import os
import time
import requests

# --- Configuración ---
HUGGINGFACE_API_KEY = os.getenv("HUGGINGFACE_API_KEY")
MODEL_ID = os.getenv("HUGGINGFACE_MODEL", "deepseek/deepseek-v3-0324")
HF_API_URL = "https://router.huggingface.co/novita/v3/openai/chat/completions"

if not HUGGINGFACE_API_KEY:
    raise ValueError("Error: La variable de entorno HUGGINGFACE_API_KEY no está definida.")

# --- Throttling ---
LAST_CALL_TIME = 0.0
MIN_INTERVAL_SECONDS = 5.0

# --- Fragmentación de la respuesta ---
MAX_CHUNK_LENGTH = 200
DELAY_BETWEEN_CHUNKS = 1.0


def query_huggingface(prompt):
    """Envía un prompt estilo chat a HuggingFace Router y devuelve la respuesta."""
    headers = {
        "Authorization": f"Bearer {HUGGINGFACE_API_KEY}",
        "Content-Type": "application/json"
    }

    payload = {
        "model": MODEL_ID,
        "messages": [
            {
                "role": "system",
                "content": "You are a helpful assistant that always responds in neutral Latin American Spanish. Ignore usernames or nicknames like 'teraflops'. Focus only on the user's input."
            },
            {
                "role": "user",
                "content": prompt
            }
        ],
        "temperature": 0.7,
        "max_tokens": 300
    }

    try:
        response = requests.post(HF_API_URL, headers=headers, json=payload, timeout=120)
        response.raise_for_status()
        return response.json()["choices"][0]["message"]["content"].strip()
    except requests.exceptions.Timeout:
        return "⏱️ El servidor de HuggingFace tardó demasiado en responder."
    except requests.exceptions.RequestException as e:
        return f"❌ Error al conectar con HuggingFace Router: {e}"


def chunk_text(text, max_length):
    """Corta 'text' en fragmentos de longitud máxima 'max_length'."""
    return [text[i:i + max_length] for i in range(0, len(text), max_length)]


def run(*args):
    """Consulta el modelo y devuelve la respuesta troceada, aplicando throttling."""
    global LAST_CALL_TIME

    if not args:
        return help()

    now = time.time()
    elapsed = now - LAST_CALL_TIME
    if elapsed < MIN_INTERVAL_SECONDS:
        wait_time = round(MIN_INTERVAL_SECONDS - elapsed, 1)
        return f"¡Espera {wait_time} seg antes de hacer otra consulta!"

    LAST_CALL_TIME = now
    prompt = " ".join(args)

    try:
        full_text = query_huggingface(prompt)
    except Exception as e:
        return f"Error al consultar HuggingFace: {e}"

    result = []
    for fragment in chunk_text(full_text, MAX_CHUNK_LENGTH):
        result.append(fragment)
        time.sleep(DELAY_BETWEEN_CHUNKS)

    return "\n".join(result)


def help():
    return f""".ollama <pregunta> - Consulta el modelo '{MODEL_ID}' en HuggingFace usando el router.

Ejemplos:
!ollama ¿Qué es la entropía en física?
!ollama Resume el argumento de Don Quijote en 3 líneas.
"""