app2.py

import os

import shutil



# --- SEARCH AND DESTROY POISONED CACHE ---

corrupted_dir = "/root/.cache/huggingface/hub/models--google--umt5-base"

if os.path.exists(corrupted_dir):

    print("[SYSTEM] Found corrupted UMT5 cache. Deleting...")

    shutil.rmtree(corrupted_dir, ignore_errors=True)

else:

    print("[SYSTEM] Cache is clean.")



# --- YOUR ORIGINAL CODE STARTS HERE ---

from flask import Flask, request, jsonify, send_from_directory # Added send_from_directory

from flask_sock import Sock

from transformers import AutoModel

import torch

import time

import json

from flask_cors import CORS



app = Flask(__name__)

CORS(app)

sock = Sock(app) # Initialize WebSocket support



print("[SYSTEM] Booting up Network Server...")

print("[SYSTEM] Loading FloodDiffusionTiny model from Hugging Face...")



# 1. Load the model

model = AutoModel.from_pretrained(

    "ShandaAI/FloodDiffusionTiny",

    trust_remote_code=True

)



# 2. Cloud Architecture Override

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

model = model.to(device)

print(f"[SYSTEM] Model loaded successfully onto device: {device}")



@app.route('/')

def serve_ui():

    # This tells Flask to send the index.html file to the user's browser

    return send_from_directory('.', 'index.html')

# --- ADD THIS NEW ROUTE HERE ---
@app.route('/<path:filename>')
def serve_static_files(filename):
    # This allows Flask to send the smpl.glb file when the browser asks for it!
    return send_from_directory('.', filename)

# --- THE NEW WEBSOCKET PIPELINE ---

@sock.route('/api/generate_stream')

def stream_motion(ws):

    print("\n[NETWORK] 🟢 WebSocket Connection Opened! Client connected.")

    

    # Keep the connection open forever

    while True:

        try:

            # 1. Wait for the live prompt from the client's text box

            raw_data = ws.receive() 

            if raw_data is None:

                continue

                

            data = json.loads(raw_data)

            text_prompt = data.get('prompt', '')

            ticket_number = data.get('ticket', 0)

            print(f"[NETWORK] Live Prompt Received: '{text_prompt}'")

            

            start_time = time.time()

            

            # 2. Server Processing (Inference)

            motion_joints = model(text_prompt, length=150, output_joints=True)

            processing_time = (time.time() - start_time) * 1000 

            

            # 3. Format Network Payload

            payload = {

                "status": "success",

                "ticket": ticket_number,

                "latency_ms": round(processing_time, 2),

                "tensor_shape": list(motion_joints.shape),

                "data": motion_joints.tolist() 

            }

            

            # 4. Push data back through the pipe instantly!

            ws.send(json.dumps(payload))

            print(f"[NETWORK] ⚡ Streamed 30 frames to client in {processing_time:.2f}ms")

            

        except Exception as e:

            print(f"[NETWORK] 🔴 WebSocket Error or Disconnect: {e}")

            break



if __name__ == '__main__':

    # --- PORT 7860 FOR HUGGING FACE ---

    app.run(host='0.0.0.0', port=7860, debug=False)