debug

example/Dockerfile

@@ -1,59 +0,0 @@
-FROM nvidia/cuda:12.4.0-devel-ubuntu22.04
-
-ARG DEBIAN_FRONTEND=noninteractive
-
-ENV PYTHONUNBUFFERED=1
-
-RUN apt-get update && apt-get install --no-install-recommends -y \
-  build-essential \
-  python3.11 \
-  python3-pip \
-  python3-dev \
-  git \
-  curl \
-  ffmpeg \
-  libglib2.0-0 \
-  libsm6 \
-  libxrender1 \
-  libxext6 \
-  ninja-build \
-  && apt-get clean && rm -rf /var/lib/apt/lists/*
-
-WORKDIR /code
-
-COPY ./requirements.txt /code/requirements.txt
-
-# Set up a new user named "user" with user ID 1000
-RUN useradd -m -u 1000 user
-# Switch to the "user" user
-USER user
-# Set home to the user's home directory
-ENV HOME=/home/user \
-    PATH=/home/user/.local/bin:$PATH
-
-# Set Python path and environment variables
-ENV PYTHONPATH=$HOME/app \
-  PYTHONUNBUFFERED=1 \
-  DATA_ROOT=/tmp/data
-
-RUN echo "Installing requirements.txt"
-RUN pip3 install --no-cache-dir --upgrade -r /code/requirements.txt
-
-# Install NVIDIA Apex with CUDA and C++ extensions
-RUN cd $HOME && \
-    git clone https://github.com/NVIDIA/apex && \
-    cd apex && \
-    NVCC_APPEND_FLAGS="--threads 4" pip install -v --disable-pip-version-check --no-cache-dir --no-build-isolation --global-option="--cpp_ext" --global-option="--cuda_ext" --global-option="--parallel" --global-option="8" ./
-
-WORKDIR $HOME/app
-
-# Copy all files and set proper ownership
-COPY --chown=user . $HOME/app
-
-# Expose the port that server.py uses (8080)
-EXPOSE 8080
-
-ENV PORT 8080
-
-# Run the HF space launcher script which sets up the correct paths
-CMD ["python3", "run_hf_space.py"]

example/client.js

@@ -1,603 +0,0 @@
-// MatrixGame WebSocket Client
-
-// WebSocket connection
-let socket = null;
-let userId = null;
-let isStreaming = false;
-let lastFrameTime = 0;
-let frameCount = 0;
-let fpsUpdateInterval = null;
-
-// DOM Elements
-const connectBtn = document.getElementById('connect-btn');
-const startStreamBtn = document.getElementById('start-stream-btn');
-const stopStreamBtn = document.getElementById('stop-stream-btn');
-const sceneSelect = document.getElementById('scene-select');
-const gameCanvas = document.getElementById('game-canvas');
-const connectionLog = document.getElementById('connection-log');
-const mousePosition = document.getElementById('mouse-position');
-const fpsCounter = document.getElementById('fps-counter');
-const mouseTrackingArea = document.getElementById('mouse-tracking-area');
-
-// Pointer Lock API support check
-const pointerLockSupported = 'pointerLockElement' in document || 
-                            'mozPointerLockElement' in document || 
-                            'webkitPointerLockElement' in document;
-
-// Keyboard DOM elements
-const keyElements = {
-    'w': document.getElementById('key-w'),
-    'a': document.getElementById('key-a'),
-    's': document.getElementById('key-s'),
-    'd': document.getElementById('key-d'),
-    'space': document.getElementById('key-space'),
-    'shift': document.getElementById('key-shift')
-};
-
-// Key mapping to action names
-const keyToAction = {
-    'w': 'forward',
-    'arrowup': 'forward',
-    'a': 'left',
-    'arrowleft': 'left',
-    's': 'back',
-    'arrowdown': 'back',
-    'd': 'right',
-    'arrowright': 'right',
-    ' ': 'jump',
-    'shift': 'attack'
-};
-
-// Key state tracking
-const keyState = {
-    'forward': false,
-    'back': false,
-    'left': false,
-    'right': false,
-    'jump': false,
-    'attack': false
-};
-
-// Mouse state
-const mouseState = {
-    x: 0,
-    y: 0,
-    captured: false
-};
-
-// Test server connectivity before establishing WebSocket
-async function testServerConnectivity() {
-    try {
-        // Get base path by extracting path from the script tag's src attribute
-        let basePath = '';
-        const scriptTags = document.getElementsByTagName('script');
-        for (const script of scriptTags) {
-            if (script.src.includes('client.js')) {
-                const url = new URL(script.src);
-                basePath = url.pathname.replace('/assets/client.js', '');
-                break;
-            }
-        }
-        
-        // Try to fetch the debug endpoint to see if the server is accessible
-        const response = await fetch(`${window.location.protocol}//${window.location.host}${basePath}/api/debug`);
-        if (!response.ok) {
-            throw new Error(`Server returned ${response.status}`);
-        }
-        
-        const debugInfo = await response.json();
-        logMessage(`Server connection test successful! Server time: ${new Date(debugInfo.server_time * 1000).toLocaleTimeString()}`);
-        
-        // Log available routes from server
-        if (debugInfo.all_routes && debugInfo.all_routes.length > 0) {
-            logMessage(`Available routes: ${debugInfo.all_routes.join(', ')}`);
-        }
-        
-        // Return the debug info for connection setup
-        return debugInfo;
-    } catch (error) {
-        logMessage(`Server connection test failed: ${error.message}`);
-        return null;
-    }
-}
-
-// Connect to WebSocket server
-async function connectWebSocket() {
-    // First test connectivity to the server
-    logMessage('Testing server connectivity...');
-    const debugInfo = await testServerConnectivity();
-    
-    // Use secure WebSocket (wss://) if the page is loaded over HTTPS
-    const protocol = window.location.protocol === 'https:' ? 'wss:' : 'ws:';
-    
-    // Get base path by extracting path from the script tag's src attribute
-    let basePath = '';
-    if (debugInfo && debugInfo.base_path) {
-        // Use base path from server if available
-        basePath = debugInfo.base_path;
-        logMessage(`Using server-provided base path: ${basePath}`);
-    } else {
-        const scriptTags = document.getElementsByTagName('script');
-        for (const script of scriptTags) {
-            if (script.src.includes('client.js')) {
-                const url = new URL(script.src);
-                basePath = url.pathname.replace('/assets/client.js', '');
-                break;
-            }
-        }
-    }
-    
-    // Try both with and without base path for WebSocket connection
-    let serverUrl = `${protocol}//${window.location.hostname}${window.location.port ? ':' + window.location.port : ''}${basePath}/ws`;
-    logMessage(`Attempting to connect to WebSocket at ${serverUrl}...`);
-    
-    // For Hugging Face Spaces, try the direct /ws path if the base path doesn't work
-    const fallbackUrl = `${protocol}//${window.location.hostname}${window.location.port ? ':' + window.location.port : ''}/ws`;
-    
-    try {
-        socket = new WebSocket(serverUrl);
-        setupWebSocketHandlers();
-        
-        // Set a timeout to try the fallback URL if the first one doesn't connect
-        setTimeout(() => {
-            if (socket.readyState !== WebSocket.OPEN && socket.readyState !== WebSocket.CONNECTING) {
-                logMessage(`Connection to ${serverUrl} failed. Trying fallback URL: ${fallbackUrl}`);
-                socket = new WebSocket(fallbackUrl);
-                setupWebSocketHandlers();
-            }
-        }, 3000);
-    } catch (error) {
-        logMessage(`Error connecting to WebSocket: ${error.message}`);
-        resetUI();
-    }
-}
-
-// Set up WebSocket event handlers
-function setupWebSocketHandlers() {
-    socket.onopen = () => {
-        logMessage('WebSocket connection established');
-        connectBtn.textContent = 'Disconnect';
-        startStreamBtn.disabled = false;
-        sceneSelect.disabled = false;
-    };
-    
-    socket.onmessage = (event) => {
-        const message = JSON.parse(event.data);
-        
-        switch (message.action) {
-            case 'welcome':
-                userId = message.userId;
-                logMessage(`Connected with user ID: ${userId}`);
-                
-                // Update scene options if server provides them
-                if (message.scenes && Array.isArray(message.scenes)) {
-                    sceneSelect.innerHTML = '';
-                    message.scenes.forEach(scene => {
-                        const option = document.createElement('option');
-                        option.value = scene;
-                        option.textContent = scene.charAt(0).toUpperCase() + scene.slice(1);
-                        sceneSelect.appendChild(option);
-                    });
-                }
-                break;
-                
-            case 'frame':
-                // Process incoming frame
-                processFrame(message);
-                break;
-                
-            case 'start_stream':
-                if (message.success) {
-                    isStreaming = true;
-                    startStreamBtn.disabled = true;
-                    stopStreamBtn.disabled = false;
-                    logMessage(`Streaming started: ${message.message}`);
-                    
-                    // Start FPS counter
-                    startFpsCounter();
-                } else {
-                    logMessage(`Error starting stream: ${message.error}`);
-                }
-                break;
-                
-            case 'stop_stream':
-                if (message.success) {
-                    isStreaming = false;
-                    startStreamBtn.disabled = false;
-                    stopStreamBtn.disabled = true;
-                    logMessage('Streaming stopped');
-                    
-                    // Stop FPS counter
-                    stopFpsCounter();
-                } else {
-                    logMessage(`Error stopping stream: ${message.error}`);
-                }
-                break;
-                
-            case 'pong':
-                // Server responded to ping
-                break;
-                
-            case 'change_scene':
-                if (message.success) {
-                    logMessage(`Scene changed to ${message.scene}`);
-                } else {
-                    logMessage(`Error changing scene: ${message.error}`);
-                }
-                break;
-                
-            default:
-                logMessage(`Received message: ${JSON.stringify(message)}`);
-        }
-    };
-    
-    socket.onclose = (event) => {
-        logMessage(`WebSocket connection closed (code: ${event.code}, reason: ${event.reason || 'none given'})`);
-        resetUI();
-    };
-    
-    socket.onerror = (error) => {
-        logMessage(`WebSocket error. This is often caused by CORS issues or the server being inaccessible.`);
-        console.error('WebSocket error:', error);
-        resetUI();
-    };
-}
-
-// Disconnect from WebSocket server
-function disconnectWebSocket() {
-    if (socket && socket.readyState === WebSocket.OPEN) {
-        // Stop streaming if active
-        if (isStreaming) {
-            sendStopStream();
-        }
-        
-        // Close the socket
-        socket.close();
-        logMessage('Disconnected from server');
-    }
-}
-
-// Start streaming frames
-function sendStartStream() {
-    if (socket && socket.readyState === WebSocket.OPEN) {
-        socket.send(JSON.stringify({
-            action: 'start_stream',
-            requestId: generateRequestId(),
-            fps: 16  // Default FPS
-        }));
-    }
-}
-
-// Stop streaming frames
-function sendStopStream() {
-    if (socket && socket.readyState === WebSocket.OPEN) {
-        socket.send(JSON.stringify({
-            action: 'stop_stream',
-            requestId: generateRequestId()
-        }));
-    }
-}
-
-// Send keyboard input to server
-function sendKeyboardInput(key, pressed) {
-    if (socket && socket.readyState === WebSocket.OPEN) {
-        socket.send(JSON.stringify({
-            action: 'keyboard_input',
-            requestId: generateRequestId(),
-            key: key,
-            pressed: pressed
-        }));
-    }
-}
-
-// Send mouse input to server
-function sendMouseInput(x, y) {
-    if (socket && socket.readyState === WebSocket.OPEN && isStreaming) {
-        socket.send(JSON.stringify({
-            action: 'mouse_input',
-            requestId: generateRequestId(),
-            x: x,
-            y: y
-        }));
-    }
-}
-
-// Change scene
-function sendChangeScene(scene) {
-    if (socket && socket.readyState === WebSocket.OPEN) {
-        socket.send(JSON.stringify({
-            action: 'change_scene',
-            requestId: generateRequestId(),
-            scene: scene
-        }));
-    }
-}
-
-// Process incoming frame
-function processFrame(message) {
-    // Update FPS calculation
-    const now = performance.now();
-    if (lastFrameTime > 0) {
-        frameCount++;
-    }
-    lastFrameTime = now;
-    
-    // Update the canvas with the new frame
-    if (message.frameData) {
-        gameCanvas.src = `data:image/jpeg;base64,${message.frameData}`;
-    }
-}
-
-// Generate a random request ID
-function generateRequestId() {
-    return Math.random().toString(36).substring(2, 15);
-}
-
-// Log message to the connection info panel
-function logMessage(message) {
-    const logEntry = document.createElement('div');
-    logEntry.className = 'log-entry';
-    
-    const timestamp = new Date().toLocaleTimeString();
-    logEntry.textContent = `[${timestamp}] ${message}`;
-    
-    connectionLog.appendChild(logEntry);
-    connectionLog.scrollTop = connectionLog.scrollHeight;
-    
-    // Limit number of log entries
-    while (connectionLog.children.length > 100) {
-        connectionLog.removeChild(connectionLog.firstChild);
-    }
-}
-
-// Start FPS counter updates
-function startFpsCounter() {
-    frameCount = 0;
-    lastFrameTime = 0;
-    
-    // Update FPS display every second
-    fpsUpdateInterval = setInterval(() => {
-        fpsCounter.textContent = `FPS: ${frameCount}`;
-        frameCount = 0;
-    }, 1000);
-}
-
-// Stop FPS counter updates
-function stopFpsCounter() {
-    if (fpsUpdateInterval) {
-        clearInterval(fpsUpdateInterval);
-        fpsUpdateInterval = null;
-    }
-    fpsCounter.textContent = 'FPS: 0';
-}
-
-// Reset UI to initial state
-function resetUI() {
-    connectBtn.textContent = 'Connect';
-    startStreamBtn.disabled = true;
-    stopStreamBtn.disabled = true;
-    sceneSelect.disabled = true;
-    
-    // Reset key indicators
-    for (const key in keyElements) {
-        keyElements[key].classList.remove('active');
-    }
-    
-    // Stop FPS counter
-    stopFpsCounter();
-    
-    // Reset streaming state
-    isStreaming = false;
-}
-
-// Event Listeners
-connectBtn.addEventListener('click', () => {
-    if (socket && socket.readyState === WebSocket.OPEN) {
-        disconnectWebSocket();
-    } else {
-        connectWebSocket();
-    }
-});
-
-startStreamBtn.addEventListener('click', sendStartStream);
-stopStreamBtn.addEventListener('click', sendStopStream);
-
-sceneSelect.addEventListener('change', () => {
-    sendChangeScene(sceneSelect.value);
-});
-
-// Keyboard event listeners
-document.addEventListener('keydown', (event) => {
-    const key = event.key.toLowerCase();
-    
-    // Map key to action
-    let action = keyToAction[key];
-    if (!action && key === ' ') {
-        action = keyToAction[' '];  // Handle spacebar
-    }
-    
-    if (action && !keyState[action]) {
-        keyState[action] = true;
-        
-        // Update visual indicator
-        const keyElement = keyElements[key] || 
-                          (key === ' ' ? keyElements['space'] : null) ||
-                          (key === 'shift' ? keyElements['shift'] : null);
-        
-        if (keyElement) {
-            keyElement.classList.add('active');
-        }
-        
-        // Send to server
-        sendKeyboardInput(action, true);
-    }
-    
-    // Prevent default actions for game controls
-    if (Object.keys(keyToAction).includes(key) || key === ' ') {
-        event.preventDefault();
-    }
-});
-
-document.addEventListener('keyup', (event) => {
-    const key = event.key.toLowerCase();
-    
-    // Map key to action
-    let action = keyToAction[key];
-    if (!action && key === ' ') {
-        action = keyToAction[' '];  // Handle spacebar
-    }
-    
-    if (action && keyState[action]) {
-        keyState[action] = false;
-        
-        // Update visual indicator
-        const keyElement = keyElements[key] || 
-                          (key === ' ' ? keyElements['space'] : null) ||
-                          (key === 'shift' ? keyElements['shift'] : null);
-        
-        if (keyElement) {
-            keyElement.classList.remove('active');
-        }
-        
-        // Send to server
-        sendKeyboardInput(action, false);
-    }
-});
-
-// Mouse capture functions
-function requestPointerLock() {
-    if (!mouseState.captured && pointerLockSupported) {
-        mouseTrackingArea.requestPointerLock = mouseTrackingArea.requestPointerLock ||
-                                            mouseTrackingArea.mozRequestPointerLock ||
-                                            mouseTrackingArea.webkitRequestPointerLock;
-        mouseTrackingArea.requestPointerLock();
-        logMessage('Mouse captured. Press ESC to release.');
-    }
-}
-
-function exitPointerLock() {
-    if (mouseState.captured) {
-        document.exitPointerLock = document.exitPointerLock ||
-                                 document.mozExitPointerLock ||
-                                 document.webkitExitPointerLock;
-        document.exitPointerLock();
-        logMessage('Mouse released.');
-    }
-}
-
-// Handle pointer lock change events
-document.addEventListener('pointerlockchange', pointerLockChangeHandler);
-document.addEventListener('mozpointerlockchange', pointerLockChangeHandler);
-document.addEventListener('webkitpointerlockchange', pointerLockChangeHandler);
-
-function pointerLockChangeHandler() {
-    if (document.pointerLockElement === mouseTrackingArea ||
-        document.mozPointerLockElement === mouseTrackingArea ||
-        document.webkitPointerLockElement === mouseTrackingArea) {
-        // Pointer is locked, enable mouse movement tracking
-        mouseState.captured = true;
-        document.addEventListener('mousemove', handleMouseMovement);
-    } else {
-        // Pointer is unlocked, disable mouse movement tracking
-        mouseState.captured = false;
-        document.removeEventListener('mousemove', handleMouseMovement);
-        // Reset mouse state
-        mouseState.x = 0;
-        mouseState.y = 0;
-        mousePosition.textContent = `Mouse: ${mouseState.x.toFixed(2)}, ${mouseState.y.toFixed(2)}`;
-        throttledSendMouseInput();
-    }
-}
-
-// Mouse tracking with pointer lock
-function handleMouseMovement(event) {
-    if (mouseState.captured) {
-        // Use movement for mouse look when captured
-        const sensitivity = 0.005; // Adjust sensitivity
-        mouseState.x += event.movementX * sensitivity;
-        mouseState.y -= event.movementY * sensitivity; // Invert Y for intuitive camera control
-        
-        // Clamp values
-        mouseState.x = Math.max(-1, Math.min(1, mouseState.x));
-        mouseState.y = Math.max(-1, Math.min(1, mouseState.y));
-        
-        // Update display
-        mousePosition.textContent = `Mouse: ${mouseState.x.toFixed(2)}, ${mouseState.y.toFixed(2)}`;
-        
-        // Send to server (throttled)
-        throttledSendMouseInput();
-    }
-}
-
-// Mouse click to capture
-mouseTrackingArea.addEventListener('click', () => {
-    if (!mouseState.captured && isStreaming) {
-        requestPointerLock();
-    }
-});
-
-// Standard mouse tracking for when pointer is not locked
-mouseTrackingArea.addEventListener('mousemove', (event) => {
-    if (!mouseState.captured) {
-        // Calculate normalized coordinates relative to the center of the tracking area
-        const rect = mouseTrackingArea.getBoundingClientRect();
-        const centerX = rect.width / 2;
-        const centerY = rect.height / 2;
-        
-        // Calculate relative position from center (-1 to 1)
-        const relX = (event.clientX - rect.left - centerX) / centerX;
-        const relY = (event.clientY - rect.top - centerY) / centerY;
-        
-        // Scale down for smoother movement (similar to conditions.py)
-        const scaleFactor = 0.05;
-        mouseState.x = relX * scaleFactor;
-        mouseState.y = -relY * scaleFactor;  // Invert Y for intuitive camera control
-        
-        // Update display
-        mousePosition.textContent = `Mouse: ${mouseState.x.toFixed(2)}, ${mouseState.y.toFixed(2)}`;
-        
-        // Send to server (throttled)
-        throttledSendMouseInput();
-    }
-});
-
-// Throttle mouse movement to avoid flooding the server
-const throttledSendMouseInput = (() => {
-    let lastSentTime = 0;
-    const interval = 50;  // milliseconds
-    
-    return () => {
-        const now = performance.now();
-        if (now - lastSentTime >= interval) {
-            sendMouseInput(mouseState.x, mouseState.y);
-            lastSentTime = now;
-        }
-    };
-})();
-
-// Toggle panel collapse/expand
-function togglePanel(panelId) {
-    const panel = document.getElementById(panelId);
-    const button = panel.querySelector('.toggle-button');
-    
-    if (panel.classList.contains('collapsed')) {
-        // Expand the panel
-        panel.classList.remove('collapsed');
-        button.textContent = '−'; // Minus sign
-    } else {
-        // Collapse the panel
-        panel.classList.add('collapsed');
-        button.textContent = '+'; // Plus sign
-    }
-}
-
-// Initialize the UI
-resetUI();
-
-// Make panel headers clickable
-document.querySelectorAll('.panel-header').forEach(header => {
-    header.addEventListener('click', () => {
-        const panelId = header.parentElement.id;
-        togglePanel(panelId);
-    });
-});

example/engine.py

@@ -1,332 +0,0 @@
-#!/usr/bin/env python3
-# -*- coding: utf-8 -*-
-
-"""
-MatrixGame Engine
-
-This module handles the core rendering and model inference for the MatrixGame project.
-"""
-
-import os
-import logging
-import argparse
-import time
-import torch
-import numpy as np
-from PIL import Image
-import cv2
-from einops import rearrange
-from diffusers.utils import load_image
-from diffusers.video_processor import VideoProcessor
-from typing import Dict, List, Tuple, Any, Optional, Union
-
-# MatrixGame specific imports
-from matrixgame.sample.pipeline_matrixgame import MatrixGameVideoPipeline
-from matrixgame.model_variants import get_dit
-from matrixgame.vae_variants import get_vae
-from matrixgame.encoder_variants import get_text_enc
-from matrixgame.model_variants.matrixgame_dit_src import MGVideoDiffusionTransformerI2V
-from matrixgame.sample.flow_matching_scheduler_matrixgame import FlowMatchDiscreteScheduler
-from teacache_forward import teacache_forward
-
-# Import utility functions
-from utils import (
-    visualize_controls, 
-    frame_to_jpeg, 
-    load_scene_frames,
-    logger
-)
-
-class MatrixGameEngine:
-    """
-    Core engine for MatrixGame model inference and frame generation.
-    """
-    def __init__(self, args: Optional[argparse.Namespace] = None):
-        """
-        Initialize the MatrixGame engine with configuration parameters.
-        
-        Args:
-            args: Optional parsed command line arguments for model configuration
-        """
-        # Set default parameters if args not provided
-        self.frame_width = getattr(args, 'frame_width', 640)
-        self.frame_height = getattr(args, 'frame_height', 360)
-        self.fps = getattr(args, 'fps', 16)
-        self.inference_steps = getattr(args, 'inference_steps', 20)
-        self.guidance_scale = getattr(args, 'guidance_scale', 6.0)
-        self.num_pre_frames = getattr(args, 'num_pre_frames', 3)
-        
-        # Initialize state
-        self.frame_count = 0
-        self.device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
-        self.weight_dtype = torch.bfloat16 if torch.cuda.is_available() else torch.float32
-        
-        # Model paths from environment or args
-        self.vae_path = os.environ.get("VAE_PATH", "./models/matrixgame/vae/")
-        self.dit_path = os.environ.get("DIT_PATH", "./models/matrixgame/dit/")
-        self.textenc_path = os.environ.get("TEXTENC_PATH", "./models/matrixgame")
-        
-        # Cache scene initial frames
-        self.scenes = {
-            'forest': load_scene_frames('forest', self.frame_width, self.frame_height),
-            'desert': load_scene_frames('desert', self.frame_width, self.frame_height),
-            'beach': load_scene_frames('beach', self.frame_width, self.frame_height),
-            'hills': load_scene_frames('hills', self.frame_width, self.frame_height),
-            'river': load_scene_frames('river', self.frame_width, self.frame_height),
-            'icy': load_scene_frames('icy', self.frame_width, self.frame_height),
-            'mushroom': load_scene_frames('mushroom', self.frame_width, self.frame_height),
-            'plain': load_scene_frames('plain', self.frame_width, self.frame_height)
-        }
-        
-        # Cache initial images for model input
-        self.scene_initial_images = {}
-        
-        # Initialize MatrixGame pipeline
-        self.model_loaded = False
-        if torch.cuda.is_available():
-            try:
-                self._init_models()
-                self.model_loaded = True
-                logger.info("MatrixGame models loaded successfully")
-            except Exception as e:
-                logger.error(f"Failed to initialize MatrixGame models: {str(e)}")
-                logger.info("Falling back to frame cycling mode")
-        else:
-            logger.warning("CUDA not available. Using frame cycling mode only.")
-    
-    def _init_models(self):
-        """Initialize MatrixGame models (VAE, text encoder, transformer)"""
-        # Initialize flow matching scheduler
-        self.scheduler = FlowMatchDiscreteScheduler(
-            shift=15.0,
-            reverse=True,
-            solver="euler"
-        )
-        
-        # Initialize VAE
-        try:
-            self.vae = get_vae("matrixgame", self.vae_path, self.weight_dtype)
-            self.vae.requires_grad_(False)
-            self.vae.eval()
-            self.vae.enable_tiling()
-            logger.info("VAE model loaded successfully")
-        except Exception as e:
-            logger.error(f"Error loading VAE model: {str(e)}")
-            raise
-        
-        # Initialize DIT (Transformer)
-        try:
-            dit = MGVideoDiffusionTransformerI2V.from_pretrained(self.dit_path)
-            dit.requires_grad_(False)
-            dit.eval()
-            logger.info("DIT model loaded successfully")
-        except Exception as e:
-            logger.error(f"Error loading DIT model: {str(e)}")
-            raise
-        
-        # Initialize text encoder
-        try:
-            self.text_enc = get_text_enc('matrixgame', self.textenc_path, weight_dtype=self.weight_dtype, i2v_type='refiner')
-            logger.info("Text encoder loaded successfully")
-        except Exception as e:
-            logger.error(f"Error loading text encoder: {str(e)}")
-            raise
-        
-        # Initialize pipeline
-        try:
-            self.pipeline = MatrixGameVideoPipeline(
-                vae=self.vae.vae,
-                text_encoder=self.text_enc,
-                transformer=dit,
-                scheduler=self.scheduler,
-            ).to(self.weight_dtype).to(self.device)
-            logger.info("Pipeline initialized successfully")
-        except Exception as e:
-            logger.error(f"Error initializing pipeline: {str(e)}")
-            raise
-        
-        # Configure teacache for the transformer
-        self.pipeline.transformer.__class__.enable_teacache = True
-        self.pipeline.transformer.__class__.cnt = 0
-        self.pipeline.transformer.__class__.num_steps = self.inference_steps
-        self.pipeline.transformer.__class__.accumulated_rel_l1_distance = 0
-        self.pipeline.transformer.__class__.rel_l1_thresh = 0.075
-        self.pipeline.transformer.__class__.previous_modulated_input = None
-        self.pipeline.transformer.__class__.previous_residual = None
-        self.pipeline.transformer.__class__.forward = teacache_forward
-        
-        # Preprocess initial images for all scenes
-        for scene_name, frames in self.scenes.items():
-            if frames:
-                # Use first frame as initial image
-                self.scene_initial_images[scene_name] = self._preprocess_image(frames[0])
-    
-    def _preprocess_image(self, image_array: np.ndarray) -> torch.Tensor:
-        """
-        Preprocess an image for the model.
-        
-        Args:
-            image_array: Input image as numpy array
-            
-        Returns:
-            torch.Tensor: Preprocessed image tensor
-        """
-        # Convert numpy array to PIL Image if needed
-        if isinstance(image_array, np.ndarray):
-            image = Image.fromarray(image_array)
-        else:
-            image = image_array
-            
-        # Preprocess for VAE
-        vae_scale_factor = 2 ** (len(self.vae.config.block_out_channels) - 1) if hasattr(self, 'vae') else 8
-        video_processor = VideoProcessor(vae_scale_factor=vae_scale_factor)
-        initial_image = video_processor.preprocess(image, height=self.frame_height, width=self.frame_width)
-        
-        # Add past frames for stability (use same frame repeated)
-        past_frames = initial_image.repeat(self.num_pre_frames, 1, 1, 1)
-        initial_image = torch.cat([initial_image, past_frames], dim=0)
-        
-        return initial_image
-    
-    def generate_frame(self, scene_name: str, keyboard_condition: Optional[List] = None, 
-                      mouse_condition: Optional[List] = None) -> bytes:
-        """
-        Generate the next frame based on current conditions using MatrixGame model.
-        
-        Args:
-            scene_name: Name of the current scene
-            keyboard_condition: Keyboard input state
-            mouse_condition: Mouse input state
-            
-        Returns:
-            bytes: JPEG bytes of the frame
-        """
-        # Check if model is loaded
-        if not self.model_loaded or not torch.cuda.is_available():
-            # Fall back to frame cycling for demo mode or if models failed to load
-            return self._fallback_frame(scene_name, keyboard_condition, mouse_condition)
-        else:
-            # Use MatrixGame model for frame generation
-            try:
-                # Get initial image for this scene
-                initial_image = self.scene_initial_images.get(scene_name)
-                if initial_image is None:
-                    # Use forest as default if we don't have an initial image for this scene
-                    initial_image = self.scene_initial_images.get('forest')
-                    if initial_image is None:
-                        # If we still don't have an initial image, fall back to frame cycling
-                        logger.error(f"No initial image available for scene {scene_name}")
-                        return self._fallback_frame(scene_name, keyboard_condition, mouse_condition)
-                
-                # Prepare input tensors (move to device and format correctly)
-                if keyboard_condition is None:
-                    keyboard_condition = [[0, 0, 0, 0, 0, 0]]
-                if mouse_condition is None:
-                    mouse_condition = [[0, 0]]
-                
-                # Convert conditions to tensors
-                keyboard_tensor = torch.tensor(keyboard_condition, dtype=torch.float32)
-                mouse_tensor = torch.tensor(mouse_condition, dtype=torch.float32)
-                
-                # Move to device and convert to correct dtype
-                keyboard_tensor = keyboard_tensor.to(self.weight_dtype).to(self.device)
-                mouse_tensor = mouse_tensor.to(self.weight_dtype).to(self.device)
-                
-                # Get the first frame from the scene for semantic conditioning
-                scene_frames = self.scenes.get(scene_name, self.scenes['forest'])
-                if not scene_frames:
-                    return self._fallback_frame(scene_name, keyboard_condition, mouse_condition)
-                
-                semantic_image = Image.fromarray(scene_frames[0])
-                
-                # Get PIL image version of the frame for visualization
-                for scene_frame in scene_frames:
-                    if isinstance(scene_frame, np.ndarray):
-                        semantic_image = Image.fromarray(scene_frame)
-                        break
-                
-                # Generate a single frame with the model
-                # Use fewer inference steps for interactive frame generation
-                with torch.no_grad():
-                    # Generate a short video (we'll just use the first frame)
-                    # We're using a short length (3 frames) for real-time performance
-                    video = self.pipeline(
-                        height=self.frame_height,
-                        width=self.frame_width,
-                        video_length=3,  # Generate a very short video for speed
-                        mouse_condition=mouse_tensor,
-                        keyboard_condition=keyboard_tensor,
-                        initial_image=initial_image,
-                        num_inference_steps=self.inference_steps,
-                        guidance_scale=self.guidance_scale,
-                        embedded_guidance_scale=None,
-                        data_type="video",
-                        vae_ver='884-16c-hy',
-                        enable_tiling=True,
-                        generator=torch.Generator(device=self.device).manual_seed(42),
-                        i2v_type='refiner',
-                        semantic_images=semantic_image
-                    ).videos[0]
-                
-                # Convert video tensor to numpy array (use first frame)
-                video_frame = video[0].permute(1, 2, 0).cpu().numpy()
-                video_frame = (video_frame * 255).astype(np.uint8)
-                frame = video_frame
-                
-                # Increment frame counter
-                self.frame_count += 1
-                
-            except Exception as e:
-                logger.error(f"Error generating frame with MatrixGame model: {str(e)}")
-                # Fall back to cycling demo frames if model generation fails
-                return self._fallback_frame(scene_name, keyboard_condition, mouse_condition)
-        
-        # Add visualization of input controls
-        frame = visualize_controls(
-            frame, keyboard_condition, mouse_condition, 
-            self.frame_width, self.frame_height
-        )
-        
-        # Convert frame to JPEG
-        return frame_to_jpeg(frame, self.frame_height, self.frame_width)
-    
-    def _fallback_frame(self, scene_name: str, keyboard_condition: Optional[List] = None, 
-                       mouse_condition: Optional[List] = None) -> bytes:
-        """
-        Generate a fallback frame when model generation fails.
-        
-        Args:
-            scene_name: Name of the current scene
-            keyboard_condition: Keyboard input state
-            mouse_condition: Mouse input state
-            
-        Returns:
-            bytes: JPEG bytes of the frame
-        """
-        scene_frames = self.scenes.get(scene_name, self.scenes['forest'])
-        frame_idx = self.frame_count % len(scene_frames)
-        frame = scene_frames[frame_idx].copy()
-        self.frame_count += 1
-        
-        # Add fallback mode indicator
-        cv2.putText(frame, "Fallback mode", 
-                  (10, self.frame_height - 20), 
-                  cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 1)
-        
-        # Add visualization of input controls
-        frame = visualize_controls(
-            frame, keyboard_condition, mouse_condition, 
-            self.frame_width, self.frame_height
-        )
-        
-        # Convert frame to JPEG
-        return frame_to_jpeg(frame, self.frame_height, self.frame_width)
-    
-    def get_valid_scenes(self) -> List[str]:
-        """
-        Get a list of valid scene names.
-        
-        Returns:
-            List[str]: List of valid scene names
-        """
-        return list(self.scenes.keys())

example/index.html

@@ -1,329 +0,0 @@
-<!DOCTYPE html>
-<html lang="en">
-<head>
-    <meta charset="UTF-8">
-    <meta name="viewport" content="width=device-width, initial-scale=1.0">
-    <title>MatrixGame Client</title>
-    <style>
-        body {
-            font-family: Arial, sans-serif;
-            margin: 0;
-            padding: 0;
-            background-color: #121212;
-            color: #e0e0e0;
-            display: flex;
-            flex-direction: column;
-            align-items: center;
-            user-select: none; /* Disable text selection */
-            -webkit-user-select: none;
-            -moz-user-select: none;
-            -ms-user-select: none;
-            overflow-x: hidden;
-        }
-        
-        .container {
-            width: 100%;
-            max-width: 100%;
-            display: flex;
-            flex-direction: column;
-            align-items: center;
-        }
-        
-        .game-area {
-            display: flex;
-            flex-direction: column;
-            align-items: center;
-            width: 100%;
-            max-height: 85vh;
-            margin: 0;
-            position: relative;
-        }
-        
-        #mouse-tracking-area {
-            position: relative;
-            width: 100%;
-            height: auto;
-            cursor: pointer; /* Show cursor as pointer to encourage clicks */
-            display: flex;
-            justify-content: center;
-            align-items: center;
-            max-height: 85vh;
-        }
-        
-        #game-canvas {
-            width: 100%;
-            height: auto;
-            max-height: 85vh;
-            object-fit: contain;
-            background-color: #000;
-            pointer-events: none; /* Prevent drag on the image */
-            -webkit-user-drag: none;
-            -khtml-user-drag: none;
-            -moz-user-drag: none;
-            -o-user-drag: none;
-            user-drag: none;
-        }
-        
-        .controls {
-            display: flex;
-            justify-content: space-between;
-            width: 100%;
-            max-width: 1200px;
-            padding: 10px;
-            background-color: rgba(0, 0, 0, 0.5);
-            position: absolute;
-            bottom: 0;
-            z-index: 10;
-            box-sizing: border-box;
-        }
-        
-        .panels-container {
-            display: flex;
-            width: 100%;
-            max-width: 1200px;
-            margin: 10px auto;
-            gap: 10px;
-        }
-        
-        .panel {
-            flex: 1;
-            background-color: #1E1E1E;
-            border-radius: 5px;
-            overflow: hidden;
-            box-shadow: 0 2px 5px rgba(0, 0, 0, 0.2);
-            transition: height 0.3s ease;
-        }
-        
-        .panel-header {
-            background-color: #272727;
-            padding: 10px 15px;
-            display: flex;
-            justify-content: space-between;
-            align-items: center;
-            cursor: pointer;
-        }
-        
-        .panel-title {
-            font-weight: bold;
-            color: #4CAF50;
-        }
-        
-        .toggle-button {
-            background: none;
-            border: none;
-            color: #e0e0e0;
-            font-size: 18px;
-            cursor: pointer;
-        }
-        
-        .toggle-button:focus {
-            outline: none;
-        }
-        
-        .panel-content {
-            padding: 15px;
-            max-height: 300px;
-            overflow-y: auto;
-            transition: all 0.3s ease;
-        }
-        
-        .collapsed .panel-content {
-            max-height: 0;
-            padding-top: 0;
-            padding-bottom: 0;
-            overflow: hidden;
-        }
-        
-        button {
-            background-color: #4CAF50;
-            color: white;
-            border: none;
-            padding: 10px 15px;
-            text-align: center;
-            text-decoration: none;
-            display: inline-block;
-            font-size: 14px;
-            border-radius: 5px;
-            cursor: pointer;
-            margin: 5px;
-            transition: background-color 0.3s;
-        }
-        
-        button:hover {
-            background-color: #45a049;
-        }
-        
-        button:disabled {
-            background-color: #cccccc;
-            cursor: not-allowed;
-        }
-        
-        select {
-            padding: 10px;
-            border-radius: 5px;
-            background-color: #2A2A2A;
-            color: #e0e0e0;
-            border: 1px solid #4CAF50;
-        }
-        
-        .status {
-            margin-top: 10px;
-            color: #4CAF50;
-        }
-        
-        .key-indicators {
-            display: flex;
-            justify-content: center;
-            margin-top: 15px;
-        }
-        
-        .key {
-            width: 40px;
-            height: 40px;
-            margin: 0 5px;
-            background-color: #2A2A2A;
-            border: 1px solid #444;
-            border-radius: 5px;
-            display: flex;
-            justify-content: center;
-            align-items: center;
-            font-weight: bold;
-            transition: background-color 0.2s;
-        }
-        
-        .key.active {
-            background-color: #4CAF50;
-            color: white;
-        }
-        
-        .key-row {
-            display: flex;
-            justify-content: center;
-            margin: 5px 0;
-        }
-        
-        .spacebar {
-            width: 150px;
-        }
-        
-        .connection-info {
-            font-family: monospace;
-            height: 100%;
-            overflow-y: auto;
-        }
-        
-        .log-entry {
-            margin: 5px 0;
-            padding: 3px;
-            border-bottom: 1px solid #333;
-        }
-        
-        .fps-counter {
-            position: absolute;
-            top: 10px;
-            right: 10px;
-            background-color: rgba(0,0,0,0.5);
-            color: #4CAF50;
-            padding: 5px;
-            border-radius: 3px;
-            font-family: monospace;
-            z-index: 20;
-        }
-        
-        
-        #mouse-position {
-            position: absolute;
-            top: 10px;
-            left: 10px;
-            background-color: rgba(0,0,0,0.5);
-            color: #4CAF50;
-            padding: 5px;
-            border-radius: 3px;
-            font-family: monospace;
-            z-index: 20;
-        }
-        
-        @media (max-width: 768px) {
-            .panels-container {
-                flex-direction: column;
-            }
-        }
-    </style>
-</head>
-<body>
-    <div class="container">
-        <div class="game-area">
-            <div id="mouse-tracking-area">
-                <img id="game-canvas" src="data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mNk+A8AAQUBAScY42YAAAAASUVORK5CYII=" alt="Game Frame">
-                <div id="mouse-position">Mouse: 0.00, 0.00</div>
-                <div class="fps-counter" id="fps-counter">FPS: 0</div>
-            </div>
-            
-            <div class="controls">
-                <button id="connect-btn">Connect</button>
-                <button id="start-stream-btn" disabled>Start Stream</button>
-                <button id="stop-stream-btn" disabled>Stop Stream</button>
-                <select id="scene-select" disabled>
-                    <option value="forest">Forest</option>
-                    <option value="desert">Desert</option>
-                    <option value="beach">Beach</option>
-                    <option value="hills">Hills</option>
-                    <option value="river">River</option>
-                    <option value="icy">Icy</option>
-                    <option value="mushroom">Mushroom</option>
-                    <option value="plain">Plain</option>
-                </select>
-            </div>
-        </div>
-        
-        <div class="panels-container">
-            <!-- Controls Panel -->
-            <div class="panel" id="controls-panel">
-                <div class="panel-header" onclick="togglePanel('controls-panel')">
-                    <div class="panel-title">Keyboard Controls</div>
-                    <button class="toggle-button">−</button>
-                </div>
-                <div class="panel-content">
-                    <div class="key-indicators">
-                        <div class="key-row">
-                            <div id="key-w" class="key">W</div>
-                        </div>
-                        <div class="key-row">
-                            <div id="key-a" class="key">A</div>
-                            <div id="key-s" class="key">S</div>
-                            <div id="key-d" class="key">D</div>
-                        </div>
-                        <div class="key-row">
-                            <div id="key-space" class="key spacebar">SPACE</div>
-                        </div>
-                        <div class="key-row">
-                            <div id="key-shift" class="key">SHIFT</div>
-                        </div>
-                    </div>
-                    <p class="status">
-                        W or ↑ = Forward, S or ↓ = Back, A or ← = Left, D or → = Right<br>
-                        Space = Jump, Shift = Attack<br>
-                        Click on game view to capture mouse (ESC to release)<br>
-                        Mouse = Look around
-                    </p>
-                </div>
-            </div>
-            
-            <!-- Connection Log Panel -->
-            <div class="panel" id="log-panel">
-                <div class="panel-header" onclick="togglePanel('log-panel')">
-                    <div class="panel-title">Connection Log</div>
-                    <button class="toggle-button">−</button>
-                </div>
-                <div class="panel-content">
-                    <div class="connection-info" id="connection-log">
-                        <div class="log-entry">Waiting to connect...</div>
-                    </div>
-                </div>
-            </div>
-        </div>
-    </div>
-    
-    <script src="./assets/client.js"></script>
-</body>
-</html>

example/requirements.txt

@@ -1,23 +0,0 @@
-diffusers==0.32.2
-einops==0.8.1
-
-#flash_attn==2.7.4.post1
-flash-attn @ https://github.com/Dao-AILab/flash-attention/releases/download/v2.7.4.post1/flash_attn-2.7.4.post1+cu12torch2.4cxx11abiFALSE-cp310-cp310-linux_x86_64.whl
-
-ftfy==6.3.1
-imageio==2.34.0
-numpy==1.24.4
-opencv_python==4.9.0.80
-opencv_python_headless==4.9.0.80
-packaging==25.0
-peft==0.14.0
-Pillow==11.2.1
-regex==2024.11.6
-safetensors==0.5.3
-torch==2.5.1
-torchvision==0.20.1 
-torchaudio==2.5.1
-transformers==4.47.1
-aiohttp==3.9.3
-jinja2==3.1.3
-python-multipart==0.0.6

example/server.py

@@ -1,649 +0,0 @@
-#!/usr/bin/env python3
-# -*- coding: utf-8 -*-
-
-"""
-MatrixGame Websocket Gaming Server
-
-This script implements a websocket server for the MatrixGame project,
-allowing real-time streaming of game frames based on player inputs.
-"""
-
-import asyncio
-import json
-import logging
-import os
-import pathlib
-import time
-import uuid
-import base64
-import argparse
-from typing import Dict, List, Any, Optional
-from aiohttp import web, WSMsgType
-
-# Import the game engine
-from engine import MatrixGameEngine
-from utils import logger, parse_model_args, setup_gpu_environment
-
-class GameSession:
-    """
-    Represents a user's gaming session.
-    Each WebSocket connection gets its own session with separate queues.
-    """
-    def __init__(self, user_id: str, ws: web.WebSocketResponse, game_manager):
-        self.user_id = user_id
-        self.ws = ws
-        self.game_manager = game_manager
-        
-        # Create action queue for this user session
-        self.action_queue = asyncio.Queue()
-        
-        # Session creation time
-        self.created_at = time.time()
-        self.last_activity = time.time()
-        
-        # Game state
-        self.current_scene = "forest"  # Default scene
-        self.is_streaming = False
-        self.stream_task = None
-        
-        # Current input state
-        self.keyboard_state = [0, 0, 0, 0, 0, 0]  # forward, back, left, right, jump, attack
-        self.mouse_state = [0, 0]  # x, y
-        
-        self.background_tasks = []
-        
-    async def start(self):
-        """Start all the queue processors for this session"""
-        self.background_tasks = [
-            asyncio.create_task(self._process_action_queue()),
-        ]
-        logger.info(f"Started game session for user {self.user_id}")
-        
-    async def stop(self):
-        """Stop all background tasks for this session"""
-        # Stop streaming if active
-        if self.is_streaming and self.stream_task:
-            self.is_streaming = False
-            self.stream_task.cancel()
-            try:
-                await self.stream_task
-            except asyncio.CancelledError:
-                pass
-        
-        # Cancel other background tasks
-        for task in self.background_tasks:
-            task.cancel()
-        
-        try:
-            # Wait for tasks to complete cancellation
-            await asyncio.gather(*self.background_tasks, return_exceptions=True)
-        except asyncio.CancelledError:
-            pass
-        
-        logger.info(f"Stopped game session for user {self.user_id}")
-        
-    async def _process_action_queue(self):
-        """Process game actions from the queue"""
-        while True:
-            data = await self.action_queue.get()
-            try:
-                action_type = data.get('action')
-                
-                if action_type == 'start_stream':
-                    result = await self._handle_start_stream(data)
-                elif action_type == 'stop_stream':
-                    result = await self._handle_stop_stream(data)
-                elif action_type == 'keyboard_input':
-                    result = await self._handle_keyboard_input(data)
-                elif action_type == 'mouse_input':
-                    result = await self._handle_mouse_input(data)
-                elif action_type == 'change_scene':
-                    result = await self._handle_scene_change(data)
-                else:
-                    result = {
-                        'action': action_type,
-                        'requestId': data.get('requestId'),
-                        'success': False,
-                        'error': f'Unknown action: {action_type}'
-                    }
-                
-                # Send response back to the client
-                await self.ws.send_json(result)
-                
-                # Update last activity time
-                self.last_activity = time.time()
-                
-            except Exception as e:
-                logger.error(f"Error processing action for user {self.user_id}: {str(e)}")
-                try:
-                    await self.ws.send_json({
-                        'action': data.get('action'),
-                        'requestId': data.get('requestId', 'unknown'),
-                        'success': False,
-                        'error': f'Error processing action: {str(e)}'
-                    })
-                except Exception as send_error:
-                    logger.error(f"Error sending error response: {send_error}")
-            finally:
-                self.action_queue.task_done()
-    
-    async def _handle_start_stream(self, data: Dict) -> Dict:
-        """Handle request to start streaming frames"""
-        if self.is_streaming:
-            return {
-                'action': 'start_stream',
-                'requestId': data.get('requestId'),
-                'success': False,
-                'error': 'Stream already active'
-            }
-        
-        fps = data.get('fps', 16)
-        self.is_streaming = True
-        self.stream_task = asyncio.create_task(self._stream_frames(fps))
-        
-        return {
-            'action': 'start_stream',
-            'requestId': data.get('requestId'),
-            'success': True,
-            'message': f'Streaming started at {fps} FPS'
-        }
-    
-    async def _handle_stop_stream(self, data: Dict) -> Dict:
-        """Handle request to stop streaming frames"""
-        if not self.is_streaming:
-            return {
-                'action': 'stop_stream',
-                'requestId': data.get('requestId'),
-                'success': False,
-                'error': 'No active stream to stop'
-            }
-        
-        self.is_streaming = False
-        if self.stream_task:
-            self.stream_task.cancel()
-            try:
-                await self.stream_task
-            except asyncio.CancelledError:
-                pass
-            self.stream_task = None
-        
-        return {
-            'action': 'stop_stream',
-            'requestId': data.get('requestId'),
-            'success': True,
-            'message': 'Streaming stopped'
-        }
-    
-    async def _handle_keyboard_input(self, data: Dict) -> Dict:
-        """Handle keyboard input from client"""
-        key = data.get('key', '')
-        pressed = data.get('pressed', False)
-        
-        # Map key to keyboard state index
-        key_map = {
-            'w': 0, 'forward': 0,
-            's': 1, 'back': 1, 'backward': 1,
-            'a': 2, 'left': 2,
-            'd': 3, 'right': 3,
-            'space': 4, 'jump': 4,
-            'shift': 5, 'attack': 5, 'ctrl': 5
-        }
-        
-        if key.lower() in key_map:
-            key_idx = key_map[key.lower()]
-            self.keyboard_state[key_idx] = 1 if pressed else 0
-        
-        return {
-            'action': 'keyboard_input',
-            'requestId': data.get('requestId'),
-            'success': True,
-            'keyboardState': self.keyboard_state
-        }
-    
-    async def _handle_mouse_input(self, data: Dict) -> Dict:
-        """Handle mouse movement/input from client"""
-        mouse_x = data.get('x', 0)
-        mouse_y = data.get('y', 0)
-        
-        # Update mouse state, normalize values between -1 and 1
-        self.mouse_state = [float(mouse_x), float(mouse_y)]
-        
-        return {
-            'action': 'mouse_input',
-            'requestId': data.get('requestId'),
-            'success': True,
-            'mouseState': self.mouse_state
-        }
-    
-    async def _handle_scene_change(self, data: Dict) -> Dict:
-        """Handle scene change requests"""
-        scene_name = data.get('scene', 'forest')
-        valid_scenes = self.game_manager.valid_scenes
-        
-        if scene_name not in valid_scenes:
-            return {
-                'action': 'change_scene',
-                'requestId': data.get('requestId'),
-                'success': False,
-                'error': f'Invalid scene: {scene_name}. Valid scenes are: {", ".join(valid_scenes)}'
-            }
-            
-        self.current_scene = scene_name
-        
-        return {
-            'action': 'change_scene',
-            'requestId': data.get('requestId'),
-            'success': True,
-            'scene': scene_name
-        }
-    
-    async def _stream_frames(self, fps: int):
-        """Stream frames to the client at the specified FPS"""
-        frame_interval = 1.0 / fps  # Time between frames in seconds
-        
-        try:
-            while self.is_streaming:
-                start_time = time.time()
-                
-                # Generate frame based on current keyboard and mouse state
-                keyboard_condition = [self.keyboard_state]
-                mouse_condition = [self.mouse_state]
-                
-                # Use the engine to generate the next frame
-                frame_bytes = self.game_manager.engine.generate_frame(
-                    self.current_scene, keyboard_condition, mouse_condition
-                )
-                
-                # Encode as base64 for sending in JSON
-                frame_base64 = base64.b64encode(frame_bytes).decode('utf-8')
-                
-                # Send frame to client
-                await self.ws.send_json({
-                    'action': 'frame',
-                    'frameData': frame_base64,
-                    'timestamp': time.time()
-                })
-                
-                # Calculate sleep time to maintain FPS
-                elapsed = time.time() - start_time
-                sleep_time = max(0, frame_interval - elapsed)
-                await asyncio.sleep(sleep_time)
-                
-        except asyncio.CancelledError:
-            logger.info(f"Frame streaming cancelled for user {self.user_id}")
-        except Exception as e:
-            logger.error(f"Error in frame streaming for user {self.user_id}: {str(e)}")
-            if self.ws.closed:
-                logger.info(f"WebSocket closed for user {self.user_id}")
-                return
-            
-            # Notify client of error
-            try:
-                await self.ws.send_json({
-                    'action': 'frame_error',
-                    'error': f'Streaming error: {str(e)}'
-                })
-            except:
-                pass
-            
-            # Stop streaming
-            self.is_streaming = False
-
-class GameManager:
-    """
-    Manages all active gaming sessions and shared resources.
-    """
-    def __init__(self, args: argparse.Namespace):
-        self.sessions = {}
-        self.session_lock = asyncio.Lock()
-        
-        # Initialize game engine
-        self.engine = MatrixGameEngine(args)
-        
-        # Load valid scenes from engine
-        self.valid_scenes = self.engine.get_valid_scenes()
-        
-    async def create_session(self, user_id: str, ws: web.WebSocketResponse) -> GameSession:
-        """Create a new game session"""
-        async with self.session_lock:
-            # Create a new session for this user
-            session = GameSession(user_id, ws, self)
-            await session.start()
-            self.sessions[user_id] = session
-            return session
-    
-    async def delete_session(self, user_id: str) -> None:
-        """Delete a game session and clean up resources"""
-        async with self.session_lock:
-            if user_id in self.sessions:
-                session = self.sessions[user_id]
-                await session.stop()
-                del self.sessions[user_id]
-                logger.info(f"Deleted game session for user {user_id}")
-    
-    def get_session(self, user_id: str) -> Optional[GameSession]:
-        """Get a game session if it exists"""
-        return self.sessions.get(user_id)
-    
-    async def close_all_sessions(self) -> None:
-        """Close all active sessions (used during shutdown)"""
-        async with self.session_lock:
-            for user_id, session in list(self.sessions.items()):
-                await session.stop()
-            self.sessions.clear()
-            logger.info("Closed all active game sessions")
-    
-    @property
-    def session_count(self) -> int:
-        """Get the number of active sessions"""
-        return len(self.sessions)
-    
-    def get_session_stats(self) -> Dict:
-        """Get statistics about active sessions"""
-        stats = {
-            'total_sessions': len(self.sessions),
-            'active_scenes': {},
-            'streaming_sessions': 0
-        }
-        
-        # Count sessions by scene and streaming status
-        for session in self.sessions.values():
-            scene = session.current_scene
-            stats['active_scenes'][scene] = stats['active_scenes'].get(scene, 0) + 1
-            if session.is_streaming:
-                stats['streaming_sessions'] += 1
-            
-        return stats
-
-# Create global game manager
-game_manager = None
-
-async def status_handler(request: web.Request) -> web.Response:
-    """Handler for API status endpoint"""
-    # Get session statistics
-    session_stats = game_manager.get_session_stats()
-    
-    return web.json_response({
-        'product': 'MatrixGame WebSocket Server',
-        'version': '1.0.0',
-        'active_sessions': session_stats,
-        'available_scenes': game_manager.valid_scenes
-    })
-
-async def root_handler(request: web.Request) -> web.Response:
-    """Handler for serving the client at the root path"""
-    client_path = pathlib.Path(__file__).parent / 'client' / 'index.html'
-    
-    with open(client_path, 'r') as file:
-        html_content = file.read()
-        
-    return web.Response(text=html_content, content_type='text/html')
-
-async def websocket_handler(request: web.Request) -> web.WebSocketResponse:
-    """Handle WebSocket connections with robust error handling"""
-    logger.info(f"WebSocket connection attempt - PATH: {request.path}, QUERY: {request.query_string}")
-    
-    # Log request headers at debug level only (could contain sensitive information)
-    logger.debug(f"WebSocket request headers: {dict(request.headers)}")
-    
-    # Prepare a WebSocket response with appropriate settings
-    ws = web.WebSocketResponse(
-        max_msg_size=1024*1024*10,  # 10MB max message size
-        timeout=60.0,
-        heartbeat=30.0  # Add heartbeat to keep connection alive
-    )
-    
-    # Check if WebSocket protocol is supported
-    if not ws.can_prepare(request):
-        logger.error("Cannot prepare WebSocket: WebSocket protocol not supported")
-        return web.Response(status=400, text="WebSocket protocol not supported")
-    
-    try:
-        logger.info("Preparing WebSocket connection...")
-        await ws.prepare(request)
-        
-        # Generate a unique user ID for this connection
-        user_id = str(uuid.uuid4())
-        
-        # Get client IP address
-        peername = request.transport.get_extra_info('peername')
-        if peername is not None:
-            client_ip = peername[0]
-        else:
-            client_ip = request.headers.get('X-Forwarded-For', 'unknown').split(',')[0].strip()
-        
-        # Log connection success
-        logger.info(f"Client {user_id} connecting from IP: {client_ip} - WebSocket connection established")
-        
-        # Mark that the session is established
-        is_session_created = False
-        
-        try:
-            # Store the user ID in the websocket for easy access
-            ws.user_id = user_id
-            
-            # Create a new session for this user
-            logger.info(f"Creating game session for user {user_id}")
-            user_session = await game_manager.create_session(user_id, ws)
-            is_session_created = True
-            logger.info(f"Game session created for user {user_id}")
-        except Exception as session_error:
-            logger.error(f"Error creating game session: {str(session_error)}", exc_info=True)
-            if not ws.closed:
-                await ws.close(code=1011, message=f"Server error: {str(session_error)}".encode())
-            if is_session_created:
-                await game_manager.delete_session(user_id)
-            return ws
-    except Exception as e:
-        logger.error(f"Error establishing WebSocket connection: {str(e)}", exc_info=True)
-        if not ws.closed and ws.prepared:
-            await ws.close(code=1011, message=f"Server error: {str(e)}".encode())
-        return ws
-    
-    # Send initial welcome message
-    try:
-        await ws.send_json({
-            'action': 'welcome',
-            'userId': user_id,
-            'message': 'Welcome to the MatrixGame WebSocket server!',
-            'scenes': game_manager.valid_scenes
-        })
-        logger.info(f"Sent welcome message to user {user_id}")
-    except Exception as welcome_error:
-        logger.error(f"Error sending welcome message: {str(welcome_error)}")
-        if not ws.closed:
-            await ws.close(code=1011, message=b"Failed to send welcome message")
-        await game_manager.delete_session(user_id)
-        return ws
-
-    try:
-        async for msg in ws:
-            if msg.type == WSMsgType.TEXT:
-                try:
-                    data = json.loads(msg.data)
-                    action = data.get('action')
-                    
-                    logger.debug(f"Received {action} message from user {user_id}")
-                    
-                    if action == 'ping':
-                        # Respond to ping immediately
-                        await ws.send_json({
-                            'action': 'pong',
-                            'requestId': data.get('requestId'),
-                            'timestamp': time.time()
-                        })
-                    else:
-                        # Route game actions to the session's action queue
-                        await user_session.action_queue.put(data)
-                        
-                except json.JSONDecodeError:
-                    logger.error(f"Invalid JSON from user {user_id}: {msg.data}")
-                    if not ws.closed:
-                        await ws.send_json({
-                            'error': 'Invalid JSON message',
-                            'success': False
-                        })
-                except Exception as e:
-                    logger.error(f"Error processing WebSocket message for user {user_id}: {str(e)}")
-                    if not ws.closed:
-                        await ws.send_json({
-                            'action': data.get('action') if 'data' in locals() else 'unknown',
-                            'success': False,
-                            'error': f'Error processing message: {str(e)}'
-                        })
-                    
-            elif msg.type == WSMsgType.ERROR:
-                logger.error(f"WebSocket error for user {user_id}: {ws.exception()}")
-                break
-                
-            elif msg.type == WSMsgType.CLOSE:
-                logger.info(f"WebSocket close received for user {user_id} (code: {msg.data}, message: {msg.extra})")
-                break
-                
-            elif msg.type == WSMsgType.CLOSING:
-                logger.info(f"WebSocket closing for user {user_id}")
-                break
-                
-            elif msg.type == WSMsgType.CLOSED:
-                logger.info(f"WebSocket already closed for user {user_id}")
-                break
-                
-    except Exception as ws_error:
-        logger.error(f"Unexpected WebSocket error for user {user_id}: {str(ws_error)}", exc_info=True)
-    finally:
-        # Cleanup session
-        try:
-            logger.info(f"Cleaning up session for user {user_id}")
-            await game_manager.delete_session(user_id)
-            logger.info(f"Connection closed for user {user_id}")
-        except Exception as cleanup_error:
-            logger.error(f"Error during session cleanup for user {user_id}: {str(cleanup_error)}")
-    
-    return ws
-
-async def init_app(args, base_path="") -> web.Application:
-    """Initialize the web application"""
-    global game_manager
-    
-    # Initialize game manager with command line args
-    game_manager = GameManager(args)
-    
-    app = web.Application(
-        client_max_size=1024**2*10  # 10MB max size
-    )
-    
-    # Add cleanup logic
-    async def cleanup(app):
-        logger.info("Shutting down server, closing all sessions...")
-        await game_manager.close_all_sessions()
-    
-    app.on_shutdown.append(cleanup)
-    
-    # Add routes with CORS headers for WebSockets
-    # Configure CORS for all routes
-    @web.middleware
-    async def cors_middleware(request, handler):
-        if request.method == 'OPTIONS':
-            # Handle preflight requests
-            resp = web.Response()
-            resp.headers['Access-Control-Allow-Origin'] = '*'
-            resp.headers['Access-Control-Allow-Methods'] = 'GET, POST, OPTIONS'
-            resp.headers['Access-Control-Allow-Headers'] = 'Content-Type, X-Requested-With'
-            return resp
-            
-        # Normal request, call the handler
-        resp = await handler(request)
-        
-        # Add CORS headers to the response
-        resp.headers['Access-Control-Allow-Origin'] = '*'
-        resp.headers['Access-Control-Allow-Methods'] = 'GET, POST, OPTIONS'
-        resp.headers['Access-Control-Allow-Headers'] = 'Content-Type, X-Requested-With'
-        return resp
-    
-    app.middlewares.append(cors_middleware)
-    
-    # Add a debug endpoint to help diagnose WebSocket issues
-    async def debug_handler(request):
-        client_ip = request.remote
-        headers = dict(request.headers)
-        server_host = request.host
-        
-        debug_info = {
-            "client_ip": client_ip,
-            "server_host": server_host,
-            "headers": headers,
-            "request_path": request.path,
-            "server_time": time.time(),
-            "base_path": base_path,
-            "websocket_route": f"{base_path}/ws",
-            "all_routes": [route.name for route in app.router.routes() if route.name],
-            "server_info": {
-                "active_sessions": game_manager.session_count,
-                "available_scenes": game_manager.valid_scenes
-            }
-        }
-        
-        return web.json_response(debug_info)
-    
-    # Set up routes with the base_path
-    # Add multiple WebSocket routes to ensure compatibility
-    logger.info(f"Setting up WebSocket route at {base_path}/ws")
-    app.router.add_get(f'{base_path}/ws', websocket_handler, name='ws_handler')
-    
-    # Also add WebSocket route at the root for Hugging Face compatibility
-    if base_path:
-        logger.info(f"Adding additional WebSocket route at /ws")
-        app.router.add_get('/ws', websocket_handler, name='ws_root_handler')
-    
-    # Add routes for API and debug endpoints
-    app.router.add_get(f'{base_path}/api/status', status_handler, name='status_handler')
-    app.router.add_get(f'{base_path}/api/debug', debug_handler, name='debug_handler')
-    
-    # Serve the client at both the base path and root path for compatibility
-    app.router.add_get(f'{base_path}/', root_handler, name='root_handler')
-    
-    # Always serve at the root path for Hugging Face Spaces compatibility
-    if base_path:
-        app.router.add_get('/', root_handler, name='root_handler_no_base')
-    
-    # Set up static file serving for the client assets
-    app.router.add_static(f'{base_path}/assets', pathlib.Path(__file__).parent / 'client', name='static_handler')
-    
-    # Add static file serving at root for compatibility
-    if base_path:
-        app.router.add_static('/assets', pathlib.Path(__file__).parent / 'client', name='static_handler_no_base')
-    
-    return app
-
-def parse_args() -> argparse.Namespace:
-    """Parse server-specific command line arguments"""
-    parser = argparse.ArgumentParser(description="MatrixGame WebSocket Server")
-    parser.add_argument("--host", type=str, default="0.0.0.0", help="Host IP to bind to")
-    parser.add_argument("--port", type=int, default=8080, help="Port to listen on")
-    parser.add_argument("--path", type=str, default="", help="Base path for the server (for proxy setups)")
-    
-    # Parse server args first
-    server_args, remaining_args = parser.parse_known_args()
-    
-    # Parse model args and combine
-    model_args = parse_model_args()
-    
-    # Combine all args
-    combined_args = argparse.Namespace(**vars(server_args), **vars(model_args))
-    
-    return combined_args
-
-if __name__ == '__main__':
-    # Configure GPU environment
-    setup_gpu_environment()
-    
-    # Parse command line arguments
-    args = parse_args()
-    
-    # Initialize app
-    loop = asyncio.get_event_loop()
-    app = loop.run_until_complete(init_app(args, base_path=args.path))
-    
-    # Start server
-    logger.info(f"Starting MatrixGame WebSocket Server at {args.host}:{args.port}")
-    web.run_app(app, host=args.host, port=args.port)

example/utils.py

@@ -1,202 +0,0 @@
-#!/usr/bin/env python3
-# -*- coding: utf-8 -*-
-
-"""
-MatrixGame Utility Functions
-
-This module contains helper functions and utilities for the MatrixGame project.
-"""
-
-import os
-import logging
-import argparse
-import torch
-import numpy as np
-import cv2
-from PIL import Image
-from typing import Dict, List, Tuple, Any, Optional, Union
-
-# Configure logging
-logging.basicConfig(
-    level=logging.INFO,
-    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
-)
-logger = logging.getLogger(__name__)
-
-def setup_gpu_environment():
-    """
-    Configure the GPU environment and log GPU information.
-    
-    Returns:
-        bool: True if CUDA is available, False otherwise
-    """
-    # Set CUDA memory allocation environment variable for better performance
-    os.environ.setdefault("PYTORCH_CUDA_ALLOC_CONF", "expandable_segments:True")
-    
-    # Check if CUDA is available and log information
-    if torch.cuda.is_available():
-        gpu_count = torch.cuda.device_count()
-        gpu_info = []
-        
-        for i in range(gpu_count):
-            gpu_name = torch.cuda.get_device_name(i)
-            gpu_memory = torch.cuda.get_device_properties(i).total_memory / (1024**3)  # Convert to GB
-            gpu_info.append(f"GPU {i}: {gpu_name} ({gpu_memory:.2f} GB)")
-        
-        logger.info(f"CUDA is available. Found {gpu_count} GPU(s):")
-        for info in gpu_info:
-            logger.info(f"  {info}")
-        return True
-    else:
-        logger.warning("CUDA is not available. Running in CPU-only mode.")
-        return False
-
-def parse_model_args() -> argparse.Namespace:
-    """
-    Parse command line arguments for model paths and configuration.
-    
-    Returns:
-        argparse.Namespace: Parsed arguments
-    """
-    parser = argparse.ArgumentParser(description="MatrixGame Model Configuration")
-    
-    # Model paths
-    parser.add_argument("--model_root", type=str, default="./models/matrixgame",
-                        help="Root directory for model files")
-    parser.add_argument("--dit_path", type=str, default=None,
-                        help="Path to DIT model. If not provided, will use MODEL_ROOT/dit/")
-    parser.add_argument("--vae_path", type=str, default=None,
-                        help="Path to VAE model. If not provided, will use MODEL_ROOT/vae/")
-    parser.add_argument("--textenc_path", type=str, default=None,
-                        help="Path to text encoder model. If not provided, will use MODEL_ROOT")
-    
-    # Model settings
-    parser.add_argument("--inference_steps", type=int, default=20,
-                        help="Number of inference steps for frame generation (lower is faster)")
-    parser.add_argument("--guidance_scale", type=float, default=6.0,
-                        help="Guidance scale for generation")
-    parser.add_argument("--frame_width", type=int, default=640,
-                        help="Width of the generated frames")
-    parser.add_argument("--frame_height", type=int, default=360,
-                        help="Height of the generated frames")
-    parser.add_argument("--num_pre_frames", type=int, default=3,
-                        help="Number of pre-frames for conditioning")
-    parser.add_argument("--fps", type=int, default=16,
-                        help="Frames per second for video")
-    
-    args = parser.parse_args()
-    
-    # Set environment variables for model paths if provided
-    if args.model_root:
-        os.environ.setdefault("MODEL_ROOT", args.model_root)
-    if args.dit_path:
-        os.environ.setdefault("DIT_PATH", args.dit_path)
-    else:
-        os.environ.setdefault("DIT_PATH", os.path.join(os.environ.get("MODEL_ROOT", "./models/matrixgame"), "dit/"))
-    if args.vae_path:
-        os.environ.setdefault("VAE_PATH", args.vae_path)
-    else:
-        os.environ.setdefault("VAE_PATH", os.path.join(os.environ.get("MODEL_ROOT", "./models/matrixgame"), "vae/"))
-    if args.textenc_path:
-        os.environ.setdefault("TEXTENC_PATH", args.textenc_path)
-    else:
-        os.environ.setdefault("TEXTENC_PATH", os.environ.get("MODEL_ROOT", "./models/matrixgame"))
-    
-    return args
-
-def visualize_controls(frame: np.ndarray, keyboard_condition: List, mouse_condition: List, 
-                       frame_width: int, frame_height: int) -> np.ndarray:
-    """
-    Visualize keyboard and mouse controls on the frame.
-    
-    Args:
-        frame: The video frame to visualize on
-        keyboard_condition: Keyboard state as a list
-        mouse_condition: Mouse state as a list
-        frame_width: Width of the frame
-        frame_height: Height of the frame
-        
-    Returns:
-        np.ndarray: Frame with visualized controls
-    """
-    # Clone the frame to avoid modifying the original
-    frame = frame.copy()
-    
-    # If we have keyboard/mouse conditions, visualize them on the frame
-    if keyboard_condition:
-        # Visualize keyboard inputs
-        keys = ["W", "S", "A", "D", "JUMP", "ATTACK"]
-        for i, key_pressed in enumerate(keyboard_condition[0]):
-            color = (0, 255, 0) if key_pressed else (100, 100, 100)
-            cv2.putText(frame, keys[i], (20 + i*100, 30), 
-                      cv2.FONT_HERSHEY_SIMPLEX, 0.7, color, 2)
-            
-    if mouse_condition:
-        # Visualize mouse movement
-        mouse_x, mouse_y = mouse_condition[0]
-        # Scale mouse values for visualization
-        offset_x = int(mouse_x * 100)
-        offset_y = int(mouse_y * 100)
-        center_x, center_y = frame_width // 2, frame_height // 2
-        cv2.circle(frame, (center_x + offset_x, center_y - offset_y), 10, (255, 0, 0), -1)
-        cv2.putText(frame, f"Mouse: {mouse_x:.2f}, {mouse_y:.2f}", 
-                   (frame_width - 250, 30), cv2.FONT_HERSHEY_SIMPLEX, 0.7, (255, 0, 0), 2)
-    
-    return frame
-
-def frame_to_jpeg(frame: np.ndarray, frame_height: int, frame_width: int) -> bytes:
-    """
-    Convert a frame to JPEG bytes.
-    
-    Args:
-        frame: The video frame to convert
-        frame_height: Height of the frame for fallback
-        frame_width: Width of the frame for fallback
-        
-    Returns:
-        bytes: JPEG bytes of the frame
-    """
-    success, buffer = cv2.imencode('.jpg', frame)
-    if not success:
-        logger.error("Failed to encode frame as JPEG")
-        # Return a blank frame
-        blank = np.ones((frame_height, frame_width, 3), dtype=np.uint8) * 100
-        success, buffer = cv2.imencode('.jpg', blank)
-        
-    return buffer.tobytes()
-
-def load_scene_frames(scene_name: str, frame_width: int, frame_height: int) -> List[np.ndarray]:
-    """
-    Load initial frames for a scene from asset directory.
-    
-    Args:
-        scene_name: Name of the scene
-        frame_width: Width to resize frames to
-        frame_height: Height to resize frames to
-        
-    Returns:
-        List[np.ndarray]: List of frames as numpy arrays
-    """
-    frames = []
-    scene_dir = f"./GameWorldScore/asset/init_image/{scene_name}"
-    
-    if os.path.exists(scene_dir):
-        image_files = sorted([f for f in os.listdir(scene_dir) if f.endswith('.png') or f.endswith('.jpg')])
-        for img_file in image_files:
-            try:
-                img_path = os.path.join(scene_dir, img_file)
-                img = Image.open(img_path).convert("RGB")
-                img = img.resize((frame_width, frame_height))
-                frames.append(np.array(img))
-            except Exception as e:
-                logger.error(f"Error loading image {img_file}: {str(e)}")
-    
-    # If no frames were loaded, create a default colored frame with text
-    if not frames:
-        frame = np.ones((frame_height, frame_height, 3), dtype=np.uint8) * 100
-        # Add scene name as text
-        cv2.putText(frame, f"Scene: {scene_name}", (50, 180), 
-                   cv2.FONT_HERSHEY_SIMPLEX, 1, (255, 255, 255), 2)
-        frames.append(frame)
-        
-    return frames

reference_example/Dockerfile

@@ -1,52 +0,0 @@
-FROM nvidia/cuda:12.4.0-devel-ubuntu22.04
-
-ARG DEBIAN_FRONTEND=noninteractive
-
-ENV PYTHONUNBUFFERED=1
-
-RUN apt-get update && apt-get install --no-install-recommends -y \
-  build-essential \
-  python3.11 \
-  python3-pip \
-  python3-dev \
-  git \
-  curl \
-  ffmpeg \
-  libglib2.0-0 \
-  libsm6 \
-  libxrender1 \
-  libxext6 \
-  && apt-get clean && rm -rf /var/lib/apt/lists/*
-
-WORKDIR /code
-
-COPY ./requirements.txt /code/requirements.txt
-
-# Set up a new user named "user" with user ID 1000
-RUN useradd -m -u 1000 user
-# Switch to the "user" user
-USER user
-# Set home to the user's home directory
-ENV HOME=/home/user \
-	PATH=/home/user/.local/bin:$PATH
-
-# Set home to the user's home directory
-ENV PYTHONPATH=$HOME/app \
-  PYTHONUNBUFFERED=1 \
-  DATA_ROOT=/tmp/data
-
-RUN echo "Installing requirements.txt"
-RUN pip3 install --no-cache-dir --upgrade -r /code/requirements.txt
-
-# yeah.. this is manual for now
-#RUN flutter build web
-
-WORKDIR $HOME/app
-
-COPY --chown=user . $HOME/app
-
-EXPOSE 8080
-
-ENV PORT 8080
-
-CMD python3 api.py

reference_example/api.py

@@ -1,297 +0,0 @@
-import asyncio
-import json
-import logging
-import os
-import pathlib
-import time
-import uuid
-from aiohttp import web, WSMsgType
-from typing import Dict, Any
-
-from api_core import VideoGenerationAPI
-from api_session import SessionManager
-from api_metrics import MetricsTracker
-from api_config import *
-
-# Configure logging
-logging.basicConfig(
-    level=logging.INFO,
-    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
-)
-logger = logging.getLogger(__name__)
-
-# Create global session and metrics managers
-session_manager = SessionManager()
-metrics_tracker = MetricsTracker()
-
-# Dictionary to track connected anonymous clients by IP address
-anon_connections = {}
-anon_connection_lock = asyncio.Lock()
-
-async def status_handler(request: web.Request) -> web.Response:
-    """Handler for API status endpoint"""
-    api = session_manager.shared_api
-    
-    # Get current busy status of all endpoints
-    endpoint_statuses = []
-    for ep in api.endpoint_manager.endpoints:
-        endpoint_statuses.append({
-            'id': ep.id,
-            'url': ep.url,
-            'busy': ep.busy,
-            'last_used': ep.last_used,
-            'error_count': ep.error_count,
-            'error_until': ep.error_until
-        })
-    
-    # Get session statistics
-    session_stats = session_manager.get_session_stats()
-    
-    # Get metrics
-    api_metrics = metrics_tracker.get_metrics()
-    
-    return web.json_response({
-        'product': PRODUCT_NAME,
-        'version': PRODUCT_VERSION,
-        'maintenance_mode': MAINTENANCE_MODE,
-        'available_endpoints': len(VIDEO_ROUND_ROBIN_ENDPOINT_URLS),
-        'endpoint_status': endpoint_statuses,
-        'active_endpoints': sum(1 for ep in endpoint_statuses if not ep['busy'] and ('error_until' not in ep or ep['error_until'] < time.time())),
-        'active_sessions': session_stats,
-        'metrics': api_metrics
-    })
-
-async def metrics_handler(request: web.Request) -> web.Response:
-    """Handler for detailed metrics endpoint (protected)"""
-    # Check for API key in header or query param
-    auth_header = request.headers.get('Authorization', '')
-    api_key = None
-    
-    if auth_header.startswith('Bearer '):
-        api_key = auth_header[7:]
-    else:
-        api_key = request.query.get('key')
-    
-    # Validate API key (using SECRET_TOKEN as the API key)
-    if not api_key or api_key != SECRET_TOKEN:
-        return web.json_response({
-            'error': 'Unauthorized'
-        }, status=401)
-    
-    # Get detailed metrics
-    detailed_metrics = metrics_tracker.get_detailed_metrics()
-    
-    return web.json_response(detailed_metrics)
-
-async def websocket_handler(request: web.Request) -> web.WebSocketResponse:
-    # Check if maintenance mode is enabled
-    if MAINTENANCE_MODE:
-        # Return an error response indicating maintenance mode
-        return web.json_response({
-            'error': 'Server is in maintenance mode',
-            'maintenance': True
-        }, status=503)  # 503 Service Unavailable
-    
-    ws = web.WebSocketResponse(
-        max_msg_size=1024*1024*20,  # 20MB max message size
-        timeout=30.0  # we want to keep things tight and short
-    )
-    
-    await ws.prepare(request)
-    
-    # Get the Hugging Face token from query parameters
-    hf_token = request.query.get('hf_token', '')
-    
-    # Generate a unique user ID for this connection
-    user_id = str(uuid.uuid4())
-    
-    # Validate the token and determine the user role
-    user_role = await session_manager.shared_api.validate_user_token(hf_token)
-    logger.info(f"User {user_id} connected with role: {user_role}")
-    
-    # Get client IP address
-    peername = request.transport.get_extra_info('peername')
-    if peername is not None:
-        client_ip = peername[0]
-    else:
-        client_ip = request.headers.get('X-Forwarded-For', 'unknown').split(',')[0].strip()
-    
-    logger.info(f"Client {user_id} connecting from IP: {client_ip} with role: {user_role}")
-    
-    # Check for anonymous user connection limits
-    if user_role == 'anon':
-        async with anon_connection_lock:
-            # Track this connection
-            anon_connections[client_ip] = anon_connections.get(client_ip, 0) + 1
-            # Store the IP so we can clean up later
-            ws.client_ip = client_ip
-            
-            # Log multiple connections from same IP but don't restrict them
-            if anon_connections[client_ip] > 1:
-                logger.info(f"Multiple anonymous connections from IP {client_ip}: {anon_connections[client_ip]} connections")
-    
-    # Store the user role in the websocket for easy access
-    ws.user_role = user_role
-    ws.user_id = user_id
-    
-    # Register with metrics
-    metrics_tracker.register_session(user_id, client_ip)
-    
-    # Create a new session for this user
-    user_session = await session_manager.create_session(user_id, user_role, ws)
-
-    try:
-        async for msg in ws:
-            if msg.type == WSMsgType.TEXT:
-                try:
-                    data = json.loads(msg.data)
-                    action = data.get('action')
-                    
-                    # Check for rate limiting
-                    request_type = 'other'
-                    if action in ['join_chat', 'leave_chat', 'chat_message']:
-                        request_type = 'chat'
-                    elif action in ['generate_video']:
-                        request_type = 'video'
-                    elif action == 'search':
-                        request_type = 'search'
-                    elif action == 'simulate':
-                        request_type = 'simulation'
-                    
-                    # Record the request for metrics
-                    await metrics_tracker.record_request(user_id, client_ip, request_type, user_role)
-                    
-                    # Check rate limits (except for admins)
-                    if user_role != 'admin' and await metrics_tracker.is_rate_limited(user_id, request_type, user_role):
-                        await ws.send_json({
-                            'action': action,
-                            'requestId': data.get('requestId'),
-                            'success': False,
-                            'error': f'Rate limit exceeded for {request_type} requests. Please try again later.'
-                        })
-                        continue
-                    
-                    # Route requests to appropriate queues
-                    if action in ['join_chat', 'leave_chat', 'chat_message']:
-                        await user_session.chat_queue.put(data)
-                    elif action in ['generate_video']:
-                        await user_session.video_queue.put(data)
-                    elif action == 'search':
-                        await user_session.search_queue.put(data)
-                    elif action == 'simulate':
-                        await user_session.simulation_queue.put(data)
-                    else:
-                        await user_session.process_generic_request(data)
-                        
-                except Exception as e:
-                    logger.error(f"Error processing WebSocket message for user {user_id}: {str(e)}")
-                    await ws.send_json({
-                        'action': data.get('action') if 'data' in locals() else 'unknown',
-                        'success': False,
-                        'error': f'Error processing message: {str(e)}'
-                    })
-                    
-            elif msg.type in (WSMsgType.ERROR, WSMsgType.CLOSE):
-                break
-                
-    finally:
-        # Cleanup session
-        await session_manager.delete_session(user_id)
-        
-        # Cleanup anonymous connection tracking
-        if getattr(ws, 'user_role', None) == 'anon' and hasattr(ws, 'client_ip'):
-            client_ip = ws.client_ip
-            async with anon_connection_lock:
-                if client_ip in anon_connections:
-                    anon_connections[client_ip] = max(0, anon_connections[client_ip] - 1)
-                    if anon_connections[client_ip] == 0:
-                        del anon_connections[client_ip]
-                    logger.info(f"Anonymous connection from {client_ip} closed. Remaining: {anon_connections.get(client_ip, 0)}")
-        
-        # Unregister from metrics
-        metrics_tracker.unregister_session(user_id, client_ip)
-        logger.info(f"Connection closed for user {user_id}")
-    
-    return ws
-
-async def init_app() -> web.Application:
-    app = web.Application(
-        client_max_size=1024**2*20  # 20MB max size
-    )
-    
-    # Add cleanup logic
-    async def cleanup(app):
-        logger.info("Shutting down server, closing all sessions...")
-        await session_manager.close_all_sessions()
-    
-    app.on_shutdown.append(cleanup)
-    
-    # Add routes
-    app.router.add_get('/ws', websocket_handler)
-    app.router.add_get('/api/status', status_handler)
-    app.router.add_get('/api/metrics', metrics_handler)
-    
-    # Set up static file serving
-    # Define the path to the public directory
-    public_path = pathlib.Path(__file__).parent / 'build' / 'web'
-    if not public_path.exists():
-        public_path.mkdir(parents=True, exist_ok=True)
-    
-    # Set up static file serving with proper security considerations
-    async def static_file_handler(request):
-        # Get the path from the request (removing leading /)
-        path_parts = request.path.lstrip('/').split('/')
-        
-        # Convert to safe path to prevent path traversal attacks
-        safe_path = public_path.joinpath(*path_parts)
-        
-        # Make sure the path is within the public directory (prevent directory traversal)
-        try:
-            safe_path = safe_path.resolve()
-            if not str(safe_path).startswith(str(public_path.resolve())):
-                return web.HTTPForbidden(text="Access denied")
-        except (ValueError, FileNotFoundError):
-            return web.HTTPNotFound()
-        
-        # If path is a directory, look for index.html
-        if safe_path.is_dir():
-            safe_path = safe_path / 'index.html'
-        
-        # Check if the file exists
-        if not safe_path.exists() or not safe_path.is_file():
-            # If not found, serve index.html (for SPA routing)
-            safe_path = public_path / 'index.html'
-            if not safe_path.exists():
-                return web.HTTPNotFound()
-        
-        # Determine content type based on file extension
-        content_type = 'text/plain'
-        ext = safe_path.suffix.lower()
-        if ext == '.html':
-            content_type = 'text/html'
-        elif ext == '.js':
-            content_type = 'application/javascript'
-        elif ext == '.css':
-            content_type = 'text/css'
-        elif ext in ('.jpg', '.jpeg'):
-            content_type = 'image/jpeg'
-        elif ext == '.png':
-            content_type = 'image/png'
-        elif ext == '.gif':
-            content_type = 'image/gif'
-        elif ext == '.svg':
-            content_type = 'image/svg+xml'
-        elif ext == '.json':
-            content_type = 'application/json'
-        
-        # Return the file with appropriate headers
-        return web.FileResponse(safe_path, headers={'Content-Type': content_type})
-    
-    # Add catch-all route for static files (lower priority than API routes)
-    app.router.add_get('/{path:.*}', static_file_handler)
-    
-    return app
-
-if __name__ == '__main__':
-    app = asyncio.run(init_app())
-    web.run_app(app, host='0.0.0.0', port=8080)

reference_example/api_config.py

@@ -1,184 +0,0 @@
-import os
-
-PRODUCT_NAME = os.environ.get('PRODUCT_NAME', 'TikSlop')
-PRODUCT_VERSION = "2.0.0"
-
-# you should use Mistral 7b instruct for good performance and accuracy balance
-TEXT_MODEL = os.environ.get('HF_TEXT_MODEL', '')
-
-# Environment variable to control maintenance mode
-MAINTENANCE_MODE = os.environ.get('MAINTENANCE_MODE', 'false').lower() in ('true', 'yes', '1', 't')
-
-# Environment variable to control how many nodes to use
-MAX_NODES = int(os.environ.get('MAX_NODES', '8'))
-
-ADMIN_ACCOUNTS = [
-    "jbilcke-hf"
-]
-
-RAW_VIDEO_ROUND_ROBIN_ENDPOINT_URLS = [
-    os.environ.get('VIDEO_ROUND_ROBIN_SERVER_1', ''),
-    os.environ.get('VIDEO_ROUND_ROBIN_SERVER_2', ''),
-    os.environ.get('VIDEO_ROUND_ROBIN_SERVER_3', ''),
-    os.environ.get('VIDEO_ROUND_ROBIN_SERVER_4', ''),
-    os.environ.get('VIDEO_ROUND_ROBIN_SERVER_5', ''),
-    os.environ.get('VIDEO_ROUND_ROBIN_SERVER_6', ''),
-    os.environ.get('VIDEO_ROUND_ROBIN_SERVER_7', ''),
-    os.environ.get('VIDEO_ROUND_ROBIN_SERVER_8', ''),
-]
-
-# Filter out empty strings from the endpoint list
-filtered_urls = [url for url in RAW_VIDEO_ROUND_ROBIN_ENDPOINT_URLS if url]
-
-# Limit the number of URLs based on MAX_NODES environment variable
-VIDEO_ROUND_ROBIN_ENDPOINT_URLS = filtered_urls[:MAX_NODES]
-
-HF_TOKEN = os.environ.get('HF_TOKEN')
-
-# use the same secret token as you used to secure your BASE_SPACE_NAME spaces
-SECRET_TOKEN = os.environ.get('SECRET_TOKEN')
-
-# altenative words we could use: "saturated, highlight, overexposed, highlighted, overlit, shaking, too bright, worst quality, inconsistent motion, blurry, jittery, distorted, cropped, watermarked, watermark, logo, subtitle, subtitles, lowres"
-NEGATIVE_PROMPT = "low quality, worst quality, deformed, distorted, disfigured, blurry, text, watermark"
-
-POSITIVE_PROMPT_SUFFIX = "high quality, cinematic, 4K, intricate details"
-
-GUIDANCE_SCALE = 1.0
-
-THUMBNAIL_FRAMES = 65
-
-# anonymous users are people browing TikSlop without being connected
-# this category suffers from regular abuse so we need to enforce strict limitations
-CONFIG_FOR_ANONYMOUS_USERS = {
-
-    # anons can only watch 2 minutes per video
-    "max_rendering_time_per_client_per_video_in_sec": 2 * 60,
-
-    "min_num_inference_steps": 2,
-    "default_num_inference_steps": 4,
-    "max_num_inference_steps": 4,
-
-    "min_num_frames": 9, # 8 + 1
-    "default_max_num_frames": 65, # 8*8 + 1
-    "max_num_frames": 65, # 8*8 + 1
-
-    "min_clip_duration_seconds": 1,
-    "default_clip_duration_seconds": 2,
-    "max_clip_duration_seconds": 2,
-
-    "min_clip_playback_speed": 0.7,
-    "default_clip_playback_speed": 0.7,
-    "max_clip_playback_speed": 0.7,
-
-    "min_clip_framerate": 8,
-    "default_clip_framerate": 16,
-    "max_clip_framerate": 16,
-
-    "min_clip_width": 544,
-    "default_clip_width": 640,
-    "max_clip_width": 640,
-
-    "min_clip_height": 320,
-    "default_clip_height": 352,
-    "max_clip_height": 352,
-}
-
-# Hugging Face users enjoy a more normal and calibrated experience
-CONFIG_FOR_STANDARD_HF_USERS = {
-    "max_rendering_time_per_client_per_video_in_sec": 15 * 60,
-
-    "min_num_inference_steps": 2,
-    "default_num_inference_steps": 4,
-    "max_num_inference_steps": 4,
-    
-    "min_num_frames": 9, # 8 + 1
-    "default_num_frames": 81, # 8*10 + 1
-    "max_num_frames": 81,
-
-    "min_clip_duration_seconds": 1,
-    "default_clip_duration_seconds": 3,
-    "max_clip_duration_seconds": 3,
-
-    "min_clip_playback_speed": 0.7,
-    "default_clip_playback_speed": 0.7,
-    "max_clip_playback_speed": 0.7,
-
-    "min_clip_framerate": 8,
-    "default_clip_framerate": 25,
-    "max_clip_framerate": 25,
-
-    "min_clip_width": 544,
-    "default_clip_width": 1152, # 928, # 1216, # 768, # 640,
-    "max_clip_width": 1152, # 928, # 1216, # 768, # 640,
-
-    "min_clip_height": 320,
-    "default_clip_height": 640, # 512, # 448, # 416,
-    "max_clip_height": 640, # 512, # 448, # 416,
-}
-
-# Hugging Face users with a Pro may enjoy an improved experience
-CONFIG_FOR_PRO_HF_USERS = {
-    "max_rendering_time_per_client_per_video_in_sec": 20 * 60,
-
-    "min_num_inference_steps": 2,
-    "default_num_inference_steps": 4,
-    "max_num_inference_steps": 4,
-    
-    "min_num_frames": 9, # 8 + 1
-    "default_num_frames": 81, # 8*10 + 1
-    "max_num_frames": 81,
-
-    "min_clip_duration_seconds": 1,
-    "default_clip_duration_seconds": 3,
-    "max_clip_duration_seconds": 3,
-
-    "min_clip_playback_speed": 0.7,
-    "default_clip_playback_speed": 0.7,
-    "max_clip_playback_speed": 0.7,
-
-    "min_clip_framerate": 8,
-    "default_clip_framerate": 25,
-    "max_clip_framerate": 25,
-
-    "min_clip_width": 544,
-    "default_clip_width": 1152, # 928, # 1216, # 768, # 640,
-    "max_clip_width": 1152, # 928, # 1216, # 768, # 640,
-
-    "min_clip_height": 320,
-    "default_clip_height": 640, # 512, # 448, # 416,
-    "max_clip_height": 640, # 512, # 448, # 416,
-}
-
-CONFIG_FOR_ADMIN_HF_USERS = {
-    "max_rendering_time_per_client_per_video_in_sec": 60 * 60,
-
-    "min_num_inference_steps": 2,
-    "default_num_inference_steps": 4,
-    "max_num_inference_steps": 4,
-
-    "min_num_frames": 9, # 8 + 1
-    "default_num_frames": 81, # (8 * 10) + 1
-    "max_num_frames": 129, # (8 * 16) + 1
-
-    "min_clip_duration_seconds": 1,
-    "default_clip_duration_seconds": 2,
-    "max_clip_duration_seconds": 4,
-
-    "min_clip_playback_speed": 0.7,
-    "default_clip_playback_speed": 0.7,
-    "max_clip_playback_speed": 1.0,
-
-    "min_clip_framerate": 8,
-    "default_clip_framerate": 30,
-    "max_clip_framerate": 60,
-
-    "min_clip_width": 544,
-    "default_clip_width": 1152, # 928, # 1216, # 768, # 640,
-    "max_clip_width": 1152, # 928, # 1216, # 768, # 640,
-
-    "min_clip_height": 320,
-    "default_clip_height": 640, # 512, # 448, # 416,
-    "max_clip_height": 640, # 512, # 448, # 416,
-}
-
-CONFIG_FOR_ADMIN_HF_USERS = CONFIG_FOR_PRO_HF_USERS

reference_example/api_core.py

@@ -1,1068 +0,0 @@
-import logging
-import os
-import io
-import re
-import base64
-import uuid
-from typing import Dict, Any, Optional, List, Literal
-from dataclasses import dataclass
-from asyncio import Lock, Queue
-import asyncio
-import time
-import datetime
-from contextlib import asynccontextmanager
-from collections import defaultdict
-from aiohttp import web, ClientSession
-from huggingface_hub import InferenceClient, HfApi
-from gradio_client import Client
-import random
-import yaml
-import json
-
-from api_config import *
-
-# User role type
-UserRole = Literal['anon', 'normal', 'pro', 'admin']
-
-# Configure logging
-logging.basicConfig(
-    level=logging.INFO,
-    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
-)
-logger = logging.getLogger(__name__)
-
-
-def generate_seed():
-    """Generate a random positive 32-bit integer seed."""
-    return random.randint(0, 2**32 - 1)
-
-def sanitize_yaml_response(response_text: str) -> str:
-    """
-    Sanitize and format AI response into valid YAML.
-    Returns properly formatted YAML string.
-    """
-
-    response_text = response_text.split("```")[0]
-
-    # Remove any markdown code block indicators and YAML document markers
-    clean_text = re.sub(r'```yaml|```|---|\.\.\.$', '', response_text.strip())
-    
-    # Split into lines and process each line
-    lines = clean_text.split('\n')
-    sanitized_lines = []
-    current_field = None
-    
-    for line in lines:
-        stripped = line.strip()
-        if not stripped:
-            continue
-            
-        # Handle field starts
-        if stripped.startswith('title:') or stripped.startswith('description:'):
-            # Ensure proper YAML format with space after colon and proper quoting
-            field_name = stripped.split(':', 1)[0]
-            field_value = stripped.split(':', 1)[1].strip().strip('"\'')
-            
-            # Quote the value if it contains special characters
-            if any(c in field_value for c in ':[]{},&*#?|-<>=!%@`'):
-                field_value = f'"{field_value}"'
-                
-            sanitized_lines.append(f"{field_name}: {field_value}")
-            current_field = field_name
-            
-        elif stripped.startswith('tags:'):
-            sanitized_lines.append('tags:')
-            current_field = 'tags'
-            
-        elif stripped.startswith('-') and current_field == 'tags':
-            # Process tag values
-            tag = stripped[1:].strip().strip('"\'')
-            if tag:
-                # Clean and format tag
-                tag = re.sub(r'[^\x00-\x7F]+', '', tag)  # Remove non-ASCII
-                tag = re.sub(r'[^a-zA-Z0-9\s-]', '', tag)  # Keep only alphanumeric and hyphen
-                tag = tag.strip().lower().replace(' ', '-')
-                if tag:
-                    sanitized_lines.append(f"  - {tag}")
-                    
-        elif current_field in ['title', 'description']:
-            # Handle multi-line title/description continuation
-            value = stripped.strip('"\'')
-            if value:
-                # Append to previous line
-                prev = sanitized_lines[-1]
-                sanitized_lines[-1] = f"{prev} {value}"
-    
-    # Ensure the YAML has all required fields
-    required_fields = {'title', 'description', 'tags'}
-    found_fields = {line.split(':')[0].strip() for line in sanitized_lines if ':' in line}
-    
-    for field in required_fields - found_fields:
-        if field == 'tags':
-            sanitized_lines.extend(['tags:', '  - default'])
-        else:
-            sanitized_lines.append(f'{field}: "No {field} provided"')
-    
-    return '\n'.join(sanitized_lines)
-
-@dataclass
-class Endpoint:
-    id: int
-    url: str
-    busy: bool = False
-    last_used: float = 0
-    error_count: int = 0
-    error_until: float = 0  # Timestamp until which this endpoint is considered in error state
-
-class EndpointManager:
-    def __init__(self):
-        self.endpoints: List[Endpoint] = []
-        self.lock = Lock()
-        self.initialize_endpoints()
-        self.last_used_index = -1  # Track the last used endpoint for round-robin
-
-    def initialize_endpoints(self):
-        """Initialize the list of endpoints"""
-        for i, url in enumerate(VIDEO_ROUND_ROBIN_ENDPOINT_URLS):
-            endpoint = Endpoint(id=i + 1, url=url)
-            self.endpoints.append(endpoint)
-
-    def _get_next_free_endpoint(self):
-        """Get the next available non-busy endpoint, or oldest endpoint if all are busy"""
-        current_time = time.time()
-        
-        # First priority: Get any non-busy and non-error endpoint
-        free_endpoints = [
-            ep for ep in self.endpoints 
-            if not ep.busy and current_time > ep.error_until
-        ]
-        
-        if free_endpoints:
-            # Return the least recently used free endpoint
-            return min(free_endpoints, key=lambda ep: ep.last_used)
-        
-        # Second priority: If all busy/error, use round-robin but skip error endpoints
-        tried_count = 0
-        next_index = self.last_used_index
-        
-        while tried_count < len(self.endpoints):
-            next_index = (next_index + 1) % len(self.endpoints)
-            tried_count += 1
-            
-            # If endpoint is not in error state, use it
-            if current_time > self.endpoints[next_index].error_until:
-                self.last_used_index = next_index
-                return self.endpoints[next_index]
-        
-        # If all endpoints are in error state, use the one with earliest error expiry
-        self.last_used_index = next_index
-        return min(self.endpoints, key=lambda ep: ep.error_until)
-
-    @asynccontextmanager
-    async def get_endpoint(self, max_wait_time: int = 10):
-        """Get the next available endpoint using a context manager"""
-        start_time = time.time()
-        endpoint = None
-        
-        try:
-            while True:
-                if time.time() - start_time > max_wait_time:
-                    raise TimeoutError(f"Could not acquire an endpoint within {max_wait_time} seconds")
-
-                async with self.lock:
-                    # Get the next available endpoint using our selection strategy
-                    endpoint = self._get_next_free_endpoint()
-                    
-                    # Mark it as busy
-                    endpoint.busy = True
-                    endpoint.last_used = time.time()
-                    #logger.info(f"Using endpoint {endpoint.id} (busy: {endpoint.busy}, last used: {endpoint.last_used})")
-                    break
-
-            yield endpoint
-
-        finally:
-            if endpoint:
-                async with self.lock:
-                    endpoint.busy = False
-                    endpoint.last_used = time.time()
-                    # We don't need to put back into queue - our strategy now picks directly from the list
-
-class ChatRoom:
-    def __init__(self):
-        self.messages = []
-        self.connected_clients = set()
-        self.max_history = 100
-
-    def add_message(self, message):
-        self.messages.append(message)
-        if len(self.messages) > self.max_history:
-            self.messages.pop(0)
-
-    def get_recent_messages(self, limit=50):
-        return self.messages[-limit:]
-
-class VideoGenerationAPI:
-    def __init__(self):
-        self.inference_client = InferenceClient(token=HF_TOKEN)
-        self.hf_api = HfApi(token=HF_TOKEN)
-        self.endpoint_manager = EndpointManager()
-        self.active_requests: Dict[str, asyncio.Future] = {}
-        self.chat_rooms = defaultdict(ChatRoom)
-        self.video_events: Dict[str, List[Dict[str, Any]]] = defaultdict(list)
-        self.event_history_limit = 50
-        # Cache for user roles to avoid repeated API calls
-        self.user_role_cache: Dict[str, Dict[str, Any]] = {}
-        # Cache expiration time (10 minutes)
-        self.cache_expiration = 600
-
-
-    def _add_event(self, video_id: str, event: Dict[str, Any]):
-        """Add an event to the video's history and maintain the size limit"""
-        events = self.video_events[video_id]
-        events.append(event)
-        if len(events) > self.event_history_limit:
-            events.pop(0)
-    
-    async def validate_user_token(self, token: str) -> UserRole:
-        """
-        Validates a Hugging Face token and determines the user's role.
-        
-        Returns one of:
-        - 'anon': Anonymous user (no token or invalid token)
-        - 'normal': Standard Hugging Face user
-        - 'pro': Hugging Face Pro user
-        - 'admin': Admin user (username in ADMIN_ACCOUNTS)
-        """
-        # If no token is provided, the user is anonymous
-        if not token:
-            return 'anon'
-        
-        # Check if we have a cached result for this token
-        current_time = time.time()
-        if token in self.user_role_cache:
-            cached_data = self.user_role_cache[token]
-            # If the cache is still valid
-            if current_time - cached_data['timestamp'] < self.cache_expiration:
-                logger.info(f"Using cached user role: {cached_data['role']}")
-                return cached_data['role']
-        
-        # No valid cache, need to check the token with the HF API
-        try:
-            # Use HF API to validate the token and get user info
-            logger.info("Validating Hugging Face token...")
-            
-            # Run in executor to avoid blocking the event loop
-            user_info = await asyncio.get_event_loop().run_in_executor(
-                None, 
-                lambda: self.hf_api.whoami(token=token)
-            )
-            
-            # Handle both object and dict response formats from whoami
-            username = user_info.get('name') if isinstance(user_info, dict) else getattr(user_info, 'name', None)
-            is_pro = user_info.get('is_pro') if isinstance(user_info, dict) else getattr(user_info, 'is_pro', False)
-            
-            if not username:
-                logger.error(f"Could not determine username from user_info: {user_info}")
-                return 'anon'
-                
-            logger.info(f"Token valid for user: {username}")
-            
-            # Determine the user role based on the information
-            user_role: UserRole
-            
-            # Check if the user is an admin
-            if username in ADMIN_ACCOUNTS:
-                user_role = 'admin'
-            # Check if the user has a pro account
-            elif is_pro:
-                user_role = 'pro'
-            else:
-                user_role = 'normal'
-            
-            # Cache the result
-            self.user_role_cache[token] = {
-                'role': user_role,
-                'timestamp': current_time,
-                'username': username
-            }
-            
-            return user_role
-            
-        except Exception as e:
-            logger.error(f"Failed to validate Hugging Face token: {str(e)}")
-            # If validation fails, the user is treated as anonymous
-            return 'anon'
-
-    async def download_video(self, url: str) -> bytes:
-        """Download video file from URL and return bytes"""
-        async with ClientSession() as session:
-            async with session.get(url) as response:
-                if response.status != 200:
-                    raise Exception(f"Failed to download video: HTTP {response.status}")
-                return await response.read()
-
-    async def search_video(self, query: str, attempt_count: int = 0) -> Optional[dict]:
-        """Generate a single search result using HF text generation"""
-        # Maximum number of attempts to generate a description without placeholder tags
-        max_attempts = 2
-        current_attempt = attempt_count
-        # Use a random temperature between 0.68 and 0.72 to generate more diverse results
-        # and prevent duplicate results from successive calls with the same prompt
-        temperature = random.uniform(0.68, 0.72)
-
-        while current_attempt <= max_attempts:
-            prompt = f"""# Instruction
-Your response MUST be a YAML object containing a title and description, consistent with what we can find on a video sharing platform.
-Format your YAML response with only those fields: "title" (a short string) and "description" (string caption of the scene). Do not add any other field.
-In the description field, describe in a very synthetic way the visuals of the first shot (first scene), eg "<STYLE>, medium close-up shot, high angle view. In the foreground a <OPTIONAL AGE> <OPTIONAL GENDER> <CHARACTERS> <ACTIONS>. In the background <DESCRIBE LOCATION, BACKGROUND CHARACTERS, OBJECTS ETC>. The scene is lit by <LIGHTING> <WEATHER>". This is just an example! you MUST replace the <TAGS>!!.
-Don't forget to replace <STYLE> etc, by the actual fields!!
-For the style, be creative, for instance you can use anything like a "documentary footage", "japanese animation", "movie scene", "tv series", "tv show", "security footage" etc.
-If the user ask for something specific eg "movie screencap", "movie scene", "documentary footage" "animation" as a style etc.
-Keep it minimalist but still descriptive, don't use bullets points, use simple words, go to the essential to describe style (cinematic, documentary footage, 3D rendering..), camera modes and angles, characters, age, gender, action, location, lighting, country, costume, time, weather, textures, color palette.. etc). Write about 80 words, and use between 2 and 3 sentences.
-The most import part is to describe the actions and movements in the scene, so don't forget that!
-Don't describe sound, so ever say things like "atmospheric music playing in the background".
-Instead describe the visual elements we can see in the background, be precise, (if there are anything, cars, objects, people, bricks, birds, clouds, trees, leaves or grass then say it so etc).
-Make the result unique and different from previous search results. ONLY RETURN YAML AND WITH ENGLISH CONTENT, NOT CHINESE - DO NOT ADD ANY OTHER COMMENT!
-
-# Context
-This is attempt {current_attempt}.
-
-# Input
-Describe the first scene/shot for: "{query}".
-
-# Output
-
-```yaml
-title: \""""
-
-            try:
-                response = await asyncio.get_event_loop().run_in_executor(
-                    None,
-                    lambda: self.inference_client.text_generation(
-                        prompt,
-                        model=TEXT_MODEL,
-                        max_new_tokens=200,
-                        temperature=temperature
-                    )
-                )
-
-                response_text = re.sub(r'^\s*\.\s*\n', '', f"title: \"{response.strip()}")
-                sanitized_yaml = sanitize_yaml_response(response_text)
-                
-                try:
-                    result = yaml.safe_load(sanitized_yaml)
-                except yaml.YAMLError as e:
-                    logger.error(f"YAML parsing failed: {str(e)}")
-                    result = None
-                
-                if not result or not isinstance(result, dict):
-                    logger.error(f"Invalid result format: {result}")
-                    current_attempt += 1
-                    temperature = random.uniform(0.68, 0.72)  # Try with different random temperature on next attempt
-                    continue
-
-                # Extract fields with defaults
-                title = str(result.get('title', '')).strip() or 'Untitled Video'
-                description = str(result.get('description', '')).strip() or 'No description available'
-                
-                # Check if the description still contains placeholder tags like <LOCATION>, <GENDER>, etc.
-                if re.search(r'<[A-Z_]+>', description):
-                    #logger.warning(f"Description still contains placeholder tags: {description}")
-                    if current_attempt < max_attempts:
-                        # Try again with a different random temperature
-                        current_attempt += 1
-                        temperature = random.uniform(0.68, 0.72)
-                        continue
-                    else:
-                        # If we've reached max attempts, use the title as description
-                        description = title
-
-                # Return valid result with all required fields
-                return {
-                    'id': str(uuid.uuid4()),
-                    'title': title,
-                    'description': description,
-                    'thumbnailUrl': '',
-                    'videoUrl': '',
-
-                    # not really used yet, maybe one day if we pre-generate or store content
-                    'isLatent': True,
-
-                    'useFixedSeed': "webcam" in description.lower(),
-
-                    'seed': generate_seed(),
-                    'views': 0,
-                    'tags': []
-                }
-
-            except Exception as e:
-                logger.error(f"Search video generation failed: {str(e)}")
-                current_attempt += 1
-                temperature = random.uniform(0.68, 0.72)  # Try with different random temperature on next attempt
-        
-        # If all attempts failed, return a simple result with title only
-        return {
-            'id': str(uuid.uuid4()),
-            'title': f"Video about {query}",
-            'description': f"Video about {query}",
-            'thumbnailUrl': '',
-            'videoUrl': '',
-            'isLatent': True,
-            'useFixedSeed': "query" in description.lower(),
-            'seed': generate_seed(),
-            'views': 0,
-            'tags': []
-        }
-
-    # The generate_thumbnail function has been removed because we now use
-    # generate_video_thumbnail for all thumbnails, which generates a video clip
-    # instead of a static image
-
-    async def generate_caption(self, title: str, description: str) -> str:
-        """Generate detailed caption using HF text generation"""
-        try:
-            prompt = f"""Generate a detailed story for a video named: "{title}"
-Visual description of the video: {description}.
-Instructions: Write the story summary, including the plot, action, what should happen.
-Make it around 200-300 words long.
-A video can be anything from a tutorial, webcam, trailer, movie, live stream etc."""
-
-            response = await asyncio.get_event_loop().run_in_executor(
-                None,
-                lambda: self.inference_client.text_generation(
-                    prompt,
-                    model=TEXT_MODEL,
-                    max_new_tokens=180,
-                    temperature=0.7
-                )
-            )
-     
-            if "Caption: " in response:
-                response = response.replace("Caption: ", "")
-            
-            chunks = f" {response} ".split(". ")
-            if len(chunks) > 1:
-                text = ". ".join(chunks[:-1])
-            else:
-                text = response
-
-            return text.strip()
-        except Exception as e:
-            logger.error(f"Error generating caption: {str(e)}")
-            return ""
-            
-    async def simulate(self, original_title: str, original_description: str, 
-                         current_description: str, condensed_history: str, 
-                         evolution_count: int = 0, chat_messages: str = '') -> dict:
-        """
-        Simulate a video by evolving its description to create a dynamic narrative.
-        
-        Args:
-            original_title: The original video title
-            original_description: The original video description
-            current_description: The current description (last evolved or original if first evolution)
-            condensed_history: A condensed summary of previous scene developments
-            evolution_count: How many times the simulation has already evolved
-            chat_messages: Chat messages from users to incorporate into the simulation
-            
-        Returns:
-            A dictionary containing the evolved description and updated condensed history
-        """
-        try:
-            # Determine if this is the first simulation
-            is_first_simulation = evolution_count == 0 or not condensed_history
-            
-            logger.info(f"simulate(): is_first_simulation={is_first_simulation}")
-                
-            # Create an appropriate prompt based on whether this is the first simulation
-            chat_section = ""
-            if chat_messages:
-                chat_section = f"""
-People are watching this content right now and have shared their thoughts. Like a game master, please take their feedback as input to adjust the story and/or the scene. Here are their messages:
-
-{chat_messages}
-"""
-
-            if is_first_simulation:
-                prompt = f"""You are tasked with evolving the narrative for a video titled: "{original_title}"
-
-Original description:
-{original_description}
-{chat_section}
-
-Instructions:
-1. Imagine the next logical scene or development that would follow this description.
-2. Create a compelling new description (200-300 words) that builds on the original but introduces new elements, developments, or perspectives.
-3. Maintain the original style, tone, and setting.
-4. If viewers have shared messages, consider their input and incorporate relevant suggestions or reactions into your narrative evolution.
-5. Also create a brief "scene history" (50-75 words) that summarizes what has happened so far.
-
-Return your response in this format:
-EVOLVED_DESCRIPTION: [your new evolved description here]
-CONDENSED_HISTORY: [your scene history summary]"""
-            else:
-                prompt = f"""You are tasked with continuing to evolve the narrative for a video titled: "{original_title}"
-
-Original description:
-{original_description}
-
-Condensed history of scenes so far:
-{condensed_history}
-
-Current description (most recent scene):
-{current_description}
-{chat_section}
-
-Instructions:
-1. Imagine the next logical scene or development that would follow the current description.
-2. Create a compelling new description (200-300 words) that builds on the narrative but introduces new elements, developments, or perspectives.
-3. Maintain consistency with the previous scenes while advancing the story.
-4. If viewers have shared messages, consider their input and incorporate relevant suggestions or reactions into your narrative evolution.
-5. Also update the condensed history (50-75 words) to include this new development.
-
-Return your response in this format:
-EVOLVED_DESCRIPTION: [your new evolved description here]
-CONDENSED_HISTORY: [your updated scene history summary]"""
-
-            # Generate the evolved description
-            response = await asyncio.get_event_loop().run_in_executor(
-                None,
-                lambda: self.inference_client.text_generation(
-                    prompt,
-                    model=TEXT_MODEL,
-                    max_new_tokens=200,
-                    temperature=0.7
-                )
-            )
-            
-            # Extract the evolved description and condensed history from the response
-            evolved_description = ""
-            new_condensed_history = ""
-            
-            # Parse the response
-            if "EVOLVED_DESCRIPTION:" in response and "CONDENSED_HISTORY:" in response:
-                parts = response.split("CONDENSED_HISTORY:")
-                if len(parts) >= 2:
-                    desc_part = parts[0].strip()
-                    if "EVOLVED_DESCRIPTION:" in desc_part:
-                        evolved_description = desc_part.split("EVOLVED_DESCRIPTION:", 1)[1].strip()
-                    new_condensed_history = parts[1].strip()
-            
-            # If parsing failed, use some fallbacks
-            if not evolved_description:
-                evolved_description = current_description
-                logger.warning(f"Failed to parse evolved description, using current description as fallback")
-            
-            if not new_condensed_history and condensed_history:
-                new_condensed_history = condensed_history
-                logger.warning(f"Failed to parse condensed history, using current history as fallback")
-            elif not new_condensed_history:
-                new_condensed_history = f"The video begins with {original_title}: {original_description[:100]}..."
-            
-            return {
-                "evolved_description": evolved_description,
-                "condensed_history": new_condensed_history
-            }
-            
-        except Exception as e:
-            logger.error(f"Error simulating video: {str(e)}")
-            return {
-                "evolved_description": current_description,
-                "condensed_history": condensed_history or f"The video shows {original_title}."
-            }
-
-
-    def get_config_value(self, role: UserRole, field: str, options: dict = None) -> Any:
-        """
-        Get the appropriate config value for a user role.
-        
-        Args:
-            role: The user role ('anon', 'normal', 'pro', 'admin')
-            field: The config field name to retrieve
-            options: Optional user-provided options that may override defaults
-            
-        Returns:
-            The config value appropriate for the user's role with respect to
-            min/max boundaries and user overrides.
-        """
-        # Select the appropriate config based on user role
-        if role == 'admin':
-            config = CONFIG_FOR_ADMIN_HF_USERS
-        elif role == 'pro':
-            config = CONFIG_FOR_PRO_HF_USERS
-        elif role == 'normal':
-            config = CONFIG_FOR_STANDARD_HF_USERS
-        else:  # Anonymous users
-            config = CONFIG_FOR_ANONYMOUS_USERS
-        
-        # Get the default value for this field from the config
-        default_value = config.get(f"default_{field}", None)
-        
-        # For fields that have min/max bounds
-        min_field = f"min_{field}"
-        max_field = f"max_{field}"
-        
-        # Check if min/max constraints exist for this field
-        has_constraints = min_field in config or max_field in config
-        
-        if not has_constraints:
-            # For fields without constraints, just return the value from config
-            return default_value
-        
-        # Get min and max values from config (if they exist)
-        min_value = config.get(min_field, None)
-        max_value = config.get(max_field, None)
-        
-        # If user provided options with this field
-        if options and field in options:
-            user_value = options[field]
-            
-            # Apply constraints if they exist
-            if min_value is not None and user_value < min_value:
-                return min_value
-            if max_value is not None and user_value > max_value:
-                return max_value
-                
-            # If within bounds, use the user's value
-            return user_value
-        
-        # If no user value, return the default
-        return default_value
-
-    async def _generate_clip_prompt(self, video_id: str, title: str, description: str) -> str:
-        """Generate a new prompt for the next clip based on event history"""
-        events = self.video_events.get(video_id, [])
-        events_json = "\n".join(json.dumps(event) for event in events)
-        
-        prompt = f"""# Context and task
-Please write the caption for a new clip.
-
-# Instructions
-1. Consider the video context and recent events
-2. Create a natural progression from previous clips
-3. Take into account user suggestions (chat messages) into the scene
-4. Don't generate hateful, political, violent or sexual content
-5. Keep visual consistency with previous clips (in most cases you should repeat the same exact description of the location, characters etc but only change a few elements. If this is a webcam scenario, don't touch the camera orientation or focus)
-6. Return ONLY the caption text, no additional formatting or explanation
-7. Write in English, about 200 words.
-8. Keep the visual style consistant, but content as well (repeat the style, character, locations, appearance etc.. across scenes, when it makes sense).
-8. Your caption must describe visual elements of the scene in details, including: camera angle and focus, people's appearance, age, look, costumes, clothes, the location visual characteristics and geometry, lighting, action, objects, weather, textures, lighting.
-
-# Examples
-Here is a demo scenario, with fake data:
-{{"time": "2024-11-29T13:36:15Z", "event": "new_stream_clip", "caption": "webcam view of a beautiful park, squirrels are playing in the lush grass, blablabla etc... (rest omitted for brevity)"}}
-{{"time": "2024-11-29T13:36:20Z", "event": "new_chat_message", "username": "MonkeyLover89", "data": "hi"}}
-{{"time": "2024-11-29T13:36:25Z", "event": "new_chat_message", "username": "MonkeyLover89", "data": "more squirrels plz"}}
-{{"time": "2024-11-29T13:36:26Z", "event": "new_stream_clip", "caption": "webcam view of a beautiful park, a lot of squirrels are playing in the lush grass, blablabla etc... (rest omitted for brevity)"}}
-
-# Real scenario and data
-
-We are inside a video titled "{title}"
-The video is described by: "{description}".
-Here is a summary of the {len(events)} most recent events:
-{events_json}
-
-# Your response
-Your caption:"""
-
-        try:
-            response = await asyncio.get_event_loop().run_in_executor(
-                None,
-                lambda: self.inference_client.text_generation(
-                    prompt,
-                    model=TEXT_MODEL,
-                    max_new_tokens=200,
-                    temperature=0.7
-                )
-            )
-            
-            # Clean up the response
-            caption = response.strip()
-            if caption.lower().startswith("caption:"):
-                caption = caption[8:].strip()
-                
-            return caption
-            
-        except Exception as e:
-            logger.error(f"Error generating clip prompt: {str(e)}")
-            # Fallback to original description if prompt generation fails
-            return description
-
-    async def generate_video_thumbnail(self, title: str, description: str, video_prompt_prefix: str, options: dict, user_role: UserRole = 'anon') -> str:
-        """
-        Generate a short, low-resolution video thumbnail for search results and previews.
-        Optimized for quick generation and low resource usage.
-        """
-        video_id = options.get('video_id', str(uuid.uuid4()))
-        seed = options.get('seed', generate_seed())
-        request_id = str(uuid.uuid4())[:8]  # Generate a short ID for logging
-        
-        logger.info(f"[{request_id}] Starting video thumbnail generation for video_id: {video_id}")
-        logger.info(f"[{request_id}] Title: '{title}', User role: {user_role}")
-        
-        # Create a more concise prompt for the thumbnail
-        clip_caption = f"{video_prompt_prefix} - {title.strip()}"
-        
-        # Add the thumbnail generation to event history
-        self._add_event(video_id, {
-            "time": datetime.datetime.utcnow().isoformat() + "Z",
-            "event": "thumbnail_generation",
-            "caption": clip_caption,
-            "seed": seed,
-            "request_id": request_id
-        })
-        
-        # Use a shorter prompt for thumbnails
-        prompt = f"{clip_caption}, {POSITIVE_PROMPT_SUFFIX}"
-        logger.info(f"[{request_id}] Using prompt: '{prompt}'")
-        
-        # Specialized configuration for thumbnails - smaller size, single frame
-        width = 512  # Reduced size for thumbnails
-        height = 288  # 16:9 aspect ratio
-        num_frames = THUMBNAIL_FRAMES  # Just one frame for static thumbnail
-        num_inference_steps = 4  # Fewer steps for faster generation
-        frame_rate = 25  # Standard frame rate
-        
-        # Optionally override with options if specified
-        width = options.get('width', width)
-        height = options.get('height', height)
-        num_frames = options.get('num_frames', num_frames)
-        num_inference_steps = options.get('num_inference_steps', num_inference_steps)
-        frame_rate = options.get('frame_rate', frame_rate)
-        
-        logger.info(f"[{request_id}] Configuration: width={width}, height={height}, frames={num_frames}, steps={num_inference_steps}, fps={frame_rate}")
-        
-        # Add thumbnail-specific tag to help debugging and metrics
-        options['thumbnail'] = True
-        
-        # Check for available endpoints before attempting generation
-        available_endpoints = sum(1 for ep in self.endpoint_manager.endpoints 
-                               if not ep.busy and time.time() > ep.error_until)
-        logger.info(f"[{request_id}] Available endpoints: {available_endpoints}/{len(self.endpoint_manager.endpoints)}")
-        
-        if available_endpoints == 0:
-            logger.error(f"[{request_id}] No available endpoints for thumbnail generation")
-            return ""
-        
-        # Use the same logic as regular video generation but with thumbnail settings
-        try:
-            # logger.info(f"[{request_id}] Generating thumbnail for video {video_id} with seed {seed}")
-            
-            start_time = time.time()
-            # Rest of thumbnail generation logic same as regular video but with optimized settings
-            result = await self._generate_video_content(
-                prompt=prompt,
-                negative_prompt=options.get('negative_prompt', NEGATIVE_PROMPT),
-                width=width,
-                height=height,
-                num_frames=num_frames,
-                num_inference_steps=num_inference_steps,
-                frame_rate=frame_rate,
-                seed=seed,
-                options=options,
-                user_role=user_role
-            )
-            duration = time.time() - start_time
-            
-            if result:
-                data_length = len(result)
-                logger.info(f"[{request_id}] Successfully generated thumbnail in {duration:.2f}s, data length: {data_length} chars")
-                return result
-            else:
-                logger.error(f"[{request_id}] Empty result returned from video generation")
-                return ""
-            
-        except Exception as e:
-            logger.error(f"[{request_id}] Error generating thumbnail: {e}")
-            if hasattr(e, "__traceback__"):
-                import traceback
-                logger.error(f"[{request_id}] Traceback: {traceback.format_exc()}")
-            return ""  # Return empty string instead of raising to avoid crashes
-    
-    async def generate_video(self, title: str, description: str, video_prompt_prefix: str, options: dict, user_role: UserRole = 'anon') -> str:
-        """Generate video using available space from pool"""
-        video_id = options.get('video_id', str(uuid.uuid4()))
-        
-        # Generate a new prompt based on event history
-        #clip_caption = await self._generate_clip_prompt(video_id, title, description)
-        clip_caption = f"{video_prompt_prefix} - {title.strip()} - {description.strip()}"
-
-        # Add the new clip to event history
-        self._add_event(video_id, {
-            "time": datetime.datetime.utcnow().isoformat() + "Z",
-            "event": "new_stream_clip",
-            "caption": clip_caption
-        })
-
-        # Use the generated caption as the prompt
-        prompt = f"{clip_caption}, {POSITIVE_PROMPT_SUFFIX}"
-        
-        # Get the config values based on user role
-        width = self.get_config_value(user_role, 'clip_width', options)
-        height = self.get_config_value(user_role, 'clip_height', options)
-        num_frames = self.get_config_value(user_role, 'num_frames', options)
-        num_inference_steps = self.get_config_value(user_role, 'num_inference_steps', options)
-        frame_rate = self.get_config_value(user_role, 'clip_framerate', options)
-        
-        # Get orientation from options
-        orientation = options.get('orientation', 'LANDSCAPE')
-        
-        # Adjust width and height based on orientation if needed
-        if orientation == 'PORTRAIT' and width > height:
-            # Swap width and height for portrait orientation
-            width, height = height, width
-            # logger.info(f"Orientation: {orientation}, swapped dimensions to width={width}, height={height}")
-        elif orientation == 'LANDSCAPE' and height > width:
-            # Swap height and width for landscape orientation
-            height, width = width, height
-            # logger.info(f"generate_video()  Orientation: {orientation}, swapped dimensions to width={width}, height={height}, steps={num_inference_steps}, fps={frame_rate} | role: {user_role}")
-        else:
-            # logger.info(f"generate_video()  Orientation: {orientation}, using original dimensions width={width}, height={height}, steps={num_inference_steps}, fps={frame_rate} | role: {user_role}")
-            pass
-        
-        # Generate the video with standard settings
-        return await self._generate_video_content(
-            prompt=prompt,
-            negative_prompt=options.get('negative_prompt', NEGATIVE_PROMPT),
-            width=width,
-            height=height,
-            num_frames=num_frames,
-            num_inference_steps=num_inference_steps,
-            frame_rate=frame_rate,
-            seed=options.get('seed', 42),
-            options=options,
-            user_role=user_role
-        )
-        
-    async def _generate_video_content(self, prompt: str, negative_prompt: str, width: int, 
-                                     height: int, num_frames: int, num_inference_steps: int, 
-                                     frame_rate: int, seed: int, options: dict, user_role: UserRole) -> str:
-        """
-        Internal method to generate video content with specific parameters.
-        Used by both regular video generation and thumbnail generation.
-        """
-        is_thumbnail = options.get('thumbnail', False)
-        request_id = options.get('request_id', str(uuid.uuid4())[:8])  # Get or generate request ID
-        video_id = options.get('video_id', 'unknown')
-        
-        # logger.info(f"[{request_id}] Generating {'thumbnail' if is_thumbnail else 'video'} for video {video_id} with seed {seed}")
-        
-        json_payload = {
-            "inputs": {
-                "prompt": prompt,
-            },
-            "parameters": {
-                # ------------------- settings for LTX-Video -----------------------
-                "negative_prompt": negative_prompt,
-                "width": width,
-                "height": height,
-                "num_frames": num_frames,
-                "num_inference_steps": num_inference_steps,
-                "guidance_scale": options.get('guidance_scale', GUIDANCE_SCALE),
-                "seed": seed,
-            
-                # ------------------- settings for Varnish -----------------------
-                "double_num_frames": False,  # <- False for real-time generation
-                "fps": frame_rate,
-                "super_resolution": False,  # <- False for real-time generation
-                "grain_amount": 0,  # No film grain (on low-res, low-quality generation the effects aren't worth it + it adds weight to the MP4 payload)
-            }
-        }
-        
-        # Add thumbnail flag to help with metrics and debugging
-        if is_thumbnail:
-            json_payload["metadata"] = {
-                "is_thumbnail": True,
-                "thumbnail_version": "1.0",
-                "request_id": request_id
-            }
-
-        # logger.info(f"[{request_id}] Waiting for an available endpoint...")
-        async with self.endpoint_manager.get_endpoint() as endpoint:
-            # logger.info(f"[{request_id}] Using endpoint {endpoint.id} for generation")
-            
-            try:
-                async with ClientSession() as session:
-                    #logger.info(f"[{request_id}] Sending request to endpoint {endpoint.id}: {endpoint.url}")
-                    start_time = time.time()
-                    
-                    # Proceed with actual request
-                    async with session.post(
-                        endpoint.url,
-                        headers={
-                            "Accept": "application/json",
-                            "Authorization": f"Bearer {HF_TOKEN}",
-                            "Content-Type": "application/json",
-                            "X-Request-ID": request_id  # Add request ID to headers
-                        },
-                        json=json_payload,
-                        timeout=12  # Extended timeout for thumbnails (was 8s)
-                    ) as response:
-                        request_duration = time.time() - start_time
-                        #logger.info(f"[{request_id}] Received response from endpoint {endpoint.id} in {request_duration:.2f}s: HTTP {response.status}")
-                        
-                        if response.status != 200:
-                            error_text = await response.text()
-                            logger.error(f"[{request_id}] Failed response: {error_text}")
-                            # Mark endpoint as in error state
-                            await self._mark_endpoint_error(endpoint)
-                            if "paused" in error_text:
-                                logger.error(f"[{request_id}] Endpoint is paused")
-                                return ""
-                            raise Exception(f"Video generation failed: HTTP {response.status} - {error_text}")
-                        
-                        result = await response.json()
-                        #logger.info(f"[{request_id}] Successfully parsed JSON response")
-                        
-                        if "error" in result:
-                            error_msg = result['error']
-                            logger.error(f"[{request_id}] Error in response: {error_msg}")
-                            # Mark endpoint as in error state
-                            await self._mark_endpoint_error(endpoint)
-                            if "paused" in str(error_msg).lower():
-                                logger.error(f"[{request_id}] Endpoint is paused")
-                                return ""
-                            raise Exception(f"Video generation failed: {error_msg}")
-                        
-                        video_data_uri = result.get("video")
-                        if not video_data_uri:
-                            logger.error(f"[{request_id}] No video data in response")
-                            # Mark endpoint as in error state
-                            await self._mark_endpoint_error(endpoint)
-                            raise Exception("No video data in response")
-                        
-                        # Get data size
-                        data_size = len(video_data_uri)
-                        #logger.info(f"[{request_id}] Received video data: {data_size} chars")
-                        
-                        # Reset error count on successful call
-                        endpoint.error_count = 0
-                        endpoint.error_until = 0
-                        
-                        return video_data_uri
-                        
-            except asyncio.TimeoutError:
-                # Handle timeout specifically
-                logger.error(f"[{request_id}] Timeout occurred after {time.time() - start_time:.2f}s")
-                await self._mark_endpoint_error(endpoint, is_timeout=True)
-                return ""
-            except Exception as e:
-                # Handle all other exceptions
-                logger.error(f"[{request_id}] Exception during video generation: {str(e)}")
-                if not isinstance(e, asyncio.TimeoutError):  # Already handled above
-                    await self._mark_endpoint_error(endpoint)
-                return ""
-                
-    async def _mark_endpoint_error(self, endpoint: Endpoint, is_timeout: bool = False):
-        """Mark an endpoint as being in error state with exponential backoff"""
-        async with self.endpoint_manager.lock:
-            endpoint.error_count += 1
-            
-            # Calculate backoff time exponentially based on error count
-            # Start with 15 seconds, then 30, 60, etc. up to a max of 5 minutes
-            # Using shorter backoffs since generation should be fast
-            backoff_seconds = min(15 * (2 ** (endpoint.error_count - 1)), 300)
-            
-            # Add extra backoff for timeouts which are more indicative of serious issues
-            if is_timeout:
-                backoff_seconds *= 2
-                
-            endpoint.error_until = time.time() + backoff_seconds
-            
-            logger.warning(
-                f"Endpoint {endpoint.id} marked as in error state (count: {endpoint.error_count}, "
-                f"unavailable until: {datetime.datetime.fromtimestamp(endpoint.error_until).strftime('%H:%M:%S')})"
-            )
-
-
-    async def handle_chat_message(self, data: dict, ws: web.WebSocketResponse) -> dict:
-        """Process and broadcast a chat message"""
-        video_id = data.get('videoId')
-        request_id = data.get('requestId')
-        
-        if not video_id:
-            return {
-                'action': 'chat_message',
-                'requestId': request_id,
-                'success': False,
-                'error': 'No video ID provided'
-            }
-
-        # Add chat message to event history
-        self._add_event(video_id, {
-            "time": datetime.datetime.utcnow().isoformat() + "Z",
-            "event": "new_chat_message",
-            "username": data.get('username', 'Anonymous'),
-            "data": data.get('content', '')
-        })
-
-        room = self.chat_rooms[video_id]
-        message_data = {k: v for k, v in data.items() if k != '_ws'}
-        room.add_message(message_data)
-        
-        for client in room.connected_clients:
-            if client != ws:
-                try:
-                    await client.send_json({
-                        'action': 'chat_message',
-                        'broadcast': True,
-                        **message_data
-                    })
-                except Exception as e:
-                    logger.error(f"Failed to broadcast to client: {e}")
-                    room.connected_clients.remove(client)
-        
-        return {
-            'action': 'chat_message',
-            'requestId': request_id,
-            'success': True,
-            'message': message_data
-        }
-
-    async def handle_join_chat(self, data: dict, ws: web.WebSocketResponse) -> dict:
-        """Handle a request to join a chat room"""
-        video_id = data.get('videoId')
-        request_id = data.get('requestId')
-        
-        if not video_id:
-            return {
-                'action': 'join_chat',
-                'requestId': request_id,
-                'success': False,
-                'error': 'No video ID provided'
-            }
-
-        room = self.chat_rooms[video_id]
-        room.connected_clients.add(ws)
-        recent_messages = room.get_recent_messages()
-        
-        return {
-            'action': 'join_chat',
-            'requestId': request_id,
-            'success': True,
-            'messages': recent_messages
-        }
-
-    async def handle_leave_chat(self, data: dict, ws: web.WebSocketResponse) -> dict:
-        """Handle a request to leave a chat room"""
-        video_id = data.get('videoId')
-        request_id = data.get('requestId')
-        
-        if not video_id:
-            return {
-                'action': 'leave_chat',
-                'requestId': request_id,
-                'success': False,
-                'error': 'No video ID provided'
-            }
-
-        room = self.chat_rooms[video_id]
-        if ws in room.connected_clients:
-            room.connected_clients.remove(ws)
-        
-        return {
-            'action': 'leave_chat',
-            'requestId': request_id,
-            'success': True
-        }

reference_example/api_metrics.py

@@ -1,185 +0,0 @@
-import time
-import logging
-import asyncio
-from collections import defaultdict
-from typing import Dict, List, Set, Optional
-import datetime
-
-logger = logging.getLogger(__name__)
-
-class MetricsTracker:
-    """
-    Tracks usage metrics across the API server.
-    """
-    def __init__(self):
-        # Total metrics since server start
-        self.total_requests = {
-            'chat': 0,
-            'video': 0, 
-            'search': 0,
-            'other': 0,
-        }
-        
-        # Per-user metrics
-        self.user_metrics = defaultdict(lambda: {
-            'requests': {
-                'chat': 0,
-                'video': 0,
-                'search': 0,
-                'other': 0,
-            },
-            'first_seen': time.time(),
-            'last_active': time.time(),
-            'role': 'anon'
-        })
-        
-        # Rate limiting buckets (per minute)
-        self.rate_limits = {
-            'anon': {
-                'video': 30,
-                'search': 45,
-                'chat': 90,
-                'other': 45
-            },
-            'normal': {
-                'video': 60,
-                'search': 90,
-                'chat': 180,
-                'other': 90 
-            },
-            'pro': {
-                'video': 120,
-                'search': 180,
-                'chat': 300,
-                'other': 180
-            },
-            'admin': {
-                'video': 240,
-                'search': 360,
-                'chat': 450,
-                'other': 360
-            }
-        }
-        
-        # Minute-based rate limiting buckets
-        self.time_buckets = defaultdict(lambda: defaultdict(lambda: defaultdict(int)))
-        
-        # Lock for thread safety
-        self.lock = asyncio.Lock()
-        
-        # Track concurrent sessions by IP
-        self.ip_sessions = defaultdict(set)
-        
-        # Server start time
-        self.start_time = time.time()
-        
-    async def record_request(self, user_id: str, ip: str, request_type: str, role: str):
-        """Record a request for metrics and rate limiting"""
-        async with self.lock:
-            # Update total metrics
-            if request_type in self.total_requests:
-                self.total_requests[request_type] += 1
-            else:
-                self.total_requests['other'] += 1
-                
-            # Update user metrics
-            user_data = self.user_metrics[user_id]
-            user_data['last_active'] = time.time()
-            user_data['role'] = role
-            
-            if request_type in user_data['requests']:
-                user_data['requests'][request_type] += 1
-            else:
-                user_data['requests']['other'] += 1
-                
-            # Update time bucket for rate limiting
-            current_minute = int(time.time() / 60)
-            self.time_buckets[user_id][current_minute][request_type] += 1
-            
-            # Clean up old time buckets (keep only last 10 minutes)
-            cutoff = current_minute - 10
-            for minute in list(self.time_buckets[user_id].keys()):
-                if minute < cutoff:
-                    del self.time_buckets[user_id][minute]
-    
-    def register_session(self, user_id: str, ip: str):
-        """Register a new session for an IP address"""
-        self.ip_sessions[ip].add(user_id)
-        
-    def unregister_session(self, user_id: str, ip: str):
-        """Unregister a session when it disconnects"""
-        if user_id in self.ip_sessions[ip]:
-            self.ip_sessions[ip].remove(user_id)
-            if not self.ip_sessions[ip]:
-                del self.ip_sessions[ip]
-    
-    def get_session_count_for_ip(self, ip: str) -> int:
-        """Get the number of active sessions for an IP address"""
-        return len(self.ip_sessions.get(ip, set()))
-    
-    async def is_rate_limited(self, user_id: str, request_type: str, role: str) -> bool:
-        """Check if a user is currently rate limited for a request type"""
-        async with self.lock:
-            current_minute = int(time.time() / 60)
-            prev_minute = current_minute - 1
-            
-            # Count requests in current and previous minute
-            current_count = self.time_buckets[user_id][current_minute][request_type]
-            prev_count = self.time_buckets[user_id][prev_minute][request_type]
-            
-            # Calculate requests per minute rate (weighted average)
-            # Weight current minute more as it's more recent
-            rate = (current_count * 0.7) + (prev_count * 0.3)
-            
-            # Get rate limit based on user role
-            limit = self.rate_limits.get(role, self.rate_limits['anon']).get(
-                request_type, self.rate_limits['anon']['other'])
-            
-            # Check if rate exceeds limit
-            return rate >= limit
-    
-    def get_metrics(self) -> Dict:
-        """Get a snapshot of current metrics"""
-        active_users = {
-            'total': len(self.user_metrics),
-            'anon': 0,
-            'normal': 0,
-            'pro': 0,
-            'admin': 0,
-        }
-        
-        # Count active users in the last 5 minutes
-        active_cutoff = time.time() - (5 * 60)
-        for user_data in self.user_metrics.values():
-            if user_data['last_active'] >= active_cutoff:
-                active_users[user_data['role']] += 1
-        
-        return {
-            'uptime_seconds': int(time.time() - self.start_time),
-            'total_requests': dict(self.total_requests),
-            'active_users': active_users,
-            'active_ips': len(self.ip_sessions),
-            'timestamp': datetime.datetime.now().isoformat()
-        }
-
-    def get_detailed_metrics(self) -> Dict:
-        """Get detailed metrics including per-user data"""
-        metrics = self.get_metrics()
-        
-        # Add anonymized user metrics
-        user_list = []
-        for user_id, data in self.user_metrics.items():
-            # Skip users inactive for more than 1 hour
-            if time.time() - data['last_active'] > 3600:
-                continue
-                
-            user_list.append({
-                'id': user_id[:8] + '...',  # Anonymize ID
-                'role': data['role'],
-                'requests': data['requests'],
-                'active_ago': int(time.time() - data['last_active']),
-                'session_duration': int(time.time() - data['first_seen'])
-            })
-        
-        metrics['users'] = user_list
-        return metrics

reference_example/api_session.py

@@ -1,569 +0,0 @@
-import asyncio
-import logging
-from typing import Dict, Set
-from aiohttp import web, WSMsgType
-import json
-import time
-import datetime
-from api_core import VideoGenerationAPI
-
-logger = logging.getLogger(__name__)
-
-class UserSession:
-    """
-    Represents a user's session with the API.
-    Each WebSocket connection gets its own session with separate queues and rate limits.
-    """
-    def __init__(self, user_id: str, user_role: str, ws: web.WebSocketResponse, shared_api):
-        self.user_id = user_id
-        self.user_role = user_role
-        self.ws = ws
-        self.shared_api = shared_api  # For shared resources like endpoint manager
-        
-        # Create separate queues for this user session
-        self.chat_queue = asyncio.Queue()
-        self.video_queue = asyncio.Queue()
-        self.search_queue = asyncio.Queue()
-        self.simulation_queue = asyncio.Queue()  # New queue for description evolution
-        
-        # Track request counts and rate limits
-        self.request_counts = {
-            'chat': 0,
-            'video': 0,
-            'search': 0,
-            'simulation': 0  # New counter for simulation requests
-        }
-        
-        # Last request timestamps for rate limiting
-        self.last_request_times = {
-            'chat': time.time(),
-            'video': time.time(),
-            'search': time.time(),
-            'simulation': time.time()  # New timestamp for simulation requests
-        }
-        
-        # Session creation time
-        self.created_at = time.time()
-        
-        self.background_tasks = []
-        
-    async def start(self):
-        """Start all the queue processors for this session"""
-        # Start background tasks for handling different request types
-        self.background_tasks = [
-            asyncio.create_task(self._process_chat_queue()),
-            asyncio.create_task(self._process_video_queue()),
-            asyncio.create_task(self._process_search_queue()),
-            asyncio.create_task(self._process_simulation_queue())  # New worker for simulation requests
-        ]
-        logger.info(f"Started session for user {self.user_id} with role {self.user_role}")
-        
-    async def stop(self):
-        """Stop all background tasks for this session"""
-        for task in self.background_tasks:
-            task.cancel()
-        
-        try:
-            # Wait for tasks to complete cancellation
-            await asyncio.gather(*self.background_tasks, return_exceptions=True)
-        except asyncio.CancelledError:
-            pass
-        
-        logger.info(f"Stopped session for user {self.user_id}")
-        
-    async def _process_chat_queue(self):
-        """High priority queue for chat operations"""
-        while True:
-            data = await self.chat_queue.get()
-            try:
-                if data['action'] == 'join_chat':
-                    result = await self.shared_api.handle_join_chat(data, self.ws)
-                elif data['action'] == 'chat_message':
-                    result = await self.shared_api.handle_chat_message(data, self.ws)
-                elif data['action'] == 'leave_chat':
-                    result = await self.shared_api.handle_leave_chat(data, self.ws)
-                # Redirect thumbnail requests to process_generic_request for consistent handling
-                elif data['action'] == 'generate_video_thumbnail':
-                    # Pass to the generic request handler to maintain consistent logic
-                    await self.process_generic_request(data)
-                    # Skip normal response handling since process_generic_request already sends a response
-                    self.chat_queue.task_done()
-                    continue
-                else:
-                    raise ValueError(f"Unknown chat action: {data['action']}")
-                    
-                await self.ws.send_json(result)
-                
-                # Update metrics
-                self.request_counts['chat'] += 1
-                self.last_request_times['chat'] = time.time()
-                
-            except Exception as e:
-                logger.error(f"Error processing chat request for user {self.user_id}: {e}")
-                try:
-                    await self.ws.send_json({
-                        'action': data['action'],
-                        'requestId': data.get('requestId'),
-                        'success': False,
-                        'error': f'Chat error: {str(e)}'
-                    })
-                except Exception as send_error:
-                    logger.error(f"Error sending error response: {send_error}")
-            finally:
-                self.chat_queue.task_done()
-
-    async def _process_video_queue(self):
-        """Process multiple video generation requests in parallel for this user"""
-        from api_config import VIDEO_ROUND_ROBIN_ENDPOINT_URLS
-        
-        active_tasks = set()
-        # Set a per-user concurrent limit based on role
-        max_concurrent = len(VIDEO_ROUND_ROBIN_ENDPOINT_URLS)
-        if self.user_role == 'anon':
-            max_concurrent = min(2, max_concurrent)  # Limit anonymous users
-        elif self.user_role == 'normal':
-            max_concurrent = min(4, max_concurrent)  # Standard users
-        # Pro and admin can use all endpoints
-
-        async def process_single_request(data):
-            try:
-                title = data.get('title', '')
-                description = data.get('description', '')
-                video_prompt_prefix = data.get('video_prompt_prefix', '')
-                options = data.get('options', {})
-                
-                # Pass the user role to generate_video
-                video_data = await self.shared_api.generate_video(
-                    title, description, video_prompt_prefix, options, self.user_role
-                )
-                
-                result = {
-                    'action': 'generate_video',
-                    'requestId': data.get('requestId'),
-                    'success': True,
-                    'video': video_data,
-                }
-                
-                await self.ws.send_json(result)
-                
-                # Update metrics
-                self.request_counts['video'] += 1
-                self.last_request_times['video'] = time.time()
-                
-            except Exception as e:
-                logger.error(f"Error processing video request for user {self.user_id}: {e}")
-                try:
-                    await self.ws.send_json({
-                        'action': 'generate_video',
-                        'requestId': data.get('requestId'),
-                        'success': False,
-                        'error': f'Video generation error: {str(e)}'
-                    })
-                except Exception as send_error:
-                    logger.error(f"Error sending error response: {send_error}")
-            finally:
-                active_tasks.discard(asyncio.current_task())
-
-        while True:
-            # Clean up completed tasks
-            active_tasks = {task for task in active_tasks if not task.done()}
-            
-            # Start new tasks if we have capacity
-            while len(active_tasks) < max_concurrent:
-                try:
-                    # Use try_get to avoid blocking if queue is empty
-                    data = await asyncio.wait_for(self.video_queue.get(), timeout=0.1)
-                    
-                    # Create and start new task
-                    task = asyncio.create_task(process_single_request(data))
-                    active_tasks.add(task)
-                    
-                except asyncio.TimeoutError:
-                    # No items in queue, break inner loop
-                    break
-                except Exception as e:
-                    logger.error(f"Error creating video generation task for user {self.user_id}: {e}")
-                    break
-
-            # Wait a short time before checking queue again
-            await asyncio.sleep(0.1)
-
-            # Handle any completed tasks' errors
-            for task in list(active_tasks):
-                if task.done():
-                    try:
-                        await task
-                    except Exception as e:
-                        logger.error(f"Task failed with error for user {self.user_id}: {e}")
-                    active_tasks.discard(task)
-
-    async def _process_search_queue(self):
-        """Medium priority queue for search operations"""
-        while True:
-            try:
-                data = await self.search_queue.get()
-                request_id = data.get('requestId')
-                query = data.get('query', '').strip()
-                attempt_count = data.get('attemptCount', 0)
-
-                # logger.info(f"Processing search request for user {self.user_id}, attempt={attempt_count}")
-
-                if not query:
-                    logger.warning(f"Empty query received in request from user {self.user_id}: {data}")
-                    result = {
-                        'action': 'search',
-                        'requestId': request_id,
-                        'success': False,
-                        'error': 'No search query provided'
-                    }
-                else:
-                    try:
-                        search_result = await self.shared_api.search_video(
-                            query,
-                            attempt_count=attempt_count
-                        )
-                        
-                        if search_result:
-                            # logger.info(f"Search successful for user {self.user_id}, query '{query}'")
-                            result = {
-                                'action': 'search',
-                                'requestId': request_id,
-                                'success': True,
-                                'result': search_result
-                            }
-                        else:
-                            # logger.warning(f"No results found for user {self.user_id}, query '{query}'")
-                            result = {
-                                'action': 'search',
-                                'requestId': request_id,
-                                'success': False,
-                                'error': 'No results found'
-                            }
-                    except Exception as e:
-                        logger.error(f"Search error for user {self.user_id}, (attempt {attempt_count}): {str(e)}")
-                        result = {
-                            'action': 'search',
-                            'requestId': request_id,
-                            'success': False,
-                            'error': f'Search error: {str(e)}'
-                        }
-
-                await self.ws.send_json(result)
-                
-                # Update metrics
-                self.request_counts['search'] += 1
-                self.last_request_times['search'] = time.time()
-                
-            except Exception as e:
-                logger.error(f"Error in search queue processor for user {self.user_id}: {str(e)}")
-                try:
-                    error_response = {
-                        'action': 'search',
-                        'requestId': data.get('requestId') if 'data' in locals() else None,
-                        'success': False,
-                        'error': f'Internal server error: {str(e)}'
-                    }
-                    await self.ws.send_json(error_response)
-                except Exception as send_error:
-                    logger.error(f"Error sending error response: {send_error}")
-            finally:
-                if 'search_queue' in self.__dict__:
-                    self.search_queue.task_done()
-                    
-    async def _process_simulation_queue(self):
-        """Dedicated queue for video simulation requests"""
-        while True:
-            try:
-                data = await self.simulation_queue.get()
-                request_id = data.get('requestId')
-                
-                # Extract parameters from the request
-                video_id = data.get('video_id', '')
-                original_title = data.get('original_title', '')
-                original_description = data.get('original_description', '')
-                current_description = data.get('current_description', '')
-                condensed_history = data.get('condensed_history', '')
-                evolution_count = data.get('evolution_count', 0)
-                chat_messages = data.get('chat_messages', '')
-                
-                logger.info(f"Processing video simulation for user {self.user_id}, video_id={video_id}, evolution_count={evolution_count}")
-                
-                # Validate required parameters
-                if not original_title or not original_description or not current_description:
-                    result = {
-                        'action': 'simulate',
-                        'requestId': request_id,
-                        'success': False,
-                        'error': 'Missing required parameters'
-                    }
-                else:
-                    try:
-                        # Call the simulate method in the API
-                        simulation_result = await self.shared_api.simulate(
-                            original_title=original_title,
-                            original_description=original_description,
-                            current_description=current_description,
-                            condensed_history=condensed_history,
-                            evolution_count=evolution_count,
-                            chat_messages=chat_messages
-                        )
-                        
-                        result = {
-                            'action': 'simulate',
-                            'requestId': request_id,
-                            'success': True,
-                            'evolved_description': simulation_result['evolved_description'],
-                            'condensed_history': simulation_result['condensed_history']
-                        }
-                    except Exception as e:
-                        logger.error(f"Error simulating video for user {self.user_id}, video_id={video_id}: {str(e)}")
-                        result = {
-                            'action': 'simulate',
-                            'requestId': request_id,
-                            'success': False,
-                            'error': f'Simulation error: {str(e)}'
-                        }
-                
-                await self.ws.send_json(result)
-                
-                # Update metrics
-                self.request_counts['simulation'] += 1
-                self.last_request_times['simulation'] = time.time()
-                
-            except Exception as e:
-                logger.error(f"Error in simulation queue processor for user {self.user_id}: {str(e)}")
-                try:
-                    error_response = {
-                        'action': 'simulate',
-                        'requestId': data.get('requestId') if 'data' in locals() else None,
-                        'success': False,
-                        'error': f'Internal server error: {str(e)}'
-                    }
-                    await self.ws.send_json(error_response)
-                except Exception as send_error:
-                    logger.error(f"Error sending error response: {send_error}")
-            finally:
-                if 'simulation_queue' in self.__dict__:
-                    self.simulation_queue.task_done()
-                    
-    async def process_generic_request(self, data: dict) -> None:
-        """Handle general requests that don't fit into specialized queues"""
-        try:
-            request_id = data.get('requestId')
-            action = data.get('action')
-            
-            def error_response(message: str):
-                return {
-                    'action': action,
-                    'requestId': request_id,
-                    'success': False,
-                    'error': message
-                }
-
-            if action == 'heartbeat':
-                # Include user role info in heartbeat response
-                await self.ws.send_json({
-                    'action': 'heartbeat',
-                    'requestId': request_id,
-                    'success': True,
-                    'user_role': self.user_role
-                })
-            
-            elif action == 'get_user_role':
-                # Return the user role information
-                await self.ws.send_json({
-                    'action': 'get_user_role',
-                    'requestId': request_id,
-                    'success': True,
-                    'user_role': self.user_role
-                })
-            
-            elif action == 'generate_caption':
-                title = data.get('params', {}).get('title')
-                description = data.get('params', {}).get('description')
-                
-                if not title or not description:
-                    await self.ws.send_json(error_response('Missing title or description'))
-                    return
-                    
-                caption = await self.shared_api.generate_caption(title, description)
-                await self.ws.send_json({
-                    'action': action,
-                    'requestId': request_id,
-                    'success': True,
-                    'caption': caption
-                })
-                
-            # evolve_description is now handled by the dedicated simulation queue processor
-                
-            elif action == 'generate_video_thumbnail':
-                title = data.get('title', '') or data.get('params', {}).get('title', '')
-                description = data.get('description', '') or data.get('params', {}).get('description', '')
-                video_prompt_prefix = data.get('video_prompt_prefix', '') or data.get('params', {}).get('video_prompt_prefix', '')
-                options = data.get('options', {}) or data.get('params', {}).get('options', {})
-                
-                if not title:
-                    await self.ws.send_json(error_response('Missing title for thumbnail generation'))
-                    return
-                
-                # Ensure the options include the thumbnail flag
-                options['thumbnail'] = True
-                
-                # Prioritize thumbnail generation with higher priority
-                options['priority'] = 'high'
-                
-                # Add small size settings if not already specified
-                if 'width' not in options:
-                    options['width'] = 512  # Default thumbnail width
-                if 'height' not in options:
-                    options['height'] = 288  # Default 16:9 aspect ratio
-                if 'num_frames' not in options:
-                    options['num_frames'] = 25  # 1 second @ 25fps
-                
-                # Let the API know this is a thumbnail for a specific video
-                options['video_id'] = data.get('video_id', f"thumbnail-{request_id}")
-                
-                logger.info(f"Generating thumbnail for video {options['video_id']} for user {self.user_id}")
-                
-                try:
-                    # Generate the thumbnail
-                    thumbnail_data = await self.shared_api.generate_video_thumbnail(
-                        title, description, video_prompt_prefix, options, self.user_role
-                    )
-                    
-                    # Respond with appropriate format based on the parameter names used in the request
-                    if 'thumbnailUrl' in data or 'thumbnailUrl' in data.get('params', {}):
-                        # Legacy format using thumbnailUrl
-                        await self.ws.send_json({
-                            'action': action,
-                            'requestId': request_id,
-                            'success': True,
-                            'thumbnailUrl': thumbnail_data or "",
-                        })
-                    else:
-                        # New format using thumbnail
-                        await self.ws.send_json({
-                            'action': action,
-                            'requestId': request_id,
-                            'success': True,
-                            'thumbnail': thumbnail_data,
-                        })
-                except Exception as e:
-                    logger.error(f"Error generating thumbnail: {str(e)}")
-                    await self.ws.send_json(error_response(f"Thumbnail generation failed: {str(e)}"))
-                
-            # Handle deprecated thumbnail actions
-            elif action == 'generate_thumbnail' or action == 'old_generate_thumbnail':
-                # Redirect to video thumbnail generation
-                logger.warning(f"Deprecated thumbnail action '{action}' used, redirecting to generate_video_thumbnail")
-                
-                # Extract parameters
-                title = data.get('title', '') or data.get('params', {}).get('title', '')
-                description = data.get('description', '') or data.get('params', {}).get('description', '')
-                
-                if not title or not description:
-                    await self.ws.send_json(error_response('Missing title or description'))
-                    return
-                
-                # Create a new request with the correct action
-                new_request = {
-                    'action': 'generate_video_thumbnail',
-                    'requestId': request_id,
-                    'title': title,
-                    'description': description,
-                    'options': {
-                        'width': 512,
-                        'height': 288,
-                        'thumbnail': True,
-                        'video_id': f"thumbnail-{request_id}"
-                    }
-                }
-                
-                # Process with the new action
-                await self.process_generic_request(new_request)
-                
-            else:
-                await self.ws.send_json(error_response(f'Unknown action: {action}'))
-                
-        except Exception as e:
-            logger.error(f"Error processing generic request for user {self.user_id}: {str(e)}")
-            try:
-                await self.ws.send_json({
-                    'action': data.get('action'),
-                    'requestId': data.get('requestId'),
-                    'success': False,
-                    'error': f'Internal server error: {str(e)}'
-                })
-            except Exception as send_error:
-                logger.error(f"Error sending error response: {send_error}")
-
-class SessionManager:
-    """
-    Manages all active user sessions and shared resources.
-    """
-    def __init__(self):
-        self.sessions = {}
-        self.shared_api = VideoGenerationAPI()  # Single instance for shared resources
-        self.session_lock = asyncio.Lock()
-    
-    async def create_session(self, user_id: str, user_role: str, ws: web.WebSocketResponse) -> UserSession:
-        """Create a new user session"""
-        async with self.session_lock:
-            # Create a new session for this user
-            session = UserSession(user_id, user_role, ws, self.shared_api)
-            await session.start()
-            self.sessions[user_id] = session
-            return session
-    
-    async def delete_session(self, user_id: str) -> None:
-        """Delete a user session and clean up resources"""
-        async with self.session_lock:
-            if user_id in self.sessions:
-                session = self.sessions[user_id]
-                await session.stop()
-                del self.sessions[user_id]
-                logger.info(f"Deleted session for user {user_id}")
-    
-    def get_session(self, user_id: str) -> UserSession:
-        """Get a user session if it exists"""
-        return self.sessions.get(user_id)
-    
-    async def close_all_sessions(self) -> None:
-        """Close all active sessions (used during shutdown)"""
-        async with self.session_lock:
-            for user_id, session in list(self.sessions.items()):
-                await session.stop()
-            self.sessions.clear()
-            logger.info("Closed all active sessions")
-    
-    @property
-    def session_count(self) -> int:
-        """Get the number of active sessions"""
-        return len(self.sessions)
-    
-    def get_session_stats(self) -> Dict:
-        """Get statistics about active sessions"""
-        stats = {
-            'total_sessions': len(self.sessions),
-            'by_role': {
-                'anon': 0,
-                'normal': 0,
-                'pro': 0,
-                'admin': 0
-            },
-            'requests': {
-                'chat': 0,
-                'video': 0,
-                'search': 0,
-                'simulation': 0
-            }
-        }
-        
-        for session in self.sessions.values():
-            stats['by_role'][session.user_role] += 1
-            stats['requests']['chat'] += session.request_counts['chat']
-            stats['requests']['video'] += session.request_counts['video']
-            stats['requests']['search'] += session.request_counts['search']
-            stats['requests']['simulation'] += session.request_counts['simulation']
-            
-        return stats

src/agent.py

@@ -5,11 +5,11 @@ from typing import Optional, Union
 import torch
 import torch.nn as nn
 
-from envs import TorchEnv, WorldModelEnv
-from models.actor_critic import ActorCritic, ActorCriticConfig, ActorCriticLossConfig
-from models.diffusion import Denoiser, DenoiserConfig, SigmaDistributionConfig
-from models.rew_end_model import RewEndModel, RewEndModelConfig
-from utils import extract_state_dict
+from src.envs import TorchEnv, WorldModelEnv
+from src.models.actor_critic import ActorCritic, ActorCriticConfig, ActorCriticLossConfig
+from src.models.diffusion import Denoiser, DenoiserConfig, SigmaDistributionConfig
+from src.models.rew_end_model import RewEndModel, RewEndModelConfig
+from src.utils import extract_state_dict
 
 
 @dataclass

src/data/dataset.py

@@ -13,7 +13,7 @@ from torch.utils.data import Dataset as TorchDataset
 from .episode import Episode
 from .segment import Segment, SegmentId
 from .utils import make_segment
-from utils import StateDictMixin
+from src.utils import StateDictMixin
 
 
 class Dataset(StateDictMixin, TorchDataset):

src/game/dataset_env.py

@@ -3,7 +3,7 @@ from typing import Any, Dict, List, Tuple
 import torch
 from torch import Tensor
 
-from data import Dataset
+from src.data import Dataset
 
 
 class DatasetEnv:

src/game/game.py

@@ -4,7 +4,7 @@ import numpy as np
 import pygame
 from PIL import Image
 
-from player.action_processing import GameAction
+from src.player.action_processing import GameAction
 from .dataset_env import DatasetEnv
 from .play_env import PlayEnv
 

src/game/play_env.py

@@ -7,11 +7,11 @@ import pygame
 import torch
 from torch import Tensor
 
-from agent import Agent
-from player.action_processing import GameAction, decode_game_action, encode_game_action, print_game_action
-from player.keymap import GAME_KEYMAP
-from data import Dataset, Episode
-from envs import WorldModelEnv
+from src.agent import Agent
+from src.player.action_processing import GameAction, decode_game_action, encode_game_action, print_game_action
+from src.player.keymap import GAME_KEYMAP
+from src.data import Dataset, Episode
+from src.envs import WorldModelEnv
 
 
 NamedEnv = namedtuple("NamedEnv", "name env")

src/main.py

@@ -8,8 +8,8 @@ import torch
 from torch.distributed import init_process_group, destroy_process_group
 import torch.multiprocessing as mp
 
-from trainer import Trainer
-from utils import skip_if_run_is_over
+from src.trainer import Trainer
+from src.utils import skip_if_run_is_over
 
 
 OmegaConf.register_new_resolver("eval", eval)

src/play.py

@@ -7,9 +7,9 @@ from hydra.utils import instantiate
 from omegaconf import DictConfig, OmegaConf
 import torch
 
-from agent import Agent
-from envs import WorldModelEnv
-from game import Game, PlayEnv
+from src.agent import Agent
+from src.envs import WorldModelEnv
+from src.game import Game, PlayEnv
 
 
 OmegaConf.register_new_resolver("eval", eval)