHASIS_RE, "$2")
+        .replace(MD_ITALIC_RE, "$1$2")
+        .replace(HTML_TAG_RE, " ")
+        .replace(INLINE_CODE_RE, " ")
+        .replace(BRACKETED_META_RE, " ")
+        .replace(/\s+/g, " ")
+        .trim();
+}

package/index.ts

@@ -16,6 +16,7 @@ import { createAgentEventStream } from "./agentResponseEvents.js";
 import { appendCustomSystemPrompt, buildAgentSystemPrompt, buildAgentTurnSystemPrompt, DEFAULT_AGENT_SYSTEM_PROMPT} from "./agent/systemPrompt.js";
 import type { ToolCallEvent } from "./agent/toolCallEvents.js";
 import type { CurrentPageContext } from "./agent/tools/getUserLocation.js";
+import { sanitizeSpeechText } from "./sanitizeSpeechText.js";
 
 type MulterFile = {
   buffer: Buffer;
@@ -540,7 +541,7 @@ export default class AdminForthAgentPlugin extends AdminForthPlugin {
             agentResponse.turnId,
           );
           const speech = await audioAdapter.synthesize({
-            text: agentResponse.text,
+            text: sanitizeSpeechText(agentResponse.text),
             stream: true,
             streamFormat: "audio",
             format: "mp3",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@adminforth/agent",
-  "version": "1.42.1",
+  "version": "1.43.0",
   "main": "dist/index.js",
   "types": "dist/index.d.ts",
   "type": "module",

package/sanitizeSpeechText.ts

@@ -0,0 +1,32 @@
+const URL_RE = /\b(?:https?:\/\/|www\.)[^\s<>()]+/gi;
+const MARKDOWN_LINK_RE = /\[([^\]]+)]\(([^)]+)\)/g;
+const MARKDOWN_IMAGE_RE = /!\[([^\]]*)]\(([^)]+)\)/g;
+const HTML_TAG_RE = /<[^>]+>/g;
+const CODE_FENCE_RE = /```[\s\S]*?```/g;
+const INLINE_CODE_RE = /`[^`]*`/g;
+const BRACKETED_META_RE = /\[(?:ref|link|url|source|sources?|debug|meta|data):[^\]]*]/gi;
+const MD_HEADING_LINE_RE = /^\s{0,3}#{1,6}\s+/gm;
+const MD_BLOCKQUOTE_RE = /^\s{0,3}>\s?/gm;
+const MD_HR_RE = /^\s{0,3}(-{3,}|\*{3,}|_{3,})\s*$/gm;
+const MD_LIST_MARKER_RE = /^\s{0,3}(?:[-*+]|\d+\.)\s+/gm;
+const MD_EMPHASIS_RE = /(\*\*|__)(.*?)\1/g;
+const MD_ITALIC_RE = /(^|[^\\])(?:\*|_)([^\s].*?)(?:\*|_)/g;
+
+export function sanitizeSpeechText(input: string): string {
+  return input
+    .replace(CODE_FENCE_RE, " ")
+    .replace(MARKDOWN_IMAGE_RE, "$1")
+    .replace(MARKDOWN_LINK_RE, "$1")
+    .replace(URL_RE, " ")
+    .replace(MD_HEADING_LINE_RE, "")
+    .replace(MD_BLOCKQUOTE_RE, "")
+    .replace(MD_HR_RE, " ")
+    .replace(MD_LIST_MARKER_RE, "")
+    .replace(MD_EMPHASIS_RE, "$2")
+    .replace(MD_ITALIC_RE, "$1$2")
+    .replace(HTML_TAG_RE, " ")
+    .replace(INLINE_CODE_RE, " ")
+    .replace(BRACKETED_META_RE, " ")
+    .replace(/\s+/g, " ")
+    .trim();
+}