Open Search MCP

Instructions

Implementation Reference

• src/tools/pdf/research.ts:17-168 (handler)

  The pdfResearch function is the main handler that executes the 'analyze_pdf' tool. It performs PDF document search based on a query, processes found PDFs with optional OCR, generates summaries and insights based on analysis depth.
  async function pdfResearch(args: ToolInput): Promise<ToolOutput> { const { query, maxDocuments = 10, documentType = 'any', includeOCR = false, forceOCR = false, sources = ['all'], dateRange, analysisDepth = 'medium' } = args; try { logger.info(`Starting PDF research for: ${query}`); if (!query || typeof query !== 'string') { throw new Error('Query parameter is required and must be a string'); } const pdfProcessor = new PDFProcessor(); // Configure search options const searchOptions: PDFSearchOptions = { query, maxDocuments, documentType: documentType as any, includeOCR, forceOCR, sources: Array.isArray(sources) ? sources : [sources], dateRange }; // Search for relevant PDFs const pdfDocuments = await pdfProcessor.searchPDFs(searchOptions); if (pdfDocuments.length === 0) { return { success: true, data: { query, documents: [], totalFound: 0, message: 'No PDF documents found for the given query', searchedAt: new Date().toISOString() }, metadata: { sources: ['pdf-research'], cached: false } }; } // Process PDFs based on analysis depth const processedDocuments = []; const maxToProcess = analysisDepth === 'shallow' ? Math.min(3, pdfDocuments.length) : analysisDepth === 'medium' ? Math.min(5, pdfDocuments.length) : Math.min(10, pdfDocuments.length); for (let i = 0; i < maxToProcess; i++) { const pdfDoc = pdfDocuments[i]; logger.info(`Processing PDF ${i + 1}/${maxToProcess}: ${pdfDoc.title}`); try { const processedPDF = await pdfProcessor.processPDF(pdfDoc, includeOCR, forceOCR); if (processedPDF) { // Create summary based on analysis depth const summary = createPDFSummary(processedPDF, analysisDepth as string); processedDocuments.push({ id: processedPDF.id, title: processedPDF.title, url: processedPDF.url, source: processedPDF.source, summary, metadata: { pages: processedPDF.content.pages, author: processedPDF.metadata.author, creationDate: processedPDF.metadata.creationDate, processingMethod: processedPDF.processing.method, ocrConfidence: processedPDF.processing.ocrConfidence }, structure: { sectionsCount: processedPDF.structure.sections.length, referencesCount: processedPDF.structure.references.length, figuresCount: processedPDF.structure.figures.length, tablesCount: processedPDF.structure.tables.length }, relevanceScore: pdfDoc.relevanceScore }); } } catch (error) { logger.warn(`Failed to process PDF: ${pdfDoc.title}`, error); // Add basic info even if processing failed processedDocuments.push({ id: pdfDoc.id, title: pdfDoc.title, url: pdfDoc.url, source: pdfDoc.source, summary: 'PDF processing failed - document available for manual review', metadata: { processingError: true }, relevanceScore: pdfDoc.relevanceScore }); } } // Generate research insights const insights = generateResearchInsights(processedDocuments, query); const result: ToolOutput = { success: true, data: { query, documents: processedDocuments, totalFound: pdfDocuments.length, totalProcessed: processedDocuments.length, insights, searchOptions: { documentType, includeOCR, forceOCR, sources: searchOptions.sources, analysisDepth }, searchedAt: new Date().toISOString() }, metadata: { sources: ['pdf-research'], cached: false } }; logger.info(`PDF research completed: ${processedDocuments.length} documents processed for ${query}`); return result; } catch (error) { logger.error(`Failed PDF research for ${query}:`, error); return { success: false, error: `Failed to conduct PDF research: ${error instanceof Error ? error.message : 'Unknown error'}`, data: null, metadata: { sources: ['pdf-research'], cached: false } }; } }
• src/tools/pdf/research.ts:339-448 (registration)

  The registerPDFResearchTools function creates and registers the 'analyze_pdf' tool using createTool, defines its metadata, input schema, and registers it with the tool registry.
  export function registerPDFResearchTools(registry: ToolRegistry): void { logger.info('Registering PDF research tools...'); // PDF Research tool const pdfResearchTool = createTool( 'analyze_pdf', 'Conduct comprehensive PDF research with document discovery, processing, and analysis', 'pdf', 'pdf-research', pdfResearch, { cacheTTL: 3600, // 1 hour cache rateLimit: 10, // 10 requests per minute requiredParams: ['query'], optionalParams: ['maxDocuments', 'documentType', 'includeOCR', 'forceOCR', 'sources', 'dateRange', 'analysisDepth'] } ); pdfResearchTool.inputSchema = { type: 'object', properties: { query: { type: 'string', description: 'Research query for PDF document search' }, maxDocuments: { type: 'number', description: 'Maximum number of documents to find (default: 10)' }, documentType: { type: 'string', description: 'Type of documents to search for', enum: ['academic', 'report', 'manual', 'any'] }, includeOCR: { type: 'boolean', description: 'Enable OCR for scanned PDFs (default: false)' }, forceOCR: { type: 'boolean', description: 'Force OCR processing even for good quality text - useful for testing OCR functionality (default: false)' }, sources: { type: 'array', items: { type: 'string' }, description: 'Sources to search (arxiv, pubmed, web, all)' }, dateRange: { type: 'object', properties: { start: { type: 'string', description: 'Start date (YYYY-MM-DD)' }, end: { type: 'string', description: 'End date (YYYY-MM-DD)' } }, description: 'Date range filter for documents' }, analysisDepth: { type: 'string', description: 'Depth of analysis (shallow, medium, deep)', enum: ['shallow', 'medium', 'deep'] } }, required: ['query'] }; registry.registerTool(pdfResearchTool); // PDF Discovery tool const pdfDiscoveryTool = createTool( 'pdf_discovery', 'Discover PDF documents without full processing - fast PDF search and listing', 'pdf', 'pdf-discovery', pdfDiscovery, { cacheTTL: 1800, // 30 minutes cache rateLimit: 15, // 15 requests per minute requiredParams: ['query'], optionalParams: ['maxResults', 'sources', 'documentType'] } ); pdfDiscoveryTool.inputSchema = { type: 'object', properties: { query: { type: 'string', description: 'Search query for PDF document discovery' }, maxResults: { type: 'number', description: 'Maximum number of results to return (default: 20)' }, sources: { type: 'array', items: { type: 'string' }, description: 'Sources to search (arxiv, pubmed, web, all)' }, documentType: { type: 'string', description: 'Type of documents to search for', enum: ['academic', 'report', 'manual', 'any'] } }, required: ['query'] }; registry.registerTool(pdfDiscoveryTool); logger.info('PDF research tools registered successfully'); }
• src/tools/pdf/research.ts:357-401 (schema)

  Input schema definition for the 'analyze_pdf' tool, specifying parameters like query (required), maxDocuments, analysisDepth, etc.
  pdfResearchTool.inputSchema = { type: 'object', properties: { query: { type: 'string', description: 'Research query for PDF document search' }, maxDocuments: { type: 'number', description: 'Maximum number of documents to find (default: 10)' }, documentType: { type: 'string', description: 'Type of documents to search for', enum: ['academic', 'report', 'manual', 'any'] }, includeOCR: { type: 'boolean', description: 'Enable OCR for scanned PDFs (default: false)' }, forceOCR: { type: 'boolean', description: 'Force OCR processing even for good quality text - useful for testing OCR functionality (default: false)' }, sources: { type: 'array', items: { type: 'string' }, description: 'Sources to search (arxiv, pubmed, web, all)' }, dateRange: { type: 'object', properties: { start: { type: 'string', description: 'Start date (YYYY-MM-DD)' }, end: { type: 'string', description: 'End date (YYYY-MM-DD)' } }, description: 'Date range filter for documents' }, analysisDepth: { type: 'string', description: 'Depth of analysis (shallow, medium, deep)', enum: ['shallow', 'medium', 'deep'] } }, required: ['query'] };
• src/index.ts:251-252 (registration)

  Call to registerPDFResearchTools in the main server initialization, which registers the 'analyze_pdf' tool among others.
  registerPDFResearchTools(this.toolRegistry); // 1 tool: analyze_pdf
• src/utils/input-validator.ts:173-182 (schema)

  Additional validation schema 'pdfAnalysis' used in input-validator for 'analyze_pdf', though the tool uses query-based input primarily.
  // PDF analysis pdfAnalysis: z.object({ filePath: CommonSchemas.filePath, extractText: z.boolean().optional().default(true), extractMetadata: z.boolean().optional().default(true), pageRange: z.object({ start: CommonSchemas.positiveInteger, end: CommonSchemas.positiveInteger, }).optional(), }),