main.py

from fastapi import FastAPI, Request
from fastapi.middleware.cors import CORSMiddleware  # Importa il middleware CORS
from pydantic import BaseModel
from huggingface_hub import InferenceClient
from datetime import datetime
from gradio_client import Client
import base64
import requests
import os
import socket
import time

#--------------------------------------------------- Definizione Server FAST API ------------------------------------------------------
app = FastAPI()
client = InferenceClient("mistralai/Mixtral-8x7B-Instruct-v0.1")

app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

class InputData(BaseModel):
    input: str
    temperature: float = 0.2
    max_new_tokens: int = 30000
    top_p: float = 0.95
    repetition_penalty: float = 1.0

class InputImage(BaseModel):
    input: str
    negativePrompt: str = '' 
    steps: int = 25
    cfg: int = 7
    seed: int = -1

#--------------------------------------------------- Generazione TESTO ------------------------------------------------------
@app.post("/Genera")
def read_root(request: Request, input_data: InputData):
    input_text = input_data.input
    temperature = input_data.temperature
    max_new_tokens = input_data.max_new_tokens
    top_p = input_data.top_p
    repetition_penalty = input_data.repetition_penalty
    history = [] 
    generated_response = generate(input_text, history, temperature, max_new_tokens, top_p, repetition_penalty)
    return {"response": generated_response}

def generate(prompt, history, temperature=0.2, max_new_tokens=30000, top_p=0.95, repetition_penalty=1.0):
    temperature = float(temperature)
    if temperature < 1e-2:
        temperature = 1e-2
    top_p = float(top_p)
    generate_kwargs = dict(
        temperature=temperature,
        max_new_tokens=max_new_tokens,
        top_p=top_p,
        repetition_penalty=repetition_penalty,
        do_sample=True,
        seed=42,
    )
    formatted_prompt = format_prompt(prompt, history)
    output = client.text_generation(formatted_prompt, **generate_kwargs, stream=False, details=False)
    return output
    
def format_prompt(message, history):
    prompt = "<s>"
    for user_prompt, bot_response in history:
        prompt += f"[INST] {user_prompt} [/INST]"
        prompt += f" {bot_response}</s> "
    now = datetime.now().strftime("%Y-%m-%d %H:%M:%S.%f")
    prompt += f"[{now}] [INST] {message} [/INST]"
    return prompt
    
#--------------------------------------------------- Generazione IMMAGINE ------------------------------------------------------
@app.post("/Immagine")
def generate_image(request: Request, input_data: InputImage):
    client = Client("https://openskyml-fast-sdxl-stable-diffusion-xl.hf.space/--replicas/545b5tw7n/")
    max_attempts = 20
    attempt = 0
    while attempt < max_attempts:
        try:
            result = client.predict(
                input_data.input,
                input_data.negativePrompt,
                input_data.steps, 
                input_data.cfg,
                1024,
                1024,
                input_data.seed,
                fn_index=0
            )
            image_url = result
            with open(image_url, 'rb') as img_file:
                img_binary = img_file.read()
                img_base64 = base64.b64encode(img_binary).decode('utf-8')
            return {"response": img_base64}
        except requests.exceptions.HTTPError as e:
            time.sleep(1)
            attempt += 1
            if attempt < max_attempts:
                continue
            else:
                return {"error": "Errore interno del server persistente"}
    return {"error": "Numero massimo di tentativi raggiunto"}

@app.get("/")
def read_general():
    return {"response": "Benvenuto. Per maggiori info: https://matteoscript-fastapi.hf.space/docs"}