MCP Power - Knowledge Search Server

Instructions

Implementation Reference

• src/tools/listDatasets.ts:61-197 (handler)

  The main handler function `handleListDatasets` that executes the `knowledge.listDatasets` tool logic. It validates input arguments using Zod schema, retrieves datasets from the registry (ready and optionally errored ones), formats a detailed markdown-like response with statistics, and handles errors.
  export async function handleListDatasets( args: unknown ): Promise<{ content: Array<{ type: string; text: string }>; isError?: boolean; }> { const startTime = Date.now(); try { // Validate arguments const parseResult = ListDatasetsArgsSchema.safeParse(args); if (!parseResult.success) { logger.warn( { errors: parseResult.error.errors, args }, 'Invalid arguments for knowledge.listDatasets' ); return { content: [ { type: 'text', text: `Invalid arguments: ${parseResult.error.errors .map((e) => `${e.path.join('.')}: ${e.message}`) .join(', ')}`, }, ], isError: true, }; } const { includeErrors } = parseResult.data; logger.info({ includeErrors }, 'Listing datasets'); // Get dataset registry const registry = getDatasetRegistry(); // Get ready datasets const readyDatasets = registry.listReady(); const results: DatasetListItem[] = readyDatasets.map((dataset) => ({ id: dataset.id, name: dataset.name, description: dataset.description || '', status: 'ready' as const, })); // Optionally include datasets with errors if (includeErrors) { const errors = registry.getErrors(); for (const error of errors) { // Extract dataset ID from manifest path (e.g., /path/to/datasets/my-dataset/manifest.json -> my-dataset) const pathParts = error.manifestPath.split('/'); const datasetId = pathParts[pathParts.length - 2] || 'unknown'; results.push({ id: datasetId, name: datasetId, // Use ID as name fallback description: 'Failed to load dataset', status: 'error' as const, error: error.error, }); } } // Calculate statistics const totalCount = results.length; const readyCount = results.filter((d) => d.status === 'ready').length; const errorCount = results.filter((d) => d.status === 'error').length; // Format response text let responseText = `Found ${totalCount} dataset${totalCount !== 1 ? 's' : ''}`; if (includeErrors && errorCount > 0) { responseText += ` (${readyCount} ready, ${errorCount} with errors)`; } responseText += ':\n\n'; // Add dataset details for (const dataset of results.sort((a, b) => a.id.localeCompare(b.id))) { responseText += `**${dataset.id}**\n`; responseText += ` Name: ${dataset.name}\n`; responseText += ` Description: ${dataset.description}\n`; responseText += ` Status: ${dataset.status}\n`; if (dataset.error) { responseText += ` Error: ${dataset.error}\n`; } responseText += '\n'; } // Add summary footer const duration = Date.now() - startTime; responseText += `---\n`; responseText += `Total: ${totalCount} datasets | Ready: ${readyCount}`; if (includeErrors && errorCount > 0) { responseText += ` | Errors: ${errorCount}`; } responseText += ` | Retrieved in ${duration}ms`; logger.info( { totalCount, readyCount, errorCount, duration, }, 'Listed datasets successfully' ); return { content: [ { type: 'text', text: responseText, }, ], }; } catch (error) { const duration = Date.now() - startTime; logger.error( { error: error instanceof Error ? error.message : String(error), duration, }, 'Failed to list datasets' ); return { content: [ { type: 'text', text: `Error listing datasets: ${ error instanceof Error ? error.message : 'Unknown error' }`, }, ], isError: true, }; } }
• src/tools/listDatasets.ts:19-24 (schema)

  Zod input validation schema for the `knowledge.listDatasets` tool, defining optional `includeErrors` boolean parameter.
  export const ListDatasetsArgsSchema = z .object({ /** Include datasets that failed to load (for diagnostics) */ includeErrors: z.boolean().optional().default(false), }) .strict();
• src/server.ts:137-143 (registration)

  Tool registration in the MCP `listTools` response, defining the name, description, and input schema for `knowledge.listDatasets`.
  name: 'knowledge.listDatasets', description: 'List all registered knowledge datasets available for searching', inputSchema: { type: 'object', properties: {} } }
• src/server.ts:159-160 (registration)

  Dispatch handler in the MCP `callTool` request switch statement that routes `knowledge.listDatasets` invocations to the `handleListDatasets` function.
  case 'knowledge.listDatasets': return await handleListDatasets(args);