app.py

import os
import copy
import base64
import requests
import tempfile
import secrets
import gradio as gr
from huggingface_hub import upload_file, InferenceClient
from PIL import Image

# --- Config ---
HF_TOKEN = os.environ.get("HF_CV_ROBOT_TOKEN")
HF_DATASET_REPO = "OppaAI/Robot_MCP"
# Model specifically for VLM (image-to-text) tasks on Hugging Face
HF_VLM_MODEL = "llava-hf/llava-interleave-qwen-0.5b-hf" # A suitable VLM model

if not HF_TOKEN:
    raise ValueError("HF_TOKEN environment variable not set.")

# Initialize the Hugging Face Inference Client
hf_client = InferenceClient(token=HF_TOKEN)

# --- Helper Functions ---
def save_and_upload_image(image_b64):
    """Save image to /tmp and upload to HF dataset."""
    image_bytes = base64.b64decode(image_b64)
    # Use a unique filename to prevent conflicts in /tmp
    local_tmp_path = f"/tmp/uploaded_image_{secrets.token_hex(8)}.jpg" 
    with open(local_tmp_path, "wb") as f:
        f.write(image_bytes)

    path_in_repo = f"images/uploaded_image_{secrets.token_hex(8)}.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:
        robot_id = payload.get("robot_id", "unknown")
        image_b64 = payload["image_b64"]

        # 1️⃣ Save & upload image
        local_tmp_path, hf_url, path_in_repo, size_bytes = save_and_upload_image(image_b64)

        # 2️⃣ Prepare prompt (optional, some models ignore this for basic image_to_text)
        # prompt = "Describe this image in detail." 
        
        # Open the image using PIL for the InferenceClient
        image = Image.open(local_tmp_path)

        # 3️⃣ Call VLM using Hugging Face Inference Client
        # Removed the problematic 'details' argument
        vlm_text = hf_client.image_to_text(
            image=image,
            model=HF_VLM_MODEL,
            # details=True # <-- REMOVED THIS LINE
        )

        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_description": vlm_text
        }

    except Exception as e:
        return {"error": str(e)}

# --- Gradio MCP Interface ---
demo = gr.Interface(
    fn=process_and_describe,
    inputs=gr.JSON(label="Input Payload (Dict format with 'image_b64')"),
    outputs=gr.JSON(label="Reply to Jetson"),
    api_name="predict"
)

if __name__ == "__main__":
    # You will need to install the required libraries:
    # pip install gradio huggingface_hub Pillow requests
    demo.launch(mcp_server=True)