From 9b6be5214cedc9a08cf4d488fe67db39803afc50 Mon Sep 17 00:00:00 2001
From: Frederic Thomas <frthomas@adaptavist.com>
Date: Wed, 14 Jan 2026 13:57:41 +0000
Subject: [PATCH] Fix TTS cleanText function to properly extract text from HTML

The cleanText() function was using broken regex patterns that resulted in
mangled HTML fragments being returned instead of clean text. This caused
TTS to only speak the last segment of responses.

Changes:
- Replace broken regex HTML parsing with DOMParser (browser native)
- Properly extract text content while preserving all readable text
- Replace <pre> and <code> tags with placeholder before text extraction
- Add fallback for edge cases where DOMParser might fail

Before: cleanText output was '</pre> distribution in version.</p>'
After: cleanText output is 'Linux Distribution...See code attached...version.'

Fixes #901

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
---
 webui/components/chat/speech/speech-store.js | 70 ++++++--------------
 1 file changed, 22 insertions(+), 48 deletions(-)

diff --git a/webui/components/chat/speech/speech-store.js b/webui/components/chat/speech/speech-store.js
index 9ca202e6c..74f4f115d 100644
--- a/webui/components/chat/speech/speech-store.js
+++ b/webui/components/chat/speech/speech-store.js
@@ -499,59 +499,33 @@ const model = {
     const codePlaceholder = SUB + "code" + SUB;
     const tablePlaceholder = SUB + "table" + SUB;
 
-    // Helper function to handle both closed and unclosed patterns
-    // replacement can be a string or null (to remove)
-    function handlePatterns(
-      inputText,
-      closedPattern,
-      unclosedPattern,
-      replacement
-    ) {
-      // Process closed patterns first
-      let processed = inputText.replace(closedPattern, replacement || "");
-
-      // If the text changed, it means we found and replaced closed patterns
-      if (processed !== inputText) {
-        return processed;
-      } else {
-        // No closed patterns found, check for unclosed ones
-        const unclosedMatch = inputText.match(unclosedPattern);
-        if (unclosedMatch) {
-          // Replace the unclosed pattern
-          return inputText.replace(unclosedPattern, replacement || "");
-        }
-      }
-
-      // No patterns found, return original
-      return inputText;
-    }
-
-    // Handle code blocks
-    text = handlePatterns(
-      text,
-      /```(?:[a-zA-Z0-9]*\n)?[\s\S]*?```/g, // closed code blocks
-      /```(?:[a-zA-Z0-9]*\n)?[\s\S]*$/g, // unclosed code blocks
-      codePlaceholder
-    );
+    // Handle code blocks BEFORE HTML parsing (markdown code blocks)
+    text = text.replace(/```(?:[a-zA-Z0-9]*\n)?[\s\S]*?```/g, codePlaceholder); // closed code blocks
+    text = text.replace(/```(?:[a-zA-Z0-9]*\n)?[\s\S]*$/g, codePlaceholder); // unclosed code blocks
 
     // Replace inline code ticks with content preserved
     text = text.replace(/`([^`]*)`/g, "$1"); // remove backticks but keep content
 
-    // Handle HTML tags
-    text = handlePatterns(
-      text,
-      /<[a-zA-Z][a-zA-Z0-9]*>.*?<\/[a-zA-Z][a-zA-Z0-9]*>/gs, // closed HTML tags
-      /<[a-zA-Z][a-zA-Z0-9]*>[\s\S]*$/g, // unclosed HTML tags
-      "" // remove HTML tags completely
-    );
+    // Parse HTML using browser's DOMParser to properly extract text content
+    try {
+      const parser = new DOMParser();
+      // Wrap in a div to handle fragments
+      const doc = parser.parseFromString(`<div>${text}</div>`, 'text/html');
 
-    // Handle self-closing HTML tags
-    text = handlePatterns(
-      text,
-      /<[a-zA-Z][a-zA-Z0-9]*(\/| [^>]*\/>)/g, // complete self-closing tags
-      /<[a-zA-Z][a-zA-Z0-9]* [^>]*$/g, // incomplete self-closing tags
-      ""
-    );
+      // Replace <pre> and <code> tags with placeholder before extracting text
+      doc.querySelectorAll('pre, code').forEach(el => {
+        el.textContent = codePlaceholder;
+      });
+
+      // Extract text content (this strips all HTML tags properly)
+      text = doc.body.textContent || "";
+    } catch (e) {
+      // Fallback: simple tag stripping if DOMParser fails
+      console.warn("[Speech Store] DOMParser failed, using fallback:", e);
+      text = text.replace(/<pre[^>]*>[\s\S]*?<\/pre>/gi, codePlaceholder);
+      text = text.replace(/<code[^>]*>[\s\S]*?<\/code>/gi, codePlaceholder);
+      text = text.replace(/<[^>]+>/g, ''); // strip remaining tags
+    }
 
     // Remove markdown links: [label](url) → label
     text = text.replace(/\[([^\]]+)\]\([^\)]+\)/g, "$1");