transcribe_audio

Implementation Reference

• src/index.js:25-64 (registration)

  Registers the 'transcribe_audio' tool on the MCP server with schema (file_path, language, model) and handler that reads an audio file and calls OpenAI-compatible speech-to-text API.

  server.tool(
    "transcribe_audio",
    {
      file_path: z.string().describe("Absolute path to the audio file to transcribe"),
      language: z
        .string()
        .optional()
        .describe("ISO-639-1 language code, e.g. 'en', 'fr'. Omit for auto-detect."),
      model: z
        .string()
        .optional()
        .describe(`Whisper model ID to use. Defaults to ${STT_MODEL}`),
    },
    async ({ file_path, language, model }) => {
      if (!fs.existsSync(file_path)) {
        return {
          content: [{ type: "text", text: `Error: file not found: ${file_path}` }],
          isError: true,
        };
      }
  
      let transcription;
      try {
        transcription = await client.audio.transcriptions.create({
          file: fs.createReadStream(file_path),
          model: model || STT_MODEL,
          ...(language ? { language } : {}),
        });
      } catch (err) {
        return {
          content: [{ type: "text", text: `Error: transcription failed: ${err.message}` }],
          isError: true,
        };
      }
  
      return {
        content: [{ type: "text", text: transcription.text }],
      };
    }
  );

• src/index.js:27-36 (schema)

  Zod schema defining input parameters: required file_path, optional language (ISO-639-1), and optional model (Whisper model ID).

  {
    file_path: z.string().describe("Absolute path to the audio file to transcribe"),
    language: z
      .string()
      .optional()
      .describe("ISO-639-1 language code, e.g. 'en', 'fr'. Omit for auto-detect."),
    model: z
      .string()
      .optional()
      .describe(`Whisper model ID to use. Defaults to ${STT_MODEL}`),

• src/index.js:38-63 (handler)

  Handler function that checks file existence, streams the audio file to the OpenAI-compatible STT API via client.audio.transcriptions.create, and returns the transcribed text.

  async ({ file_path, language, model }) => {
    if (!fs.existsSync(file_path)) {
      return {
        content: [{ type: "text", text: `Error: file not found: ${file_path}` }],
        isError: true,
      };
    }
  
    let transcription;
    try {
      transcription = await client.audio.transcriptions.create({
        file: fs.createReadStream(file_path),
        model: model || STT_MODEL,
        ...(language ? { language } : {}),
      });
    } catch (err) {
      return {
        content: [{ type: "text", text: `Error: transcription failed: ${err.message}` }],
        isError: true,
      };
    }
  
    return {
      content: [{ type: "text", text: transcription.text }],
    };
  }