#!/usr/bin/env python3.11 """ AI Video Generation Swarm - Main Application Lazy loading version - no model loading at startup """ import os import sys import json import uuid import time import asyncio import hashlib from datetime import datetime, timedelta from typing import Optional, Dict, Any, List from contextlib import asynccontextmanager import torch from fastapi import FastAPI, HTTPException, Depends, WebSocket, WebSocketDisconnect, UploadFile, File, Form from fastapi.responses import HTMLResponse, FileResponse, JSONResponse from fastapi.staticfiles import StaticFiles from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials from jose import JWTError, jwt from passlib.context import CryptContext import boto3 from botocore.config import Config # ─────────────────────────────────────────────────────────────── # CONFIGURATION # ─────────────────────────────────────────────────────────────── HF_TOKEN = os.getenv("HF_TOKEN", "") PASS_KEY = os.getenv("PASS_KEY", "default_secret_change_me") R2_ACCOUNT_ID = os.getenv("R2_ACCOUNT_ID", "") R2_BUCKET_NAME = os.getenv("R2_BUCKET_NAME", "video-ai-swarm") R2_ACCESS_KEY = os.getenv("R2_ACCESS_KEY_ID", "") R2_SECRET_KEY = os.getenv("R2_SECRET_ACCESS_KEY", "") R2_PUBLIC_DOMAIN = os.getenv("R2_PUBLIC_DOMAIN", "") SECRET_KEY = hashlib.sha256(PASS_KEY.encode()).hexdigest() ALGORITHM = "HS256" ACCESS_TOKEN_EXPIRE_HOURS = 24 GPU_MAP = { "orchestrator": 7, "video_primary": 0, "video_backup": 1, "face_restore": 2, "frame_interpolate": 3, "video_upscale": 4, "image_generator": 5, "audio_generator": 6, "quality_evaluator": 6, } # ─────────────────────────────────────────────────────────────── # AUTHENTICATION # ─────────────────────────────────────────────────────────────── pwd_context = CryptContext(schemes=["bcrypt"], deprecated="auto") security = HTTPBearer() def create_access_token(data: dict, expires_delta: Optional[timedelta] = None): to_encode = data.copy() expire = datetime.utcnow() + (expires_delta or timedelta(hours=ACCESS_TOKEN_EXPIRE_HOURS)) to_encode.update({"exp": expire}) return jwt.encode(to_encode, SECRET_KEY, algorithm=ALGORITHM) def verify_token(credentials: HTTPAuthorizationCredentials = Depends(security)): token = credentials.credentials try: payload = jwt.decode(token, SECRET_KEY, algorithms=[ALGORITHM]) return payload except JWTError: raise HTTPException(status_code=401, detail="Invalid or expired token") def verify_pass_key(key: str) -> bool: return key == PASS_KEY # ─────────────────────────────────────────────────────────────── # R2 CLIENT # ─────────────────────────────────────────────────────────────── class R2Client: def __init__(self): try: self.s3 = boto3.client( "s3", endpoint_url=f"https://{R2_ACCOUNT_ID}.r2.cloudflarestorage.com", region_name="auto", aws_access_key_id=R2_ACCESS_KEY, aws_secret_access_key=R2_SECRET_KEY, config=Config(signature_version="s3v4") ) except Exception as e: print(f"[R2] Warning: R2 not configured: {e}") self.s3 = None self.bucket = R2_BUCKET_NAME self.public_domain = R2_PUBLIC_DOMAIN def upload_file(self, local_path: str, key: str): if not self.s3: return None self.s3.upload_file(local_path, self.bucket, key) return f"{self.public_domain}/{key}" def list_videos(self, prefix: str = "outputs/") -> List[dict]: if not self.s3: return [] try: response = self.s3.list_objects_v2(Bucket=self.bucket, Prefix=prefix) items = [] for obj in response.get("Contents", []): key = obj["Key"] if key.endswith(".mp4"): items.append({ "key": key, "url": f"{self.public_domain}/{key}", "size": obj["Size"], "last_modified": obj["LastModified"].isoformat() }) return sorted(items, key=lambda x: x["last_modified"], reverse=True) except Exception as e: print(f"[R2] List error: {e}") return [] r2_client = R2Client() # ─────────────────────────────────────────────────────────────── # JOB MANAGER & STATE # ─────────────────────────────────────────────────────────────── class JobManager: def __init__(self): self.jobs: Dict[str, dict] = {} self.active_websockets: List[WebSocket] = [] self.log_websockets: List[WebSocket] = [] self.logs: List[dict] = [] def create_job(self, params: dict) -> str: job_id = str(uuid.uuid4())[:8] self.jobs[job_id] = { "id": job_id, "status": "queued", "progress": 0, "stage": "pending", "params": params, "created_at": datetime.utcnow().isoformat(), "completed_at": None, "output_url": None, "logs": [], "retries": 0, } return job_id def update_job(self, job_id: str, **kwargs): if job_id in self.jobs: self.jobs[job_id].update(kwargs) def add_log(self, level: str, agent: str, message: str): entry = { "timestamp": datetime.utcnow().isoformat(), "level": level, "agent": agent, "message": message } self.logs.append(entry) if len(self.logs) > 10000: self.logs = self.logs[-10000:] return entry job_manager = JobManager() # ─────────────────────────────────────────────────────────────── # WEBSOCKET MANAGER # ─────────────────────────────────────────────────────────────── class ConnectionManager: def __init__(self): self.active_connections: List[WebSocket] = [] async def connect(self, websocket: WebSocket): await websocket.accept() self.active_connections.append(websocket) def disconnect(self, websocket: WebSocket): if websocket in self.active_connections: self.active_connections.remove(websocket) async def broadcast(self, message: dict): disconnected = [] for conn in self.active_connections: try: await conn.send_json(message) except: disconnected.append(conn) for conn in disconnected: self.disconnect(conn) ws_manager = ConnectionManager() # ─────────────────────────────────────────────────────────────── # LAZY AGENT LOADING (no loading at startup!) # ─────────────────────────────────────────────────────────────── _agents_loaded = False _orchestrator = None _agent_lock = asyncio.Lock() async def load_agents(): """Lazy load all agents - called on first generation request""" global _agents_loaded, _orchestrator async with _agent_lock: if _agents_loaded: return print("[INFO] Starting lazy agent loading...") print(f"[INFO] CUDA available: {torch.cuda.is_available()}") print(f"[INFO] GPU count: {torch.cuda.device_count()}") sys.path.insert(0, "/workspace") try: from agents.orchestrator import OrchestratorAgent _orchestrator = OrchestratorAgent(job_manager, ws_manager, r2_client, GPU_MAP) _agents_loaded = True print("[INFO] All agents loaded successfully") except Exception as e: print(f"[ERROR] Failed to load agents: {e}") import traceback traceback.print_exc() _agents_loaded = False # ─────────────────────────────────────────────────────────────── # PIPELINE EXECUTOR # ─────────────────────────────────────────────────────────────── async def run_pipeline(job_id: str): """Execute the full video generation pipeline""" job = job_manager.jobs[job_id] try: await load_agents() if _orchestrator is None: raise RuntimeError("Orchestrator not available") await _orchestrator.execute(job_id, job["params"]) except Exception as e: job_manager.update_job(job_id, status="failed", error=str(e)) job_manager.add_log("ERROR", "ORCHESTRATOR", f"Job {job_id} failed: {e}") await ws_manager.broadcast({ "type": "job_failed", "job_id": job_id, "error": str(e) }) # ─────────────────────────────────────────────────────────────── # FASTAPI APP # ─────────────────────────────────────────────────────────────── @asynccontextmanager async def lifespan(app: FastAPI): print("="*60) print("[INFO] AI Video Swarm starting up...") print("[INFO] Models will be loaded LAZILY on first request") print("[INFO] This prevents startup timeout") print("="*60) yield print("[INFO] AI Video Swarm shutting down...") app = FastAPI(title="AI Video Swarm", version="1.0.0", lifespan=lifespan) # ─────────────────────────────────────────────────────────────── # HTML ROUTES # ─────────────────────────────────────────────────────────────── @app.get("/", response_class=HTMLResponse) async def root(): with open("/workspace/pass.html", "r", encoding="utf-8") as f: return HTMLResponse(content=f.read()) @app.get("/app", response_class=HTMLResponse) async def main_app(): with open("/workspace/index.html", "r", encoding="utf-8") as f: return HTMLResponse(content=f.read()) @app.get("/logs", response_class=HTMLResponse) async def logs_page(): with open("/workspace/logs.html", "r", encoding="utf-8") as f: return HTMLResponse(content=f.read()) # ─────────────────────────────────────────────────────────────── # AUTH ENDPOINTS # ─────────────────────────────────────────────────────────────── @app.post("/api/auth/login") async def login(data: dict): key = data.get("password", "") if not verify_pass_key(key): raise HTTPException(status_code=401, detail="Invalid password") token = create_access_token({"sub": "user", "auth": True}) return {"access_token": token, "token_type": "bearer"} @app.post("/api/auth/verify") async def verify_auth(payload: dict = Depends(verify_token)): return {"valid": True, "user": payload.get("sub")} # ─────────────────────────────────────────────────────────────── # SETTINGS ENDPOINTS # ─────────────────────────────────────────────────────────────── _user_settings = {} @app.get("/api/settings") async def get_settings(payload: dict = Depends(verify_token)): return _user_settings @app.post("/api/settings") async def save_settings(data: dict, payload: dict = Depends(verify_token)): global _user_settings _user_settings.update(data) return {"status": "saved", "settings": _user_settings} # ─────────────────────────────────────────────────────────────── # GENERATION ENDPOINTS # ─────────────────────────────────────────────────────────────── @app.post("/api/generate") async def generate_video(data: dict, payload: dict = Depends(verify_token)): job_id = job_manager.create_job(data) job_manager.add_log("TASK", "ORCHESTRATOR", f"Starting video generation job #{job_id}") # Start pipeline in background asyncio.create_task(run_pipeline(job_id)) return {"job_id": job_id, "status": "queued"} @app.get("/api/status/{job_id}") async def get_status(job_id: str, payload: dict = Depends(verify_token)): if job_id not in job_manager.jobs: raise HTTPException(status_code=404, detail="Job not found") return job_manager.jobs[job_id] # ─────────────────────────────────────────────────────────────── # GALLERY ENDPOINTS # ─────────────────────────────────────────────────────────────── @app.get("/api/videos") async def list_videos(payload: dict = Depends(verify_token)): try: return r2_client.list_videos() except Exception as e: return [] @app.get("/api/videos/{video_id}/download") async def get_download_url(video_id: str, payload: dict = Depends(verify_token)): return {"url": f"{R2_PUBLIC_DOMAIN}/outputs/{video_id}"} # ─────────────────────────────────────────────────────────────── # LOGS ENDPOINTS # ─────────────────────────────────────────────────────────────── @app.get("/api/logs") async def get_logs(limit: int = 100, offset: int = 0, payload: dict = Depends(verify_token)): logs = job_manager.logs[offset:offset+limit] return {"logs": logs, "total": len(job_manager.logs)} @app.post("/api/logs/clear") async def clear_logs(payload: dict = Depends(verify_token)): job_manager.logs = [] return {"status": "cleared"} # ─────────────────────────────────────────────────────────────── # WEBSOCKET ENDPOINTS # ─────────────────────────────────────────────────────────────── @app.websocket("/ws/progress") async def ws_progress(websocket: WebSocket): await ws_manager.connect(websocket) try: while True: data = await websocket.receive_text() pass except WebSocketDisconnect: ws_manager.disconnect(websocket) @app.websocket("/ws/logs") async def ws_logs(websocket: WebSocket): await websocket.accept() job_manager.log_websockets.append(websocket) try: while True: data = await websocket.receive_text() if data == "ping": await websocket.send_text("pong") except WebSocketDisconnect: if websocket in job_manager.log_websockets: job_manager.log_websockets.remove(websocket) # ─────────────────────────────────────────────────────────────── # HEALTH & INFO # ─────────────────────────────────────────────────────────────── @app.get("/health") async def health(): return { "status": "healthy", "cuda": torch.cuda.is_available(), "gpus": torch.cuda.device_count(), "agents_loaded": _agents_loaded } @app.get("/api/info") async def info(): return { "name": "AI Video Swarm", "version": "1.0.0", "hardware": "8xL40S", "models": { "wan_2.6": "Wan-AI/Wan2.6-I2V-14B-720P", "ltx": "Lightricks/LTX-2.3", "flux": "black-forest-labs/FLUX.1-dev", }, "agents_loaded": _agents_loaded } # ─────────────────────────────────────────────────────────────── # MAIN # ─────────────────────────────────────────────────────────────── if __name__ == "__main__": import uvicorn uvicorn.run(app, host="0.0.0.0", port=7860, log_level="info")