@tpitre/story-ui 4.5.2 → 4.6.0

package/dist/story-generator/validateStory.js

@@ -334,8 +334,136 @@ function fixUnexpectedTokens(code) {
     for (const fix of fixes) {
         fixedCode = fixedCode.replace(fix.pattern, fix.replacement);
     }
+    // CRITICAL: Fix unescaped apostrophes in title strings
+    // LLMs often generate titles like: title: 'Women's Athletic' instead of 'Women\'s Athletic'
+    fixedCode = fixUnescapedApostrophesInTitles(fixedCode);
     return fixedCode;
 }
+/**
+ * Fixes unescaped apostrophes in title strings
+ * LLMs often generate: title: 'Women's Athletic Dashboard'
+ * Which should be: title: 'Women\'s Athletic Dashboard'
+ *
+ * Also handles cases where some apostrophes are escaped and others aren't:
+ * title: 'Women\'s Athletic Dashboard's Athletic Dashboard'
+ */
+function fixUnescapedApostrophesInTitles(code) {
+    const lines = code.split('\n');
+    for (let i = 0; i < lines.length; i++) {
+        const line = lines[i];
+        // Skip if no title property
+        if (!line.includes('title:'))
+            continue;
+        // Find title: followed by a quote character
+        const titleStartMatch = line.match(/title:\s*(['"])/);
+        if (!titleStartMatch)
+            continue;
+        const quoteChar = titleStartMatch[1];
+        const titleKeywordIndex = line.indexOf('title:');
+        const openQuoteIndex = line.indexOf(quoteChar, titleKeywordIndex);
+        // Find the closing quote by looking for quote followed by comma or brace at end of line
+        // The pattern is: quote + optional whitespace + (comma or brace) + optional whitespace + end
+        const endPattern = new RegExp(`${quoteChar === "'" ? "'" : '"'}\\s*[,}]\\s*$`);
+        const endMatch = line.match(endPattern);
+        if (!endMatch)
+            continue;
+        const closeQuoteIndex = line.lastIndexOf(quoteChar);
+        if (closeQuoteIndex <= openQuoteIndex)
+            continue;
+        // Extract content between opening and closing quotes
+        const content = line.substring(openQuoteIndex + 1, closeQuoteIndex);
+        // Process the content character by character to escape unescaped quotes
+        let fixedContent = '';
+        for (let j = 0; j < content.length; j++) {
+            const char = content[j];
+            const prevChar = j > 0 ? content[j - 1] : '';
+            if (char === quoteChar && prevChar !== '\\') {
+                // This is an unescaped quote inside the string - escape it
+                fixedContent += '\\' + char;
+            }
+            else {
+                fixedContent += char;
+            }
+        }
+        // Only modify if we actually made changes
+        if (fixedContent !== content) {
+            const beforeQuote = line.substring(0, openQuoteIndex + 1);
+            const afterQuote = line.substring(closeQuoteIndex);
+            lines[i] = beforeQuote + fixedContent + afterQuote;
+        }
+    }
+    let fixedCode = lines.join('\n');
+    // Fix duplicated title segments (LLM sometimes repeats parts of the title)
+    // e.g., "Women's Athletic Dashboard's Athletic Dashboard" -> "Women's Athletic Dashboard"
+    fixedCode = fixDuplicatedTitleSegments(fixedCode);
+    return fixedCode;
+}
+/**
+ * Fixes duplicated segments in titles
+ * LLMs sometimes generate: "Women's Athletic Dashboard's Athletic Dashboard"
+ * Which should be: "Women's Athletic Dashboard"
+ */
+function fixDuplicatedTitleSegments(code) {
+    const lines = code.split('\n');
+    for (let i = 0; i < lines.length; i++) {
+        const line = lines[i];
+        if (!line.includes('title:'))
+            continue;
+        // Find title: followed by a quote
+        const titleMatch = line.match(/title:\s*(['"])/);
+        if (!titleMatch)
+            continue;
+        const quoteChar = titleMatch[1];
+        const openQuoteIndex = line.indexOf(quoteChar, line.indexOf('title:'));
+        const closeQuoteIndex = line.lastIndexOf(quoteChar);
+        if (closeQuoteIndex <= openQuoteIndex)
+            continue;
+        // Extract title content (handling escaped quotes)
+        const content = line.substring(openQuoteIndex + 1, closeQuoteIndex);
+        // Normalize escaped quotes for word processing
+        const normalizedContent = content.replace(/\\'/g, "'");
+        // Split into words
+        const words = normalizedContent.split(/\s+/);
+        // Look for repeated word sequences
+        let modified = false;
+        let newWords = [...words];
+        for (let windowSize = 2; windowSize <= Math.floor(words.length / 2); windowSize++) {
+            for (let j = 0; j <= newWords.length - windowSize * 2; j++) {
+                const segment1 = newWords.slice(j, j + windowSize).join(' ');
+                const segment2 = newWords.slice(j + windowSize, j + windowSize * 2).join(' ');
+                if (segment1 === segment2 && segment1.length > 3) {
+                    newWords = [...newWords.slice(0, j + windowSize), ...newWords.slice(j + windowSize * 2)];
+                    modified = true;
+                    break;
+                }
+            }
+            if (modified)
+                break;
+        }
+        // Also look for possessive duplications: "Dashboard's Athletic" ... "Dashboard"
+        // Pattern: something ending with 's + words + that same word
+        if (!modified) {
+            const possessivePattern = /^(.+?)(['\\]?s\s+)(.+?)\s+\1$/i;
+            const match = newWords.join(' ').match(possessivePattern);
+            if (match) {
+                newWords = (match[1] + match[2] + match[3]).split(/\s+/);
+                modified = true;
+            }
+        }
+        if (modified) {
+            // Re-escape apostrophes for the output
+            let newContent = newWords.join(' ');
+            if (quoteChar === "'") {
+                // Re-escape internal single quotes
+                newContent = newContent.replace(/(?<!\\)'/g, "\\'");
+            }
+            const beforeQuote = line.substring(0, openQuoteIndex + 1);
+            const afterQuote = line.substring(closeQuoteIndex);
+            lines[i] = beforeQuote + newContent + afterQuote;
+        }
+    }
+    return lines.join('\n');
+}
 /**
  * Fixes unterminated string literals
  */