Serper Search MCP Server

/** * Handler implementation for the Deep Research tool */ import { ResearchOrchestrator } from '../../orchestration/researchOrchestrator.js'; import { ResearchRequest, ResearchResult } from '../../../domain/types/research.js'; import { ErrorCode, McpError } from '@modelcontextprotocol/sdk/types.js'; /** * Handles requests to the deep-research tool */ export class ResearchToolHandler { private researchOrchestrator: ResearchOrchestrator; /** * Creates a new research tool handler */ constructor(researchOrchestrator: ResearchOrchestrator) { this.researchOrchestrator = researchOrchestrator; } /** * Handles the research tool request * @param args Tool input arguments * @returns Formatted tool response */ async handleResearchRequest(args: any) { try { // Validate and create research request const request: ResearchRequest = { query: args.query, depth: args.depth || 'standard', maxSources: args.maxSources || 10 }; console.log(`Deep Research request: ${JSON.stringify(request)}`); // Execute the research process const result = await this.researchOrchestrator.executeResearch(request); // Return formatted results return { content: [ { type: 'text', text: this.formatResearchResult(result), }, ], }; } catch (error) { console.error('Error in research tool handler:', error); if (error instanceof McpError) { throw error; } // Handle unexpected errors throw new McpError( ErrorCode.InternalError, `Unexpected error during research: ${(error as Error).message}` ); } } /** * Formats the research result as rich markdown with enhanced citations and metadata */ private formatResearchResult(result: ResearchResult): string { // Start with the main answer let formattedResult = result.answer; // Add research methodology section if not already included if (!formattedResult.includes('## Research Methodology')) { formattedResult += `\n\n## Research Methodology\n`; formattedResult += `This analysis was generated through a multi-stage research process:\n\n`; formattedResult += `1. **Query Analysis**: The original query was analyzed to identify key dimensions and knowledge requirements\n`; formattedResult += `2. **Strategic Decomposition**: ${result.subQueriesGenerated} targeted sub-queries were generated to explore different aspects\n`; formattedResult += `3. **Comprehensive Search**: ${result.searchesPerformed} distinct searches were performed to gather information\n`; formattedResult += `4. **Source Collection**: ${result.citations.length} relevant sources were identified and analyzed\n`; formattedResult += `5. **Critical Synthesis**: Information was evaluated, compared, and integrated to form a comprehensive analysis\n\n`; // Add sub-query information if available if (result.subQueriesGenerated > 0) { formattedResult += `### Research Coverage\n`; formattedResult += `The research process explored multiple dimensions of the topic to ensure comprehensive coverage:\n\n`; formattedResult += `- **Information Breadth**: Multiple perspectives and viewpoints were incorporated\n`; formattedResult += `- **Analytical Depth**: Core concepts were examined in detail with supporting evidence\n`; formattedResult += `- **Critical Evaluation**: Sources were assessed for reliability and relevance\n`; } } // Add source quality assessment if not already included and there are citations if (!formattedResult.includes('## Source Assessment') && result.citations.length > 0) { formattedResult += `\n\n## Source Assessment\n`; formattedResult += `The information in this report comes from ${result.citations.length} distinct sources. `; // Group sources by approximate domain const domains = new Map<string, number>(); result.citations.forEach(citation => { try { const url = new URL(citation.url); const domain = url.hostname; domains.set(domain, (domains.get(domain) || 0) + 1); } catch (e) { // Skip invalid URLs } }); if (domains.size > 0) { formattedResult += `Sources include:\n\n`; // List the top domains Array.from(domains.entries()) .sort((a, b) => b[1] - a[1]) .slice(0, 5) .forEach(([domain, count]) => { formattedResult += `- ${domain}: ${count} source${count > 1 ? 's' : ''}\n`; }); } } // Add sources section if not already included if (!formattedResult.includes('## Sources')) { formattedResult += `\n\n## Sources\n\n`; // Sort citations by relevance score before listing const sortedCitations = [...result.citations] .sort((a, b) => b.relevanceScore - a.relevanceScore); // Add sources with numbers matching citations in the text sortedCitations.forEach((citation, index) => { const sourceNum = index + 1; formattedResult += `[${sourceNum}] **${citation.title}**\n`; formattedResult += ` URL: ${citation.url}\n`; if (citation.snippet && citation.snippet.trim().length > 0) { formattedResult += ` Summary: ${citation.snippet}\n`; } formattedResult += `\n`; }); } return formattedResult; } }