/**
 * Document Parsing Tools
 * Extracts text from PDF, EPUB, HTML, and images (OCR).
 * Adapted from Big RAG plugin with graceful fallback for missing dependencies.
 */

import fs from 'fs/promises';
import path from 'path';
import { z } from 'zod';

// ──────────────────────────────────────────────────────
// PDF Parser
// ──────────────────────────────────────────────────────

import { validatePath } from '../pathSecurity';

function safePath(filePath: string, workspaceRoot: string): string {
  return validatePath(workspaceRoot, workspaceRoot, filePath);
}

export async function parse_pdf(file_path: string, workspaceRoot: string = process.cwd()): Promise<{ success: boolean; text?: string; pages?: number; error?: string }> {
  file_path = safePath(file_path, workspaceRoot);
  try {
    const stat = await fs.stat(file_path);
    if (stat.size > 100 * 1024 * 1024) { // 100MB limit
      return { success: false, error: 'File exceeds 100MB size limit' };
    }

    const content = await fs.readFile(file_path);
    
    // Try pdf-parse first (if available)
    try {
      const pdfParseModule = await import('pdf-parse') as any;
      const pdfParse = pdfParseModule.default || pdfParseModule;
      if (typeof pdfParse === 'function') {
        const data = await pdfParse(content);
        return { success: true, text: data.text, pages: data.numpages };
      }
    } catch (parseErr) {
      // pdf-parse not available or failed - fall back to basic extraction
      console.warn('[DocumentParsers] pdf-parse unavailable, using fallback:', parseErr instanceof Error ? parseErr.message : String(parseErr));
    }

    // Fallback: try to extract text from PDF structure (limited but works without deps)
    const text = extractTextFromPdfFallback(content);
    
    if (text && text.length > 0) {
      return { success: true, text, pages: estimatePages(text) };
    }

    // If all else fails, try OCR on first page as image
    try {
      const ocrResult = await ocr_image(file_path, workspaceRoot);
      if (ocrResult.success && ocrResult.text) {
        return { success: true, text: ocrResult.text, pages: 1 };
      }
    } catch {}

    return { success: false, error: 'Could not extract text from PDF. Try installing pdf-parse or use OCR mode.' };
  } catch (error) {
    return { success: false, error: error instanceof Error ? error.message : String(error) };
  }
}

// Basic PDF text extraction without external deps
function extractTextFromPdfFallback(content: Buffer): string {
  const text = content.toString('utf-8', 0, Math.min(100000, content.length));
  
  // Look for stream data between "stream" and "endstream" markers
  const streamRegex = /stream\r?\n([\s\S]*?)\r?\nendstream/g;
  let matches: string[] = [];
  let match;
  
  while ((match = streamRegex.exec(text)) !== null) {
    matches.push(match[1]);
  }

  // Clean up extracted text (remove PDF operators, escape sequences)
  return matches
    .map(chunk => chunk
      .replace(/\\[nrt()<>[\]{}\/%]/g, ' ') // Remove PDF escapes
      .replace(/\r\n/g, '\n')
      .replace(/[^\x20-\x7E\n]/g, '') // Keep only printable ASCII + newlines
      .trim()
    )
    .filter(Boolean)
    .join('\n');
}

function estimatePages(text: string): number {
  const lines = text.split('\n').filter(l => l.trim().length > 0);
  return Math.max(1, Math.ceil(lines.length / 50)); // Rough estimate: ~50 lines per page
}

// ──────────────────────────────────────────────────────
// EPUB Parser
// ──────────────────────────────────────────────────────

export async function parse_epub(file_path: string, workspaceRoot: string = process.cwd()): Promise<{ success: boolean; title?: string; chapters?: Array<{ title: string; content: string }>; error?: string }> {
  file_path = safePath(file_path, workspaceRoot);
  try {
    const stat = await fs.stat(file_path);
    if (stat.size > 50 * 1024 * 1024) { // 50MB limit
      return { success: false, error: 'File exceeds 50MB size limit' };
    }

    const content = await fs.readFile(file_path);
    
    // Try epub2 first (if available)
    try {
      // @ts-ignore - optional dependency
      const epubModule = await import('epub2') as any;
      const EPub = epubModule.EPub || epubModule.default?.EPub;
      if (!EPub) throw new Error('EPUB module not exported');
      
      return new Promise((resolve, reject) => {
        const epub = new EPub(file_path);
        
        epub.on('error', (error: Error) => {
          console.error('[DocumentParsers] EPUB parsing error:', error.message);
          resolve({ success: false, error: `EPUB parse error: ${error.message}` });
        });

        const stripHtml = (input: string) => input.replace(/<[^>]*>/g, ' ');
        
        epub.load((err?: Error) => {
          if (err) {
            resolve({ success: false, error: `EPUB load error: ${err.message}` });
            return;
          }

          const manifest = (epub as any).manifest || {};
          const chapters: Array<{ title: string; content: string }> = [];
          
          // Get spine order
          const spine = (epub as any).spine || [];
          
          for (const chapterId of spine) {
            try {
              const entry = manifest[chapterId];
              if (!entry) continue;
              
              const href = entry.href || entry.path || '';
              const mediaType = entry['media-type'] || entry.mediaType || 'text/html';
              
              // Read chapter content
              const chapterContent = epub.getChapterSync?.(href);
              if (chapterContent) {
                chapters.push({
                  title: stripHtml(chapterContent.title || '').trim() || `Chapter ${chapters.length + 1}`,
                  content: stripHtml(chapterContent.content || '')
                });
              }
            } catch {}
          }

          const title = (epub as any).metadata?.title || path.basename(file_path, '.epub');
          
          resolve({ 
            success: true, 
            title, 
            chapters: chapters.length > 0 ? chapters : undefined 
          });
        });
      });
    } catch (parseErr) {
      console.warn('[DocumentParsers] epub2 unavailable:', parseErr instanceof Error ? parseErr.message : String(parseErr));
    }

    // Fallback: try to extract from ZIP structure (EPUB is a ZIP file)
    const text = extractTextFromEpubFallback(content);
    
    if (text && text.length > 0) {
      return { success: true, title: path.basename(file_path, '.epub'), chapters: [{ title: 'Content', content: text }] };
    }

    return { success: false, error: 'Could not parse EPUB. Try installing epub2 or use a different format.' };
  } catch (error) {
    return { success: false, error: error instanceof Error ? error.message : String(error) };
  }
}

function extractTextFromEpubFallback(content: Buffer): string {
  // EPUB is ZIP - try to find text files in it
  const text = content.toString('utf-8', 0, Math.min(50000, content.length));
  
  // Look for HTML/XML content between tags
  const htmlRegex = /<body[^>]*>([\s\S]*?)<\/body>/gi;
  let matches: string[] = [];
  let match;
  
  while ((match = htmlRegex.exec(text)) !== null) {
    matches.push(match[1]);
  }

  // Strip HTML tags and clean up
  return matches
    .map(chunk => chunk.replace(/<[^>]*>/g, ' ').trim())
    .filter(Boolean)
    .join('\n\n');
}

// ──────────────────────────────────────────────────────
// HTML Parser
// ──────────────────────────────────────────────────────

export async function parse_html(file_path: string, options?: { extract_links?: boolean }, workspaceRoot: string = process.cwd()): Promise<{ success: boolean; title?: string; body_text?: string; links?: Array<{ href: string; text: string }>; error?: string }> {
  file_path = safePath(file_path, workspaceRoot);
  try {
    const content = await fs.readFile(file_path, 'utf-8');
    
    // Try cheerio first (if available)
    try {
      // @ts-ignore - optional dependency
      const cheerioModule = await import('cheerio') as any;
      const cheerio = cheerioModule.default || cheerioModule;
      if (!cheerio?.load) throw new Error('Cheerio not exported');
      
      const $ = cheerio.load(content);
      
      $('script, style, noscript').remove();
      
      const title = $('title').text().trim() || path.basename(file_path, '.html') || path.basename(file_path, '.htm');
      const bodyText = $('body').text().replace(/\s+/g, ' ').trim();
      
      // Extract links
      const links: Array<{ href: string; text: string }> = [];
      $('a[href]').each((_i: number, el: any) => {
        const href = $(el).attr('href') || '';
        const text = $(el).text().trim();
        if (href && text) {
          links.push({ href, text });
        }
      });

      return { success: true, title, body_text: bodyText, links: links.length > 0 ? links : undefined };
    } catch (parseErr) {
      console.warn('[DocumentParsers] cheerio unavailable:', parseErr instanceof Error ? parseErr.message : String(parseErr));
    }

    // Fallback: basic regex-based extraction
    const titleMatch = content.match(/<title[^>]*>([\s\S]*?)<\/title>/i);
    const bodyMatch = content.match(/<body[^>]*>([\s\S]*?)<\/body>/i);
    
    return {
      success: true,
      title: titleMatch?.[1]?.trim() || path.basename(file_path),
      body_text: bodyMatch ? stripHtml(bodyMatch[1]) : '',
      links: extractLinksFallback(content)
    };
  } catch (error) {
    return { success: false, error: error instanceof Error ? error.message : String(error) };
  }
}

function stripHtml(html: string): string {
  return html.replace(/<[^>]*>/g, ' ').replace(/\s+/g, ' ').trim();
}

function extractLinksFallback(content: string): Array<{ href: string; text: string }> {
  const links: Array<{ href: string; text: string }> = [];
  const linkRegex = /<a\s+href=["']([^"']+)["'][^>]*>([\s\S]*?)<\/a>/gi;
  let match;
  
  while ((match = linkRegex.exec(content)) !== null) {
    links.push({ href: match[1], text: stripHtml(match[2]) });
  }
  
  return links;
}

// ──────────────────────────────────────────────────────
// Image OCR Parser
// ──────────────────────────────────────────────────────

export async function ocr_image(file_path: string, workspaceRoot: string = process.cwd()): Promise<{ success: boolean; text?: string; confidence?: number; error?: string }> {
  file_path = safePath(file_path, workspaceRoot);
  try {
    const stat = await fs.stat(file_path);
    if (stat.size > 20 * 1024 * 1024) { // 20MB limit
      return { success: false, error: 'Image exceeds 20MB size limit' };
    }

    // Try tesseract.js first (if available and enabled)
    try {
      // @ts-ignore - optional dependency
      const tesseractModule = await import('tesseract.js') as any;
      const createWorker = tesseractModule.createWorker || tesseractModule.default?.createWorker;
      if (!createWorker) throw new Error('Tesseract not exported');
      
      const worker = await createWorker('eng');
      
      const { data: { text, confidence } } = await worker.recognize(file_path);
      await worker.terminate();
      
      return { 
        success: true, 
        text: text.replace(/\s+/g, ' ').trim(),
        confidence 
      };
    } catch (ocrErr) {
      console.warn('[DocumentParsers] OCR unavailable:', ocrErr instanceof Error ? ocrErr.message : String(ocrErr));
    }

    return { success: false, error: 'OCR requires Tesseract.js and Tesseract OCR engine installed. Enable in plugin settings.' };
  } catch (error) {
    return { success: false, error: error instanceof Error ? error.message : String(error) };
  }
}

// ──────────────────────────────────────────────────────
// Zod schemas for tool parameters
// ──────────────────────────────────────────────────────

export const parsePdfSchema = z.object({
  file_path: z.string().min(1).describe('Path to the PDF file'),
});

export const parseEpubSchema = z.object({
  file_path: z.string().min(1).describe('Path to the EPUB file'),
});

export const parseHtmlSchema = z.object({
  file_path: z.string().min(1).describe('Path to the HTML file'),
  options: z.object({
    extract_links: z.boolean().optional().default(true),
  }).optional(),
});

export const ocrImageSchema = z.object({
  file_path: z.string().min(1).describe('Path to the image file (PNG, JPG, BMP)'),
});


// ──────────────────────────────────────────────────────
// Text Chunker for RAG
// ──────────────────────────────────────────────────────

export async function chunk_text(text: string, options?: { 
  chunk_size?: number; 
  overlap?: number; 
}): Promise<{ success: boolean; chunks: Array<{ index: number; content: string }>; total_chunks: number; error?: string }> {
  try {
    const chunkSize = options?.chunk_size ?? 500; // chars per chunk
    const overlap = options?.overlap ?? 50; // chars of overlap between chunks
    
    if (!text || text.length === 0) {
      return { success: false, chunks: [], total_chunks: 0, error: 'Input text is empty' };
    }
    
    const chunks: Array<{ index: number; content: string }> = [];
    let start = 0;
    let index = 0;
    
    while (start < text.length) {
      const end = Math.min(start + chunkSize, text.length);
      let chunkEnd = end;
      
      // Try to break at sentence boundary if possible
      if (end < text.length) {
        const lastPeriod = text.lastIndexOf('.', end - 1);
        const lastNewline = text.lastIndexOf('\n', end - 1);
        const lastSpace = text.lastIndexOf(' ', end - 1);
        
        // Prefer sentence boundary, then newline, then space
        let breakPoint = Math.max(lastPeriod, lastNewline, lastSpace);
        if (breakPoint > start + chunkSize * 0.5) {
          chunkEnd = breakPoint + 1; // Include the period/newline/space
        } else {
          chunkEnd = end;
        }
      }
      
      const chunkContent = text.slice(start, chunkEnd).trim();
      if (chunkContent.length > 0) {
        chunks.push({ index: index++, content: chunkContent });
      }
      
      start = chunkEnd - overlap;
      if (start <= 0) start = 0; // Prevent infinite loop on small text
    }
    
    return { success: true, chunks, total_chunks: chunks.length };
  } catch (error) {
    return { success: false, chunks: [], total_chunks: 0, error: error instanceof Error ? error.message : String(error) };
  }
}

/**
 * Document Parsing Tools
 * Extracts text from PDF, EPUB, HTML, and images (OCR).
 * Adapted from Big RAG plugin with graceful fallback for missing dependencies.
 */

import fs from 'fs/promises';
import path from 'path';
import { z } from 'zod';

// ──────────────────────────────────────────────────────
// PDF Parser
// ──────────────────────────────────────────────────────

import { validatePath } from '../pathSecurity';

function safePath(filePath: string, workspaceRoot: string): string {
  return validatePath(workspaceRoot, workspaceRoot, filePath);
}

export async function parse_pdf(file_path: string, workspaceRoot: string = process.cwd()): Promise<{ success: boolean; text?: string; pages?: number; error?: string }> {
  file_path = safePath(file_path, workspaceRoot);
  try {
    const stat = await fs.stat(file_path);
    if (stat.size > 100 * 1024 * 1024) { // 100MB limit
      return { success: false, error: 'File exceeds 100MB size limit' };
    }

    const content = await fs.readFile(file_path);
    
    // Try pdf-parse first (if available)
    try {
      const pdfParseModule = await import('pdf-parse') as any;
      const pdfParse = pdfParseModule.default || pdfParseModule;
      if (typeof pdfParse === 'function') {
        const data = await pdfParse(content);
        return { success: true, text: data.text, pages: data.numpages };
      }
    } catch (parseErr) {
      // pdf-parse not available or failed - fall back to basic extraction
      console.warn('[DocumentParsers] pdf-parse unavailable, using fallback:', parseErr instanceof Error ? parseErr.message : String(parseErr));
    }

    // Fallback: try to extract text from PDF structure (limited but works without deps)
    const text = extractTextFromPdfFallback(content);
    
    if (text && text.length > 0) {
      return { success: true, text, pages: estimatePages(text) };
    }

    // If all else fails, try OCR on first page as image
    try {
      const ocrResult = await ocr_image(file_path, workspaceRoot);
      if (ocrResult.success && ocrResult.text) {
        return { success: true, text: ocrResult.text, pages: 1 };
      }
    } catch {}

    return { success: false, error: 'Could not extract text from PDF. Try installing pdf-parse or use OCR mode.' };
  } catch (error) {
    return { success: false, error: error instanceof Error ? error.message : String(error) };
  }
}

// Basic PDF text extraction without external deps
function extractTextFromPdfFallback(content: Buffer): string {
  const text = content.toString('utf-8', 0, Math.min(100000, content.length));
  
  // Look for stream data between "stream" and "endstream" markers
  const streamRegex = /stream\r?\n([\s\S]*?)\r?\nendstream/g;
  let matches: string[] = [];
  let match;
  
  while ((match = streamRegex.exec(text)) !== null) {
    matches.push(match[1]);
  }

  // Clean up extracted text (remove PDF operators, escape sequences)
  return matches
    .map(chunk => chunk
      .replace(/\\[nrt()<>[\]{}\/%]/g, ' ') // Remove PDF escapes
      .replace(/\r\n/g, '\n')
      .replace(/[^\x20-\x7E\n]/g, '') // Keep only printable ASCII + newlines
      .trim()
    )
    .filter(Boolean)
    .join('\n');
}

function estimatePages(text: string): number {
  const lines = text.split('\n').filter(l => l.trim().length > 0);
  return Math.max(1, Math.ceil(lines.length / 50)); // Rough estimate: ~50 lines per page
}

// ──────────────────────────────────────────────────────
// EPUB Parser
// ──────────────────────────────────────────────────────

export async function parse_epub(file_path: string, workspaceRoot: string = process.cwd()): Promise<{ success: boolean; title?: string; chapters?: Array<{ title: string; content: string }>; error?: string }> {
  file_path = safePath(file_path, workspaceRoot);
  try {
    const stat = await fs.stat(file_path);
    if (stat.size > 50 * 1024 * 1024) { // 50MB limit
      return { success: false, error: 'File exceeds 50MB size limit' };
    }

    const content = await fs.readFile(file_path);
    
    // Try epub2 first (if available)
    try {
      // @ts-ignore - optional dependency
      const epubModule = await import('epub2') as any;
      const EPub = epubModule.EPub || epubModule.default?.EPub;
      if (!EPub) throw new Error('EPUB module not exported');
      
      return new Promise((resolve, reject) => {
        const epub = new EPub(file_path);
        
        epub.on('error', (error: Error) => {
          console.error('[DocumentParsers] EPUB parsing error:', error.message);
          resolve({ success: false, error: `EPUB parse error: ${error.message}` });
        });

        const stripHtml = (input: string) => input.replace(/<[^>]*>/g, ' ');
        
        epub.load((err?: Error) => {
          if (err) {
            resolve({ success: false, error: `EPUB load error: ${err.message}` });
            return;
          }

          const manifest = (epub as any).manifest || {};
          const chapters: Array<{ title: string; content: string }> = [];
          
          // Get spine order
          const spine = (epub as any).spine || [];
          
          for (const chapterId of spine) {
            try {
              const entry = manifest[chapterId];
              if (!entry) continue;
              
              const href = entry.href || entry.path || '';
              const mediaType = entry['media-type'] || entry.mediaType || 'text/html';
              
              // Read chapter content
              const chapterContent = epub.getChapterSync?.(href);
              if (chapterContent) {
                chapters.push({
                  title: stripHtml(chapterContent.title || '').trim() || `Chapter ${chapters.length + 1}`,
                  content: stripHtml(chapterContent.content || '')
                });
              }
            } catch {}
          }

          const title = (epub as any).metadata?.title || path.basename(file_path, '.epub');
          
          resolve({ 
            success: true, 
            title, 
            chapters: chapters.length > 0 ? chapters : undefined 
          });
        });
      });
    } catch (parseErr) {
      console.warn('[DocumentParsers] epub2 unavailable:', parseErr instanceof Error ? parseErr.message : String(parseErr));
    }

    // Fallback: try to extract from ZIP structure (EPUB is a ZIP file)
    const text = extractTextFromEpubFallback(content);
    
    if (text && text.length > 0) {
      return { success: true, title: path.basename(file_path, '.epub'), chapters: [{ title: 'Content', content: text }] };
    }

    return { success: false, error: 'Could not parse EPUB. Try installing epub2 or use a different format.' };
  } catch (error) {
    return { success: false, error: error instanceof Error ? error.message : String(error) };
  }
}

function extractTextFromEpubFallback(content: Buffer): string {
  // EPUB is ZIP - try to find text files in it
  const text = content.toString('utf-8', 0, Math.min(50000, content.length));
  
  // Look for HTML/XML content between tags
  const htmlRegex = /<body[^>]*>([\s\S]*?)<\/body>/gi;
  let matches: string[] = [];
  let match;
  
  while ((match = htmlRegex.exec(text)) !== null) {
    matches.push(match[1]);
  }

  // Strip HTML tags and clean up
  return matches
    .map(chunk => chunk.replace(/<[^>]*>/g, ' ').trim())
    .filter(Boolean)
    .join('\n\n');
}

// ──────────────────────────────────────────────────────
// HTML Parser
// ──────────────────────────────────────────────────────

export async function parse_html(file_path: string, options?: { extract_links?: boolean }, workspaceRoot: string = process.cwd()): Promise<{ success: boolean; title?: string; body_text?: string; links?: Array<{ href: string; text: string }>; error?: string }> {
  file_path = safePath(file_path, workspaceRoot);
  try {
    const content = await fs.readFile(file_path, 'utf-8');
    
    // Try cheerio first (if available)
    try {
      // @ts-ignore - optional dependency
      const cheerioModule = await import('cheerio') as any;
      const cheerio = cheerioModule.default || cheerioModule;
      if (!cheerio?.load) throw new Error('Cheerio not exported');
      
      const $ = cheerio.load(content);
      
      $('script, style, noscript').remove();
      
      const title = $('title').text().trim() || path.basename(file_path, '.html') || path.basename(file_path, '.htm');
      const bodyText = $('body').text().replace(/\s+/g, ' ').trim();
      
      // Extract links
      const links: Array<{ href: string; text: string }> = [];
      $('a[href]').each((_i: number, el: any) => {
        const href = $(el).attr('href') || '';
        const text = $(el).text().trim();
        if (href && text) {
          links.push({ href, text });
        }
      });

      return { success: true, title, body_text: bodyText, links: links.length > 0 ? links : undefined };
    } catch (parseErr) {
      console.warn('[DocumentParsers] cheerio unavailable:', parseErr instanceof Error ? parseErr.message : String(parseErr));
    }

    // Fallback: basic regex-based extraction
    const titleMatch = content.match(/<title[^>]*>([\s\S]*?)<\/title>/i);
    const bodyMatch = content.match(/<body[^>]*>([\s\S]*?)<\/body>/i);
    
    return {
      success: true,
      title: titleMatch?.[1]?.trim() || path.basename(file_path),
      body_text: bodyMatch ? stripHtml(bodyMatch[1]) : '',
      links: extractLinksFallback(content)
    };
  } catch (error) {
    return { success: false, error: error instanceof Error ? error.message : String(error) };
  }
}

function stripHtml(html: string): string {
  return html.replace(/<[^>]*>/g, ' ').replace(/\s+/g, ' ').trim();
}

function extractLinksFallback(content: string): Array<{ href: string; text: string }> {
  const links: Array<{ href: string; text: string }> = [];
  const linkRegex = /<a\s+href=["']([^"']+)["'][^>]*>([\s\S]*?)<\/a>/gi;
  let match;
  
  while ((match = linkRegex.exec(content)) !== null) {
    links.push({ href: match[1], text: stripHtml(match[2]) });
  }
  
  return links;
}

// ──────────────────────────────────────────────────────
// Image OCR Parser
// ──────────────────────────────────────────────────────

export async function ocr_image(file_path: string, workspaceRoot: string = process.cwd()): Promise<{ success: boolean; text?: string; confidence?: number; error?: string }> {
  file_path = safePath(file_path, workspaceRoot);
  try {
    const stat = await fs.stat(file_path);
    if (stat.size > 20 * 1024 * 1024) { // 20MB limit
      return { success: false, error: 'Image exceeds 20MB size limit' };
    }

    // Try tesseract.js first (if available and enabled)
    try {
      // @ts-ignore - optional dependency
      const tesseractModule = await import('tesseract.js') as any;
      const createWorker = tesseractModule.createWorker || tesseractModule.default?.createWorker;
      if (!createWorker) throw new Error('Tesseract not exported');
      
      const worker = await createWorker('eng');
      
      const { data: { text, confidence } } = await worker.recognize(file_path);
      await worker.terminate();
      
      return { 
        success: true, 
        text: text.replace(/\s+/g, ' ').trim(),
        confidence 
      };
    } catch (ocrErr) {
      console.warn('[DocumentParsers] OCR unavailable:', ocrErr instanceof Error ? ocrErr.message : String(ocrErr));
    }

    return { success: false, error: 'OCR requires Tesseract.js and Tesseract OCR engine installed. Enable in plugin settings.' };
  } catch (error) {
    return { success: false, error: error instanceof Error ? error.message : String(error) };
  }
}

// ──────────────────────────────────────────────────────
// Zod schemas for tool parameters
// ──────────────────────────────────────────────────────

export const parsePdfSchema = z.object({
  file_path: z.string().min(1).describe('Path to the PDF file'),
});

export const parseEpubSchema = z.object({
  file_path: z.string().min(1).describe('Path to the EPUB file'),
});

export const parseHtmlSchema = z.object({
  file_path: z.string().min(1).describe('Path to the HTML file'),
  options: z.object({
    extract_links: z.boolean().optional().default(true),
  }).optional(),
});

export const ocrImageSchema = z.object({
  file_path: z.string().min(1).describe('Path to the image file (PNG, JPG, BMP)'),
});


// ──────────────────────────────────────────────────────
// Text Chunker for RAG
// ──────────────────────────────────────────────────────

export async function chunk_text(text: string, options?: { 
  chunk_size?: number; 
  overlap?: number; 
}): Promise<{ success: boolean; chunks: Array<{ index: number; content: string }>; total_chunks: number; error?: string }> {
  try {
    const chunkSize = options?.chunk_size ?? 500; // chars per chunk
    const overlap = options?.overlap ?? 50; // chars of overlap between chunks
    
    if (!text || text.length === 0) {
      return { success: false, chunks: [], total_chunks: 0, error: 'Input text is empty' };
    }
    
    const chunks: Array<{ index: number; content: string }> = [];
    let start = 0;
    let index = 0;
    
    while (start < text.length) {
      const end = Math.min(start + chunkSize, text.length);
      let chunkEnd = end;
      
      // Try to break at sentence boundary if possible
      if (end < text.length) {
        const lastPeriod = text.lastIndexOf('.', end - 1);
        const lastNewline = text.lastIndexOf('\n', end - 1);
        const lastSpace = text.lastIndexOf(' ', end - 1);
        
        // Prefer sentence boundary, then newline, then space
        let breakPoint = Math.max(lastPeriod, lastNewline, lastSpace);
        if (breakPoint > start + chunkSize * 0.5) {
          chunkEnd = breakPoint + 1; // Include the period/newline/space
        } else {
          chunkEnd = end;
        }
      }
      
      const chunkContent = text.slice(start, chunkEnd).trim();
      if (chunkContent.length > 0) {
        chunks.push({ index: index++, content: chunkContent });
      }
      
      start = chunkEnd - overlap;
      if (start <= 0) start = 0; // Prevent infinite loop on small text
    }
    
    return { success: true, chunks, total_chunks: chunks.length };
  } catch (error) {
    return { success: false, chunks: [], total_chunks: 0, error: error instanceof Error ? error.message : String(error) };
  }
}