mcp-summarizer

Implementation Reference

• src/index.ts:23-45 (handler)

  The asynchronous handler function that generates a summary of the input text using the Google Gemini model via the 'ai' SDK. It constructs a prompt, calls generateText, and returns the result in MCP format, with error handling.

  async ({ text, maxLength, language }) => {
      try {
          const prompt = `Please summarize the following text in ${language}, keeping the summary within ${maxLength} characters:\n\n${text}`;
  
          const model = google.chat("gemini-1.5-pro");
          const result = await generateText({
              model: model,
              prompt: prompt,
              maxTokens: maxLength,
              temperature: 0.5
          });
  
          return {
              content: [{
                  type: "text",
                  text: result.text
              }]
          };
      } catch (error) {
          console.error('Summarization error:', error);
          throw new Error('Failed to generate summary');
      }
  }

• src/index.ts:18-22 (schema)

  Zod schema for the 'summarize' tool inputs: 'text' (required string, min length 1), 'maxLength' (optional number, default 200), 'language' (optional string, default 'en').

  {
      text: z.string().min(1),
      maxLength: z.number().optional().default(200),
      language: z.string().optional().default("en")
  },

• src/index.ts:17-46 (registration)

  The server.tool() call that registers the 'summarize' tool, specifying its input schema and handler function.

  server.tool("summarize",
      {
          text: z.string().min(1),
          maxLength: z.number().optional().default(200),
          language: z.string().optional().default("en")
      },
      async ({ text, maxLength, language }) => {
          try {
              const prompt = `Please summarize the following text in ${language}, keeping the summary within ${maxLength} characters:\n\n${text}`;
  
              const model = google.chat("gemini-1.5-pro");
              const result = await generateText({
                  model: model,
                  prompt: prompt,
                  maxTokens: maxLength,
                  temperature: 0.5
              });
  
              return {
                  content: [{
                      type: "text",
                      text: result.text
                  }]
              };
          } catch (error) {
              console.error('Summarization error:', error);
              throw new Error('Failed to generate summary');
          }
      }
  );