TestingWebApp/flask_app/app/services/llm_client.py

"""OpenAI-совместимый клиент Chat Completions (порт `services/llmClient.js`)."""
from __future__ import annotations

import os
from dataclasses import dataclass
from typing import Optional

import urllib.request
import urllib.error
import json as _json


class LlmError(Exception):
    """Ошибка работы с LLM API."""

    def __init__(self, message: str, code: str = 'llm_error', status: int | None = None):
        super().__init__(message)
        self.code = code
        self.status = status


@dataclass
class LlmConfig:
    provider: str
    api_key: str
    base_url: str
    model: str


def get_llm_config() -> Optional[LlmConfig]:
    if k := os.environ.get('DEEPSEEK_API_KEY'):
        return LlmConfig(
            provider='deepseek',
            api_key=k,
            base_url=(os.environ.get('LLM_BASE_URL') or 'https://api.deepseek.com/v1').rstrip('/'),
            model=os.environ.get('LLM_MODEL') or 'deepseek-chat',
        )
    if k := os.environ.get('OPENAI_API_KEY'):
        return LlmConfig(
            provider='openai',
            api_key=k,
            base_url=(os.environ.get('LLM_BASE_URL') or 'https://api.openai.com/v1').rstrip('/'),
            model=os.environ.get('LLM_MODEL') or 'gpt-4o-mini',
        )
    return None


def chat_completion_text_content(
    cfg: LlmConfig,
    system: str,
    user: str,
    temperature: float = 0.25,
    timeout: int = 120,
) -> str:
    """Возвращает `assistant.message.content` (строку)."""
    body: dict = {
        'model': cfg.model,
        'messages': [
            {'role': 'system', 'content': system},
            {'role': 'user', 'content': user},
        ],
        'temperature': temperature,
    }
    if (os.environ.get('LLM_NO_JSON') or '').strip() != '1':
        body['response_format'] = {'type': 'json_object'}

    req = urllib.request.Request(
        f'{cfg.base_url}/chat/completions',
        data=_json.dumps(body).encode('utf-8'),
        headers={
            'Content-Type': 'application/json',
            'Authorization': f'Bearer {cfg.api_key}',
        },
        method='POST',
    )
    try:
        with urllib.request.urlopen(req, timeout=timeout) as resp:
            data = _json.loads(resp.read().decode('utf-8'))
    except urllib.error.HTTPError as e:
        text = ''
        try:
            text = e.read().decode('utf-8', errors='replace')
        except Exception:
            pass
        raise LlmError(
            f'LLM {e.code}: {(text or "").replace(chr(10), " ")[:280]}',
            code='llm_http',
            status=e.code,
        )
    except (urllib.error.URLError, TimeoutError) as e:
        msg = str(getattr(e, 'reason', '') or e)
        if 'timed out' in msg.lower():
            raise LlmError('Превышен таймаут ожидания ответа LLM (120 с).', code='llm_timeout')
        raise LlmError(f'Сбой сети при обращении к LLM: {msg}', code='llm_network')

    try:
        content = data['choices'][0]['message']['content']
    except (KeyError, IndexError, TypeError):
        content = None
    if not isinstance(content, str) or not content.strip():
        raise LlmError('Пустой content в ответе API.', code='llm_empty')
    return content


def ping_llm(timeout: int = 30) -> dict:
    """Smoke-проверка подключения к LLM. Не бросает исключений — всё в результате.

    Возвращает: {'ok': bool, 'provider', 'model', 'error'?, 'latencyMs'?, 'sample'?}
    """
    import time

    cfg = get_llm_config()
    if cfg is None:
        return {
            'ok': False,
            'configured': False,
            'error': 'Ключ не задан. Задайте DEEPSEEK_API_KEY или OPENAI_API_KEY в .env.',
        }
    started = time.monotonic()
    try:
        raw = chat_completion_text_content(
            cfg,
            'Отвечай ТОЛЬКО JSON: {"ok": true}.',
            'ping',
            temperature=0.0,
            timeout=timeout,
        )
        ms = int((time.monotonic() - started) * 1000)
        return {
            'ok': True,
            'configured': True,
            'provider': cfg.provider,
            'model': cfg.model,
            'latencyMs': ms,
            'sample': raw[:120],
        }
    except LlmError as e:
        ms = int((time.monotonic() - started) * 1000)
        return {
            'ok': False,
            'configured': True,
            'provider': cfg.provider,
            'model': cfg.model,
            'latencyMs': ms,
            'error': str(e),
            'code': e.code,
        }
    except Exception as e:
        return {
            'ok': False,
            'configured': True,
            'provider': cfg.provider,
            'model': cfg.model,
            'error': f'{type(e).__name__}: {e}',
            'code': 'unknown',
        }