ra-h-os/src/components/focus/source/ContentDetector.ts

/**
 * Content type detection for Source Reader
 * Analyzes raw content to determine the best formatting approach
 */

export type ContentType = 'transcript' | 'book' | 'markdown' | 'article' | 'raw';

// Timestamp patterns for transcript detection
const TIMESTAMP_PATTERNS = [
  /^\[?\d{1,2}:\d{2}(:\d{2})?\]?\s*/,   // [00:00] or [00:00:00]
  /^\d{1,2}:\d{2}(:\d{2})?\s*[-–—]\s*/,  // 00:00 - or 00:00:00 -
  /^\(\d{1,2}:\d{2}(:\d{2})?\)\s*/,      // (00:00) or (00:00:00)
  /^\[\d+(\.\d+)?s\]\s*/,                // [0.1s] or [12.5s] - decimal seconds
];

// Simple book detection - look for explicit chapter markers
// We no longer try to parse numbered sections (too unreliable)
const BOOK_INDICATORS = [
  /\b(CHAPTER|Chapter)\s+\d+/i,
  /\b(Part|PART)\s+(I|II|III|IV|V|VI|VII|VIII|IX|X|\d+)\b/i,
  /\bTable\s+of\s+Contents\b/i,
  /\bPREFACE\b/,
  /\bINTRODUCTION\b/,
  /\bAPPENDIX\b/,
];

// Markdown syntax markers
const MARKDOWN_MARKERS = [
  /^#{1,6}\s/m,           // Headers
  /\*\*[^*]+\*\*/,        // Bold
  /\*[^*]+\*/,            // Italic (but not bold)
  /^[-*+]\s/m,            // Unordered lists
  /^\d+\.\s/m,            // Ordered lists
  /```[\s\S]*?```/,       // Code blocks
  /`[^`]+`/,              // Inline code
  /\[.+\]\(.+\)/,         // Links
];

/**
 * Detect the content type of raw source content
 */
export function detectContentType(content: string): ContentType {
  if (!content || content.length < 50) return 'raw';

  const lines = content.split('\n').slice(0, 50); // Check first 50 lines

  // 1. Check for transcript patterns (timestamps)
  const timestampLines = lines.filter(line =>
    TIMESTAMP_PATTERNS.some(pattern => pattern.test(line.trim()))
  );
  if (timestampLines.length >= 3) return 'transcript';

  // 2. Check for book indicators (explicit markers like "Chapter", "Table of Contents")
  const bookIndicatorCount = BOOK_INDICATORS.filter(pattern => pattern.test(content)).length;
  if (bookIndicatorCount >= 2) return 'book';
  
  // Also consider very long content as book-like (>20K chars)
  if (content.length > 20000 && bookIndicatorCount >= 1) return 'book';

  // 3. Check for markdown patterns
  const markdownScore = calculateMarkdownScore(content);
  if (markdownScore > 0.3) return 'markdown';

  // 4. Check for article structure (paragraphs with clear breaks)
  const paragraphs = content.split(/\n\n+/).filter(p => p.trim().length > 100);
  if (paragraphs.length >= 3) return 'article';

  // 5. Default to raw
  return 'raw';
}

/**
 * Calculate a "markdown-ness" score for content
 * Returns a value between 0 and 1
 */
function calculateMarkdownScore(content: string): number {
  const matchCount = MARKDOWN_MARKERS.filter(m => m.test(content)).length;
  return matchCount / MARKDOWN_MARKERS.length;
}

/**
 * Get a human-readable label for content type
 */
export function getContentTypeLabel(type: ContentType): string {
  switch (type) {
    case 'transcript': return 'Transcript';
    case 'book': return 'Book/Chapter';
    case 'markdown': return 'Markdown';
    case 'article': return 'Article';
    case 'raw': return 'Plain Text';
  }
}