app.py

import os
import gradio as gr
import requests

HF_TOKEN = os.getenv("HF_API_TOKEN")
MODEL_NAME = os.getenv("MODEL_NAME", "google/flan-t5-base")

API_URL = f"https://api-inference.huggingface.co/models/{MODEL_NAME}"
HEADERS = {"Authorization": f"Bearer {HF_TOKEN}"}

def query(payload):
    response = requests.post(API_URL, headers=HEADERS, json=payload)
    if response.status_code != 200:
        return f"❌ Ошибка API: {response.status_code} {response.text}"
    return response.json()

def chat(message, history):
    if history is None:
        history = []
    # Добавляем запрос пользователя с пустым ответом
    history.append((message, ""))

    response = query({"inputs": message})

    # Парсим ответ — пытаемся получить строку
    output = ""
    if isinstance(response, dict):
        if "error" in response:
            output = response["error"]
        elif "generated_text" in response:
            output = response["generated_text"]
        elif isinstance(response.get("choices"), list) and len(response["choices"]) > 0:
            output = response["choices"][0].get("message", {}).get("content", "")
        else:
            output = str(response)
    elif isinstance(response, list) and len(response) > 0 and "generated_text" in response[0]:
        output = response[0]["generated_text"]
    else:
        output = str(response)

    # Гарантируем, что output — строка
    if not isinstance(output, str):
        output = str(output)

    # Обновляем последний элемент в истории — теперь с ответом
    history[-1] = (message, output)
    return history, history

chat_ui = gr.ChatInterface(fn=chat, title="FlareGPT")

if __name__ == "__main__":
    chat_ui.launch()