Fix Claude.ai parsing (#102)

* Update ClaudeExtractor to use 'font-claude-response' class and normalize content extraction

* Add Claude.ai pattern to ExtractorRegistry and fix document reference in ConversationExtractor

Author: imadreamerboy

src/extractor-registry.ts

@@ -68,6 +68,7 @@ export class ExtractorRegistry {
 
 		this.register({
 			patterns: [
+				'claude.ai',
 				/^https?:\/\/claude\.ai\/(chat|share)\/.*/
 			],
 			extractor: ClaudeExtractor

src/extractors/_conversation.ts

@@ -16,7 +16,7 @@ export abstract class ConversationExtractor extends BaseExtractor {
 		const rawContentHtml = this.createContentHtml(messages, footnotes);
 
 		// Create a temporary document to run Defuddle on our content
-		const tempDoc = document.implementation.createHTMLDocument();
+		const tempDoc = this.document.implementation.createHTMLDocument();
 		const container = tempDoc.createElement('article');
 		container.innerHTML = rawContentHtml;
 		tempDoc.body.appendChild(container);

src/extractors/claude.ts

@@ -7,7 +7,7 @@ export class ClaudeExtractor extends ConversationExtractor {
 	constructor(document: Document, url: string) {
 		super(document, url);
 		// Find all message blocks - both user and assistant messages
-		this.articles = document.querySelectorAll('div[data-testid="user-message"], div[data-testid="assistant-message"], div.font-claude-message');
+		this.articles = document.querySelectorAll('div[data-testid="user-message"], div[data-testid="assistant-message"], div.font-claude-response');
 	}
 
 	canExtract(): boolean {
@@ -33,16 +33,19 @@ export class ClaudeExtractor extends ConversationExtractor {
 				else {
 					return;
 				}
-			} else if (article.classList.contains('font-claude-message')) {
+			} else if (article.classList.contains('font-claude-response')) {
 				// Handle Claude messages
 				role = 'assistant';
-				content = article.innerHTML;
+				const assistantBody = (article.querySelector('.standard-markdown') as HTMLElement) || (article as HTMLElement);
+				content = assistantBody.innerHTML;
 			} else {
 				// Skip unknown elements
 				return;
 			}
 
 			if (content) {
+				// Normalize content similar to ChatGPT extractor
+				content = content.replace(/\u200B/g, '').replace(/<p[^>]*>\s*<\/p>/g, '');
 				messages.push({
 					author: role === 'you' ? 'You' : 'Claude',
 					content: content.trim(),