app.py

from fastapi import FastAPI, HTTPException
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel
from typing import Optional
import edge_tts
import httpx
import base64
import os
from groq import Groq

app = FastAPI(title="Anima - AI Avatar Chat")

# CORS para permitir frontend
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

# Groq client
client = Groq(api_key=os.environ.get("GROQ_API_KEY"))

# Configurações
VOICE = "pt-BR-FranciscaNeural"
SADTALKER_API = os.environ.get("SADTALKER_API", "https://madras1-sadtalker-api.hf.space")

class ChatRequest(BaseModel):
    message: str
    history: list = []
    avatar_image: Optional[str] = None  # Base64 da imagem do avatar
    use_video: bool = False   # Se True, gera vídeo com SadTalker

class ChatResponse(BaseModel):
    text: str
    audio_base64: Optional[str] = None
    video_base64: Optional[str] = None

@app.get("/health")
async def health():
    return {"status": "ok", "sadtalker_api": SADTALKER_API}

@app.post("/chat", response_model=ChatResponse)
async def chat(request: ChatRequest):
    try:
        # 1. Chama o LLM (Groq)
        messages = [
            {"role": "system", "content": "Você é Anima, uma assistente virtual amigável e prestativa. Responda de forma natural e concisa em português brasileiro."}
        ]
        
        for msg in request.history:
            messages.append(msg)
        
        messages.append({"role": "user", "content": request.message})
        
        completion = client.chat.completions.create(
            model="llama-3.1-8b-instant",
            messages=messages,
            temperature=0.7,
            max_tokens=500,
        )
        
        response_text = completion.choices[0].message.content
        
        # 2. Gera áudio com Edge-TTS
        communicate = edge_tts.Communicate(response_text, VOICE)
        audio_data = b""
        
        async for chunk in communicate.stream():
            if chunk["type"] == "audio":
                audio_data += chunk["data"]
        
        audio_base64 = base64.b64encode(audio_data).decode("utf-8")
        
        # 3. Se pediu vídeo e tem imagem, chama SadTalker
        video_base64 = None
        if request.use_video and request.avatar_image:
            try:
                video_base64 = await generate_video(request.avatar_image, audio_base64)
            except Exception as e:
                print(f"SadTalker error: {e}")
                # Continua sem vídeo se falhar
        
        return ChatResponse(
            text=response_text,
            audio_base64=audio_base64,
            video_base64=video_base64
        )
        
    except Exception as e:
        print(f"Chat error: {e}")
        raise HTTPException(status_code=500, detail=str(e))

async def generate_video(image_base64: str, audio_base64: str) -> str:
    """Chama SadTalker API para gerar vídeo"""
    async with httpx.AsyncClient(timeout=180.0) as http_client:
        # Gradio API endpoint - usa o endpoint correto
        response = await http_client.post(
            f"{SADTALKER_API}/run/predict",
            json={
                "data": [
                    {"path": None, "data": f"data:image/png;base64,{image_base64}"},
                    {"path": None, "data": f"data:audio/mp3;base64,{audio_base64}"}
                ]
            }
        )
        
        if response.status_code != 200:
            raise Exception(f"SadTalker API error: {response.status_code} - {response.text}")
        
        result = response.json()
        
        # O resultado pode ser um path ou base64
        video_data = result.get("data", [None])[0]
        
        if video_data:
            if isinstance(video_data, dict) and "data" in video_data:
                # É base64 no formato data URL
                return video_data["data"].split(",")[1] if "," in video_data["data"] else video_data["data"]
            elif isinstance(video_data, str) and video_data.startswith("data:"):
                return video_data.split(",")[1]
            elif isinstance(video_data, str):
                # É um path, precisa baixar
                video_response = await http_client.get(f"{SADTALKER_API}/file={video_data}")
                return base64.b64encode(video_response.content).decode("utf-8")
        
        raise Exception("No video in response")

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