refactor: Enhance markdown processing for text-to-speech conversion

- Introduced new regex patterns to remove thinking and reasoning blocks from markdown input.
- Added functionality to strip emojis from the text, improving clarity for TTS.
- Implemented HTML entity decoding to ensure proper text representation.
- Replaced the existing sanitization method with a more comprehensive markdown-to-text conversion approach, enhancing performance and maintainability.
This commit is contained in:
cogwheel0
2025-10-20 23:53:07 +05:30
parent 6ea7b3231c
commit e05a560966
2 changed files with 49 additions and 48 deletions

View File

@@ -11,6 +11,7 @@ import '../../../core/utils/reasoning_parser.dart';
import '../../../core/utils/message_segments.dart';
import '../../../core/utils/tool_calls_parser.dart';
import '../../../core/models/chat_message.dart';
import '../../../core/utils/markdown_to_text.dart';
import '../providers/text_to_speech_provider.dart';
import 'enhanced_image_attachment.dart';
import 'package:conduit/l10n/app_localizations.dart';
@@ -23,21 +24,6 @@ import '../../../core/utils/debug_logger.dart';
import 'sources/openwebui_sources.dart';
import '../providers/assistant_response_builder_provider.dart';
// Pre-compiled regex patterns for TTS sanitization (performance optimization)
final _ttsCodeBlockPattern = RegExp(r'```');
final _ttsInlineCodePattern = RegExp(r'`');
final _ttsImagePattern = RegExp(r'!\[(.*?)\]\((.*?)\)');
final _ttsLinkPattern = RegExp(r'\[(.*?)\]\((.*?)\)');
final _ttsBoldPattern1 = RegExp(r'\*\*');
final _ttsBoldPattern2 = RegExp(r'__');
final _ttsItalicPattern1 = RegExp(r'\*');
final _ttsItalicPattern2 = RegExp(r'_');
final _ttsStrikePattern = RegExp(r'~');
final _ttsListPattern = RegExp(r'^[-*+]\s+', multiLine: true);
final _ttsQuotePattern = RegExp(r'^>\s?', multiLine: true);
final _ttsMultiSpacePattern = RegExp(r'[ \t]{2,}');
final _ttsMultiNewlinePattern = RegExp(r'\n{3,}');
// Pre-compiled regex patterns for image processing (performance optimization)
final _base64ImagePattern = RegExp(r'data:image/[^;]+;base64,[A-Za-z0-9+/]+=*');
final _fileIdPattern = RegExp(r'/api/v1/files/([^/]+)/content');
@@ -258,7 +244,7 @@ class _AssistantMessageWidgetState extends ConsumerState<AssistantMessageWidget>
String _buildTtsPlainText(List<MessageSegment> segments, String fallback) {
if (segments.isEmpty) {
return _sanitizeForSpeech(fallback);
return MarkdownToText.convert(fallback);
}
final buffer = StringBuffer();
@@ -267,7 +253,7 @@ class _AssistantMessageWidgetState extends ConsumerState<AssistantMessageWidget>
continue;
}
final text = segment.text ?? '';
final sanitized = _sanitizeForSpeech(text);
final sanitized = MarkdownToText.convert(text);
if (sanitized.isEmpty) {
continue;
}
@@ -280,38 +266,11 @@ class _AssistantMessageWidgetState extends ConsumerState<AssistantMessageWidget>
final result = buffer.toString().trim();
if (result.isEmpty) {
return _sanitizeForSpeech(fallback);
return MarkdownToText.convert(fallback);
}
return result;
}
String _sanitizeForSpeech(String input) {
if (input.isEmpty) {
return '';
}
var text = input;
// Use pre-compiled regex patterns for better performance
text = text.replaceAll(_ttsCodeBlockPattern, ' ');
text = text.replaceAll(_ttsInlineCodePattern, '');
text = text.replaceAll(_ttsImagePattern, r'$1');
text = text.replaceAll(_ttsLinkPattern, r'$1');
text = text.replaceAll(_ttsBoldPattern1, '');
text = text.replaceAll(_ttsBoldPattern2, '');
text = text.replaceAll(_ttsItalicPattern1, '');
text = text.replaceAll(_ttsItalicPattern2, '');
text = text.replaceAll(_ttsStrikePattern, '');
text = text.replaceAll(_ttsListPattern, '');
text = text.replaceAll(_ttsQuotePattern, '');
text = text.replaceAll('&nbsp;', ' ');
text = text.replaceAll('&amp;', '&');
text = text.replaceAll('&lt;', '<');
text = text.replaceAll('&gt;', '>');
text = text.replaceAll(_ttsMultiSpacePattern, ' ');
text = text.replaceAll(_ttsMultiNewlinePattern, '\n\n');
return text.trim();
}
// No streaming-specific markdown fixes needed here; handled by Markdown widget
Widget _buildToolCallTile(ToolCallEntry tc) {