neuroverse

Instructions

Implementation Reference

• npm/src/index.ts:439-444 (handler)

  The tool handler for "neuroverse_transcribe" that calls the transcription service.

  async (params) => {
    const text = await transcribeAudio(params.audio_path);
    return {
      content: [{ type: "text" as const, text: JSON.stringify({ text }, null, 2) }],
    };
  }

• npm/src/core/voice.ts:32-43 (handler)

  The implementation logic for audio transcription.

  export async function transcribeAudio(audioPath: string): Promise<string> {
    if (!config.whisperEndpoint) throw new Error("Whisper endpoint not configured.");
    
    // Note: For a real implementation, you'd use form-data and passing the file stream.
    // This serves as the integration pattern.
    try {
      const response = await axios.post(
        config.whisperEndpoint,
        { file_path: audioPath }, // Adjust based on your Whisper REST API spec
        { headers: { "Content-Type": "application/json" } }
      );
      return response.data.text ?? "Transcription returned empty.";

• npm/src/index.ts:414-418 (schema)

  Zod schema definition for input validation of the transcribe tool.

  const TranscribeSchema = z
    .object({
      audio_path: z.string().min(1).describe("Absolute path to the audio file to transcribe"),
    })
    .strict();

• npm/src/index.ts:420-438 (registration)

  Registration of the "neuroverse_transcribe" tool in the MCP server.

  server.registerTool(
    "neuroverse_transcribe",
    {
      title: "Transcribe Audio",
      description: `Transcribe an audio file to text using Whisper STT.
  
  Args:
    - audio_path (string): Absolute path to the audio file
  
  Returns:
    JSON with the transcribed text`,
      inputSchema: TranscribeSchema,
      annotations: {
        readOnlyHint: true,
        destructiveHint: false,
        idempotentHint: true,
        openWorldHint: true,
      },
    },