search_pdf

Instructions

Implementation Reference

• src/index.ts:78-100 (registration)

  Tool registration: 'search_pdf' tool definition with name, description, and inputSchema for filePath, searchTerm, and caseSensitive parameters.

  {
    name: 'search_pdf',
    description: 'Search for text in a PDF and return all matches with context',
    inputSchema: {
      type: 'object',
      properties: {
        filePath: {
          type: 'string',
          description: 'Absolute path to the PDF file',
        },
        searchTerm: {
          type: 'string',
          description: 'Text to search for in the PDF',
        },
        caseSensitive: {
          type: 'boolean',
          description: 'Whether the search should be case sensitive',
          default: false,
        },
      },
      required: ['filePath', 'searchTerm'],
    },
  },

• src/index.ts:242-259 (handler)

  Handler: 'search_pdf' case in the CallToolRequestSchema handler. It extracts filePath, searchTerm, and optional caseSensitive from args, calls searchInPDF(), and returns results as JSON.

  case 'search_pdf': {
    const { filePath, searchTerm, caseSensitive } = args as {
      filePath: string;
      searchTerm: string;
      caseSensitive?: boolean;
    };
    
    const results = await searchInPDF(filePath, searchTerm, caseSensitive);
    
    return {
      content: [
        {
          type: 'text',
          text: JSON.stringify(results, null, 2),
        },
      ],
    };
  }

• src/pdf-tools.ts:148-189 (helper)

  Helper: searchInPDF() - the core implementation that reads a PDF file, iterates through each page, uses a regex to find matches, and returns SearchResult objects with page number, matched text, context (50 chars on each side), and position.

  export async function searchInPDF(
    filePath: string,
    searchTerm: string,
    caseSensitive: boolean = false
  ): Promise<SearchResult[]> {
    try {
      const dataBuffer = await fs.readFile(filePath);
      const parser = new PDFParse({ data: dataBuffer });
      const info = await parser.getInfo();
      
      const results: SearchResult[] = [];
      const searchRegex = new RegExp(
        searchTerm.replace(/[.*+?^${}()|[\]\\]/g, '\\$&'),
        caseSensitive ? 'g' : 'gi'
      );
  
      // Search through each page
      for (let pageNo = 1; pageNo <= info.total; pageNo++) {
        const pageResult = await parser.getText({ partial: [pageNo] });
        const pageText = pageResult.text;
        
        let match;
        while ((match = searchRegex.exec(pageText)) !== null) {
          const contextStart = Math.max(0, match.index - 50);
          const contextEnd = Math.min(pageText.length, match.index + match[0].length + 50);
          const context = pageText.substring(contextStart, contextEnd).replace(/\n/g, ' ');
  
          results.push({
            page: pageNo,
            text: match[0],
            context: `...${context}...`,
            position: match.index,
          });
        }
      }
      
      await parser.destroy();
      return results;
    } catch (error) {
      throw new Error(`Failed to search PDF: ${error instanceof Error ? error.message : String(error)}`);
    }
  }

• src/types.ts:36-41 (schema)

  Schema: SearchResult interface defining the type returned by searchInPDF - includes page number, matched text, context (with surrounding chars), and position index.

  export interface SearchResult {
    page: number;
    text: string;
    context: string;
    position: number;
  }