Spaces:

guardiancc
/

dubai

Sleeping

File size: 9,706 Bytes

import traceback
import uuid
from models.whisper import model
import modules.register as register 
from processor import generate_audio
import json
from fastapi import FastAPI, HTTPException
from pydantic import BaseModel, HttpUrl
from fastapi.middleware.cors import CORSMiddleware
from fastapi.openapi.docs import get_swagger_ui_html
import os
import requests
from modules.audio import convert, get_audio_duration
from modules.r2 import upload_to_s3, upload_image_to_s3
import threading
import queue
from diffusers import DiffusionPipeline
import torch
from datetime import datetime
import random
import numpy as np

SAVE_DIR = "saved_images"  
if not os.path.exists(SAVE_DIR):
    os.makedirs(SAVE_DIR, exist_ok=True)

device = "cuda" if torch.cuda.is_available() else "cpu"
repo_id = "black-forest-labs/FLUX.1-dev"
adapter_id = "guardiancc/lora"

pipeline = DiffusionPipeline.from_pretrained(repo_id, torch_dtype=torch.bfloat16)
pipeline.load_lora_weights(adapter_id)
pipeline = pipeline.to(device)

MAX_SEED = np.iinfo(np.int32).max
MAX_IMAGE_SIZE = 1024

vpv_webhook = os.environ.get("VPV_WEBHOOK")

app = FastAPI(title="Minha API", description="API de exemplo com FastAPI e Swagger", version="1.0.0")
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)


def save_generated_image(image):
    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
    unique_id = str(uuid.uuid4())[:8]
    filename = f"{timestamp}_{unique_id}.png"
    filepath = os.path.join(SAVE_DIR, filename)
    image.save(filepath)
    
    return filepath

def inference_image(prompt):
    seed = random.randint(0, MAX_SEED)
    generator = torch.Generator(device=device).manual_seed(seed)
    image = pipeline(
        prompt=prompt,
        guidance_scale=3.5,
        num_inference_steps=20,
        width=512,
        height=512,
        generator=generator,
        joint_attention_kwargs={"scale": 0.8},
    ).images[0]
    
    filepath = save_generated_image(image, prompt)
    url = upload_image_to_s3(filepath, os.path.basename(filepath), "png")
    os.unlink(filepath)
    return url
    
def download_file(url: str) -> str:
    """
    Baixa um arquivo da URL fornecida e o salva no diretório 'downloads/'.
    O nome do arquivo é extraído da URL automaticamente.
    """
    try:
        os.makedirs("downloads", exist_ok=True)

        file_name = os.path.basename(url.split("?")[0]) 
        save_path = os.path.join("downloads", file_name)

        response = requests.get(url)
        response.raise_for_status() 

        with open(save_path, 'wb') as f:
            f.write(response.content)

        return save_path
    except requests.exceptions.RequestException as e:
        raise Exception(f"Erro ao baixar o arquivo: {e}")

@app.get("/test", include_in_schema=False)
def test():
    return {"ok": True}

@app.get("/", include_in_schema=False)
async def custom_swagger_ui_html():
    return get_swagger_ui_html(openapi_url="/openapi.json", title="Alert Pix Ai v2")

@app.get("/openapi.json", include_in_schema=False)
async def openapi():
    with open("swagger.json") as f:
        return json.load(f)


class ProcessRequest(BaseModel):
    key: str
    text: str
    id: str
    receiver: str
    webhook: str
    censor: bool = False 
    offset: float = -0.3 
    format: str = "wav"
    speed: float = 0.8
    crossfade: float = 0.1
    
class ProcessImage(BaseModel):
    prompt: str
    id: str
    receiver: str
    webhook: str

q = queue.Queue()
image_queue = queue.Queue()

def process_queue(q):
    while True:
        try:
            key, censor, offset, text, format, speed, crossfade, id, receiver, webhook = q.get(timeout=5)  
            audio = generate_audio(key, text, censor, offset, speed=speed, crossfade=crossfade) 
            convertedAudioPath = convert(audio, format)
            duration = get_audio_duration(convertedAudioPath)
            audioUrl = upload_to_s3(convertedAudioPath, f"{id}", format)
            os.remove(audio)
            os.remove(convertedAudioPath)
                
            payload = {
                    "id": id,
                    "duration": duration,
                    "receiver": receiver,
                    "url": audioUrl
                }
                
            requests.post(webhook, json=payload)
        except Exception as e:
            print(e)
        finally:
            q.task_done()
            
def process_image(q):
    while True:
        try:
            prompt, id, receiver, webhook = q.get(timeout=5)  
            image = inference_image(prompt)
                
            payload = {
                    "id": id,
                    "receiver": receiver,
                    "url": image,
                    "type": "image"
                }
                
            requests.post(webhook, json=payload)
        except Exception as e:
            print(e)
        finally:
            q.task_done()

        
worker_thread = threading.Thread(target=process_queue, args=(q,))
worker_thread.start()

imagge_worker = threading.Thread(target=process_queue, args=(q,))
imagge_worker.start()
        
@app.post("/process")
def process_audio(payload: ProcessRequest):
    key = payload.key
    censor = payload.censor
    offset = payload.offset
    text = payload.text
    format = payload.format
    speed = payload.speed
    crossfade = payload.crossfade
    id = payload.id
    receiver = payload.receiver
    webhook = payload.webhook
    
    if len(text) >= 1000:
        raise HTTPException(status_code=500, detail=str(e))

    try:
        q.put((key, censor, offset, text, format, speed, crossfade, id, receiver, webhook))
        return {"success": True, "err": ""}
        
    except ValueError as e:
        raise HTTPException(status_code=400, detail=str(e))
    
    except Exception as e:
        error_trace = traceback.format_exc() 
        dc_callback = "https://discord.com/api/webhooks/1285586984898662511/QNVvY2rtoKICamlXsC1BreBaYjS9341jz9ANCDBzayXt4C7v-vTFzKfUtKQkwW7BwpfP"
        
        data = {
            "content": "",
            "tts": False,
            "embeds": [
                {
                    "type": "rich",
                    "title": f"Erro aconteceu na IA - MIMIC - processo",
                    "description": f"Erro: {str(e)}\n\nDetalhes do erro:\n```{error_trace}```"
                }
            ]
        }
        
        headers = {
            "Content-Type": "application/json",
            "Accept": "application/json",
        }
        requests.post(dc_callback, headers=headers, data=json.dumps(data))
        raise HTTPException(status_code=500, detail=str(e))

@app.post("/image")
def process_image(payload: ProcessImage):
    prompt = payload.prompt
    id = payload.id
    receiver = payload.receiver
    webhook = payload.webhook
    
    if len(prompt) <= 5:
        raise HTTPException(status_code=500, detail=str(e))

    try:
        image_queue.put(( prompt, id, receiver, webhook))
        return {"success": True, "err": ""}
        
    except ValueError as e:
        raise HTTPException(status_code=400, detail=str(e))
    
    except Exception as e:
        error_trace = traceback.format_exc() 
        dc_callback = "https://discord.com/api/webhooks/1285586984898662511/QNVvY2rtoKICamlXsC1BreBaYjS9341jz9ANCDBzayXt4C7v-vTFzKfUtKQkwW7BwpfP"
        
        data = {
            "content": "",
            "tts": False,
            "embeds": [
                {
                    "type": "rich",
                    "title": f"Erro aconteceu na IA - MIMIC - 2 ia",
                    "description": f"Erro: {str(e)}\n\nDetalhes do erro:\n```{error_trace}```"
                }
            ]
        }
        
        headers = {
            "Content-Type": "application/json",
            "Accept": "application/json",
        }
        requests.post(dc_callback, headers=headers, data=json.dumps(data))
        raise HTTPException(status_code=500, detail=str(e))


class TrainRequest(BaseModel):
    audio: HttpUrl 
    key: str
    endpoint: str
    id: str
    
@app.post("/train")
def create_item(payload: TrainRequest):
    audio = payload.audio
    key = payload.key
    endpoint = payload.endpoint

    try:
        src = download_file(str(audio))
        data = register.process_audio(src, key)

        for i in range(3):
            try:
                payload = {"success": True, "id": payload.id}
                requests.post(endpoint, json=payload)
                break
            except Exception as e:
                pass
            
        return data
        
    except ValueError as e:
        raise HTTPException(status_code=400, detail=str(e))
    except Exception as e:
        error_trace = traceback.format_exc() 
        dc_callback = "https://discord.com/api/webhooks/1285586984898662511/QNVvY2rtoKICamlXsC1BreBaYjS9341jz9ANCDBzayXt4C7v-vTFzKfUtKQkwW7BwpfP"
        
        data = {
            "content": "",
            "tts": False,
            "embeds": [
                {
                    "type": "rich",
                    "title": f"Erro aconteceu na IA -MIMIC - treinar",
                    "description": f"Erro: {str(e)}\n\nDetalhes do erro:\n```{error_trace}```"
                }
            ]
        }
        
        headers = {
            "Content-Type": "application/json",
            "Accept": "application/json",
        }
        requests.post(dc_callback, headers=headers, data=json.dumps(data))

        raise HTTPException(status_code=500, detail=str(e))

if __name__ == "__main__":
    import uvicorn
    uvicorn.run("app:app", host="0.0.0.0", port=7860)