Document Extractor MCP Server

Instructions

Implementation Reference

• src/index.js:898-951 (handler)

  The tool `search_documents` is registered using `server.tool` and uses `searchDocuments` to fetch the data.

  const searchDocumentsTool = server.tool(
    'search_documents',
    'Search documents by title or content using full-text search',
    {
      query: z.string().min(1, 'Query cannot be empty').describe('Search query to find documents (searches title and content)'),
      limit: z.number().min(1).max(100).optional().default(50).describe('Maximum number of results to return (default: 50)')
    },    async ({ query, limit = 50 }) => {
      try {
        // Only authenticate when tool is actually invoked
        await authenticateWhenNeeded();
        
        const result = await searchDocuments(query, limit);
        
        if (result.items.length === 0) {
          return {
            content: [
              {
                type: 'text',
                text: `🔍 No documents found matching "${query}"`
              }
            ]
          };
        }
        
        const searchResults = result.items.map(doc => 
          `**${doc.title}** (ID: ${doc.id})\n` +
          `Source: ${doc.metadata?.source || 'Unknown'}\n` +
          `Domain: ${doc.metadata?.domain || 'Unknown'}\n` +
          `Created: ${new Date(doc.created).toLocaleString()}\n` +
          `${doc.metadata?.url ? `URL: ${doc.metadata.url}\n` : ''}` +
          `Preview: ${doc.content.substring(0, 150)}...\n`
        ).join('\n---\n');
        
        return {
          content: [
            {
              type: 'text',
              text: `🔍 Found ${result.items.length} documents matching "${query}":\n\n${searchResults}`
            }
          ]
        };
      } catch (error) {
        return {
          content: [
            {
              type: 'text',
              text: `❌ Error: ${error.message}`
            }
          ],
          isError: true
        };
      }
    }
  );

• src/index.js:600-620 (handler)

  The `searchDocuments` function is the core logic that performs the actual full-text search against the PocketBase collection.

  // Search documents in PocketBase (with lazy initialization)
  async function searchDocuments(query, limit = 50) {
    try {
      await authenticateWhenNeeded();
      
      if (!DOCUMENTS_COLLECTION) {
        initializeConfig();
      }
      
      const records = await pb.collection(DOCUMENTS_COLLECTION).getList(1, limit, {
        filter: `title ~ "${query}" || content ~ "${query}"`,
        sort: '-created'
      });
      
      debugLog('Documents searched in PocketBase', { query, count: records.items.length });
      return records;
    } catch (error) {
      debugLog('Error searching documents', { error: error.message });
      throw new Error(`Failed to search documents: ${error.message}`);
    }
  }