core/interviewer.py

# core/interviewer.py
"""
🇬🇧 Interviewer logic module (no instructions)
Generates random MBTI-style questions using a fine-tuned model.

🇷🇺 Модуль интервьюера.
Использует fine-tuned модель для генерации вопросов без промптов и инструкций.
"""

import random, torch, re
from transformers import AutoModelForSeq2SeqLM, T5Tokenizer

# --------------------------------------------------------------
# 1️⃣ Настройки модели
# --------------------------------------------------------------
QG_MODEL = "f3nsmart/ft-flan-t5-base-qgen"

# ❗ Используем "slow" SentencePiece токенайзер
tokenizer = T5Tokenizer.from_pretrained(QG_MODEL, use_fast=False)
model = AutoModelForSeq2SeqLM.from_pretrained(QG_MODEL)

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device).eval()

print(f"✅ Loaded interviewer model (slow tokenizer): {QG_MODEL}")

# --------------------------------------------------------------
# 2️⃣ Базовые seed-промпты (без инструкций)
# --------------------------------------------------------------
PROMPTS = [
    "Personality and emotions.",
    "Human motivation and choices.",
    "Self-awareness and reflection.",
    "Personal growth and behavior.",
    "How people make decisions.",
]

# --------------------------------------------------------------
# 3️⃣ Очистка текста
# --------------------------------------------------------------
def _clean_question(text: str) -> str:
    """Берёт первую фразу с '?', обрезает лишнее"""
    text = text.strip()
    m = re.search(r"(.+?\?)", text)
    if m:
        text = m.group(1)
    text = text.replace("\n", " ").strip()
    if len(text.split()) < 3:
        text = text.capitalize()
        if not text.endswith("?"):
            text += "?"
    return text


# --------------------------------------------------------------
# 4️⃣ Генерация вопроса
# --------------------------------------------------------------
def generate_question(user_id: str = "default_user", **kwargs) -> str:
    """
    Генерирует один MBTI-вопрос без инструкций.
    """
    prompt = random.choice(PROMPTS)
    inputs = tokenizer(prompt, return_tensors="pt", truncation=True).to(device)
    with torch.no_grad():
        out = model.generate(
            **inputs,
            do_sample=True,
            top_p=0.9,
            temperature=0.9,
            repetition_penalty=1.1,
            max_new_tokens=60,
        )
    text = tokenizer.decode(out[0], skip_special_tokens=True)
    question = _clean_question(text)
    return question