app.py

import os
import base64
import gradio as gr
from huggingface_hub import upload_file, InferenceClient
import json
from fastmcp import MCP, MCPClient
from playsound import playsound
from gtts import gTTS

# --- Config ---
HF_DATASET_REPO = "OppaAI/Robot_MCP"
HF_VLM_MODEL = "Qwen/Qwen2.5-VL-7B-Instruct" 

# --- MCP server instance ---
mcp = MCP()  # 用於定義工具

# --- MCP Tool ---
@mcp.tools()
def say_hi(text="Hi!"):
    # 1️⃣ 生成 mp3
    tts = gTTS(text=text, lang="en")
    tmp_path = "/tmp/say_hi.mp3"
    tts.save(tmp_path)
    
    # 2️⃣ 播放音檔
    playsound(tmp_path)
    
    return f"Played: {text}"

# --- Helper Functions ---
def save_and_upload_image(image_b64, hf_token):
    image_bytes = base64.b64decode(image_b64)
    local_tmp_path = "/tmp/tmp.jpg"
    with open(local_tmp_path, "wb") as f:
        f.write(image_bytes)

    path_in_repo = "images/tmp.jpg"
    upload_file(
        path_or_fileobj=local_tmp_path,
        path_in_repo=path_in_repo,
        repo_id=HF_DATASET_REPO,
        token=hf_token,
        repo_type="dataset"
    )

    hf_image_url = f"https://huggingface.co/datasets/{HF_DATASET_REPO}/resolve/main/{path_in_repo}"
    return local_tmp_path, hf_image_url, path_in_repo, len(image_bytes)

# --- Main MCP function ---
def process_and_describe(payload: dict):
    try:
        hf_token = payload.get("hf_token")
        if not hf_token:
            return {"error": "HF token not provided in payload."}

        robot_id = payload.get("robot_id", "unknown")
        image_b64 = payload.get("image_b64")
        if not image_b64:
            return {"error": "No image provided."}

        local_tmp_path, hf_url, path_in_repo, size_bytes = save_and_upload_image(image_b64, hf_token)
        hf_client = InferenceClient(token=hf_token)

        system_prompt = """
        You are a helpful robot assistant.
        1. Describe the image in detail.
        2. Suggest what the robot should do next.
           - Human figure → say 'Hi'.
        Always respond in JSON:
        {"description": "...", "action": "say_hi"}
        """

        messages_payload = [
            {"role": "system", "content": system_prompt},
            {"role": "user", "content": [
                {"type": "text", "text": "Here is an image."},
                {"type": "image_url", "image_url": {"url": f"data:image/jpeg;base64,{image_b64}"}}
            ]}
        ]

        chat_completion = hf_client.chat.completions.create(
            model=HF_VLM_MODEL,
            messages=messages_payload,
            max_tokens=200
        )

        vlm_text = chat_completion.choices[0].message.content.strip()
        action_data = {}
        try:
            action_data = json.loads(vlm_text)
        except Exception:
            action_data = {"description": vlm_text, "action": "unknown"}

        # --- Call MCP tool ---
        vlm_action = action_data.get("action")
        tool_result = None
        if vlm_action == "say_hi":
            tool_result = say_hi(text="Hi!")  # 這裡會生成 /tmp/say_hi.mp3

        return {
            "saved_to_hf_hub": True,
            "repo_id": HF_DATASET_REPO,
            "path_in_repo": path_in_repo,
            "image_url": hf_url,
            "file_size_bytes": size_bytes,
            "robot_id": robot_id,
            "vlm_response": vlm_text,
            "vlm_action": vlm_action,
            "vlm_description": action_data.get("description", ""),
            "tool_result": tool_result
        }

    except Exception as e:
        return {"error": f"An API error occurred: {str(e)}"}

# --- Gradio MCP Interface ---
demo = gr.Interface(
    fn=process_and_describe,
    inputs=gr.JSON(label="Input Payload"),
    outputs=gr.JSON(label="Reply to Jetson"),
    api_name="predict"
)

if __name__ == "__main__":
    demo.launch(mcp_server=True)