import gradio as gr import json import base64 import requests import os HF_TOKEN = os.environ.get("HF_CV_ROBOT_TOKEN") MODEL = "Qwen/Qwen2-VL-7B-Instruct" if not HF_TOKEN: print("ERROR: HF_CV_ROBOT_TOKEN environment variable not set.") def process(payload: dict): try: if not HF_TOKEN: return {"error": "Hugging Face token is missing. Please check Space secrets."} robot_id = payload.get("robot_id", "unknown") image_b64 = payload["image_b64"] # Router API payload using "type": "file" for base64 headers = {"Authorization": f"Bearer {HF_TOKEN}"} data = { "model": MODEL, "messages": [ { "role": "user", "content": [ {"type": "text", "text": "Describe this image in detail."}, {"type": "file", "file": image_b64} # <- 這裡用 file 直接放 base64 ] } ] } resp = requests.post( "https://router.huggingface.co/v1/chat/completions", headers=headers, json=data, timeout=60 ) if resp.status_code != 200: print(f"VLM API error: {resp.status_code}, {resp.text}") return {"error": f"VLM API error: {resp.status_code}, {resp.text}"} try: vlm_text = resp.json()["choices"][0]["message"]["content"][0]["text"] except (KeyError, IndexError, json.JSONDecodeError) as e: return {"error": f"Failed to parse VLM response: {e}, Response text: {resp.text}"} return { "received": True, "robot_id": robot_id, "vllm_analysis": vlm_text } except Exception as e: print(f"An unexpected error occurred: {e}") return {"error": str(e)} demo = gr.Interface( fn=process, inputs=gr.JSON(label="Input Payload (Dict format)"), outputs=gr.JSON(label="Reply to Jetson"), api_name="predict" ) if __name__ == "__main__": demo.launch(mcp_server=True)