"""소개문·메타설명·FAQ 생성 — 겹들을 엮어 결과를 만드는 자리."""
import hashlib
import json
from dataclasses import dataclass, field
from typing import Optional

import httpx

from common.enums import PlaceCategory
from common.logger import LOG
from services.grounding.copy import FactInput, faq_polarity_ok, ground_check
from services.llm import provider
from services.llm.errors import LlmError
from services.llm.errors import LlmInvalidOutput as GeminiInvalidOutput
from services.llm.errors import LlmNotConfigured as GeminiNotConfigured
from services.prompts.copy import RESPONSE_SCHEMA, build_prompt


def is_configured() -> bool:
    """호출측(copy_service.py, place_service.py 등)은 이 겹만 안다 — 어느 공급자가 활성인지는 몰라도 된다."""
    return provider.active().is_configured()


@dataclass
class GeneratedFaq:
    question: str
    answer: str
    fact_keys: list[str] = field(default_factory=list)


@dataclass
class GeneratedCopy:
    """생성 결과."""

    intro: Optional[str] = None
    intro_fact_keys: list[str] = field(default_factory=list)
    meta_description: Optional[str] = None
    faqs: list[GeneratedFaq] = field(default_factory=list)
    rejected: list[tuple[str, str]] = field(default_factory=list)
    source: str = ""  # "openai:gpt-5.6-luna" 형식 — copy_steps.py 가 fact 출처 표기에 쓴다


def _unit_facts(unit_summaries: Optional[list[dict]]) -> list[FactInput]:
    """객실·프로그램 요약을 근거 fact 로 펼친다."""
    out: list[FactInput] = []
    for unit in unit_summaries or []:
        name = str(unit.get("name") or "").strip()
        labels = unit.get("labels") or {}
        if name:
            out.append(FactInput(key=f"unit:{name}", label="객실·프로그램명", value=name))
        for key, value in (unit.get("facts") or {}).items():
            if value is None or str(value).strip() == "":
                continue
            spec = labels.get(key) or {}
            out.append(FactInput(
                key=f"{name}:{key}" if name else key,
                label=spec.get("label") or key,
                value=str(value),
                unit=spec.get("unit"),
            ))
    return out


def _valid_keys(claimed: list, allowed: set[str]) -> list[str]:
    """모델이 적어준 근거 key 중 실제로 존재하는 것만 남긴다(없는 key 를 지어내기도 한다)."""
    return [k for k in (claimed or []) if isinstance(k, str) and k in allowed]


async def generate_copy(
    place_name: str,
    category: PlaceCategory,
    facts: list[FactInput],
    *,
    unit_summaries: Optional[list[dict]] = None,
    records: Optional[list[str]] = None,
    suggested_questions: Optional[list[str]] = None,
    max_faqs: int = 8,
    model: Optional[str] = None,
    max_retries: int = 2,
    client: Optional[httpx.AsyncClient] = None,
) -> GeneratedCopy:
    """확보된 fact 만으로 소개문·메타설명·FAQ 를 만든다."""
    llm = provider.active()
    if not llm.is_configured():
        raise GeminiNotConfigured(f"{llm.__name__.rsplit('.', 1)[-1].upper()}_API_KEY 가 설정되지 않았다")
    model = model or llm.DEFAULT_MODEL
    # 사업장 fact 가 없어도 객실·메뉴 근거가 있으면 쓴다.
    unit_grounding = _unit_facts(unit_summaries)
    if not facts and not unit_grounding:
        LOG.i(f"[llm-text] '{place_name}' 근거 fact 0건 — 생성하지 않는다(호출 없음)")
        return GeneratedCopy(rejected=[("(전체)", "근거 fact 가 없다 — 생성하지 않았다")])

    # 검증에 쓸 근거 = 넘겨받은 fact + 객실 요약 + 상호명(상호에 숫자가 있어도 근거로 본다)
    grounding = list(facts) + unit_grounding
    grounding.append(FactInput(key="place_name", label="상호명", value=place_name))
    allowed_keys = {f.key for f in facts} | {f.key for f in grounding}

    prompt = build_prompt(place_name, category, facts, max_faqs, unit_grounding, records, suggested_questions)

    owns_client = client is None
    client = client or httpx.AsyncClient(timeout=httpx.Timeout(120.0, connect=10.0))
    try:
        llm_result = await llm.generate(
            client, model, prompt=prompt, response_schema=RESPONSE_SCHEMA, temperature=0.2, max_retries=max_retries,
        )
    finally:
        if owns_client:
            await client.aclose()

    parsed = llm_result.json
    usage = llm_result.usage

    result = GeneratedCopy(source=f"{llm.__name__.rsplit('.', 1)[-1]}:{model}")

    # ── 소개문 ──
    intro = (parsed.get("intro") or "").strip()
    if intro:
        ok, reasons = ground_check(intro, grounding)
        if ok:
            result.intro = intro
            result.intro_fact_keys = _valid_keys(parsed.get("intro_fact_keys"), allowed_keys)
        else:
            result.rejected.append((intro, " / ".join(reasons)))

    # ── 메타 설명 ──
    meta_desc = (parsed.get("meta_description") or "").strip()
    if meta_desc:
        ok, reasons = ground_check(meta_desc, grounding)
        if ok:
            result.meta_description = meta_desc
        else:
            result.rejected.append((meta_desc, " / ".join(reasons)))

    # ── FAQ ── 항목마다 따로 검사한다.
    for item in (parsed.get("faqs") or [])[:max_faqs]:
        question = (item.get("question") or "").strip()
        answer = (item.get("answer") or "").strip()
        if not question or not answer:
            continue
        keys = _valid_keys(item.get("fact_keys"), allowed_keys)
        if not keys:
            # 근거를 못 대는 FAQ 는 버린다 — 사실인지 확인할 방법이 없다.
            result.rejected.append((question, "근거 fact_keys 가 없다"))
            continue
        ok, reasons = ground_check(f"{question} {answer}", grounding)
        # 질문은 주장이 아니라 값-반대 판정에서 빠진다.
        polar_ok, polar_reasons = faq_polarity_ok(question, answer, grounding)
        if not ok or not polar_ok:
            result.rejected.append((question, " / ".join(reasons + polar_reasons)))
            continue
        result.faqs.append(GeneratedFaq(question=question, answer=answer, fact_keys=keys))

    LOG.i(
        f"[llm-text] '{place_name}' 생성 — 소개문 {'O' if result.intro else 'X'} · "
        f"메타 {'O' if result.meta_description else 'X'} · FAQ {len(result.faqs)}건 · "
        f"반려 {len(result.rejected)}건 · model={model} · "
        f"tokens in={usage.input_tokens} out={usage.output_tokens} · 약 ${llm.price(model, usage)}"
    )
    return result


# ── 요약(summarize_text) ──────────────────────────────────────────────────
_SUMMARY_CACHE: dict[str, str] = {}
_SUMMARY_CACHE_MAX = 500
_SUMMARY_PROMPT = (
    "다음 숙소 소개에서 핵심 특징 1~2개만 골라 한국어 한 문장, 공백 포함 60~80자로 요약해줘. "
    "원문에 없는 사실이나 과장 표현을 추가하지 말고, 선택한 사실의 조건과 부정 표현을 유지해. "
    "반복되는 상호명, 인사말, 홍보 수식어는 생략해. "
    "요약문만 출력하고 다른 말은 붙이지 마.\n\n"
)


async def summarize_text(
    text: str,
    *,
    model: Optional[str] = None,
    max_retries: int = 2,
    client: Optional[httpx.AsyncClient] = None,
) -> Optional[str]:
    """캔버스 미리보기용 축약문."""
    stripped = text.strip()
    if not stripped:
        return None
    llm = provider.active()
    if not llm.is_configured():
        return None
    model = model or llm.DEFAULT_MODEL

    # 길이 기준을 바꾼 뒤 이전 길이의 요약을 재사용하지 않도록 프롬프트도 키에 넣는다.
    cache_key = hashlib.sha256((_SUMMARY_PROMPT + stripped).encode("utf-8")).hexdigest()
    cached = _SUMMARY_CACHE.get(cache_key)
    if cached is not None:
        return cached

    owns_client = client is None
    client = client or httpx.AsyncClient(timeout=httpx.Timeout(60.0, connect=10.0))
    try:
        result = await llm.generate(client, model, prompt=_SUMMARY_PROMPT + stripped, temperature=0.2, max_retries=max_retries)
        summary = result.text.strip()
    except LlmError as ex:
        LOG.w(f"[llm-text] 요약 실패: {ex}")
        return None
    finally:
        if owns_client:
            await client.aclose()

    if not summary:
        return None
    usage = result.usage
    LOG.i(
        f"[llm-text] 요약 {len(stripped)}자 → {len(summary)}자 · model={model} · "
        f"tokens in={usage.input_tokens} out={usage.output_tokens} · 약 ${llm.price(model, usage)}"
    )
    if len(_SUMMARY_CACHE) >= _SUMMARY_CACHE_MAX:
        _SUMMARY_CACHE.clear()  # 간단한 캐시 상한 — 관리 도구 트래픽 규모에는 LRU 가 과하다.
    _SUMMARY_CACHE[cache_key] = summary
    return summary


@dataclass
class GeneratedSong:
    """가사 생성 결과."""

    title: str
    lyrics: str
    style: str


async def generate_song(
    place_name: str,
    category: PlaceCategory,
    *,
    region: str,
    grounding: list[str],
    intro: str = "",
    model: Optional[str] = None,
    max_retries: int = 2,
    client: Optional[httpx.AsyncClient] = None,
) -> GeneratedSong:
    """이 업소의 노래 가사를 쓴다."""
    llm = provider.active()
    if not llm.is_configured():
        raise GeminiNotConfigured("API 키가 설정되지 않았다")
    if not grounding and not (intro or "").strip():
        raise GeminiInvalidOutput("가사를 쓸 재료가 없다 — 확인된 fact 도 소개문도 없다")
    model = model or llm.DEFAULT_MODEL

    from common.category_schema import get_schema
    from services.prompts.song import RESPONSE_SCHEMA as SONG_SCHEMA, build_prompt as build_song_prompt

    prompt = build_song_prompt(place_name, get_schema(category).label, region, grounding, intro)

    owns_client = client is None
    client = client or httpx.AsyncClient(timeout=httpx.Timeout(120.0, connect=10.0))
    try:
        llm_result = await llm.generate(
            client, model, prompt=prompt, response_schema=SONG_SCHEMA, temperature=0.9, max_retries=max_retries,
        )
    finally:
        if owns_client:
            await client.aclose()

    parsed = llm_result.json or {}
    title = (parsed.get("title") or "").strip()
    lyrics = (parsed.get("lyrics") or "").strip()
    style = (parsed.get("style") or "").strip()
    if not lyrics:
        raise GeminiInvalidOutput("가사가 비어 있다")

    usage = llm_result.usage
    LOG.i(
        f"[llm-text] '{place_name}' 가사 — '{title}' ({style}) · {len(lyrics)}자 · "
        f"tokens in={usage.input_tokens} out={usage.output_tokens} · 약 ${llm.price(model, usage)}"
    )
    # 제목이 비면 상호를 쓴다 — 빈 제목은 플레이어에서 빈 줄로 보인다.
    return GeneratedSong(title=title or place_name, lyrics=lyrics, style=style or "acoustic ballad")


async def generate_social_post(place_name, facts, link_url, provider=2, *, client=None,
                               region='', category=''):
    """실제 게시 문자열을 검증한다."""
    from services.prompts import social
    from services.external.social import adapter, weighted_length, URL
    from services.llm import provider as llm_provider
    import unicodedata

    if not facts:
        raise GeminiInvalidOutput('NO_GROUNDED_FACTS')
    llm = llm_provider.active()
    if not llm.is_configured():
        raise GeminiNotConfigured(f"{llm.__name__.rsplit('.', 1)[-1].upper()}_API_KEY 가 설정되지 않았다")
    limit = adapter(provider).weighted_limit()
    allowed = {f.key for f in facts}
    feedback = ''
    owns = client is None
    client = client or httpx.AsyncClient(timeout=45)
    try:
        for _ in range(3):
            prompt = social.build_prompt(
                place_name, facts, limit - weighted_length('\n\n' + link_url, provider), feedback,
                region=region, category=category)
            result = await llm.generate(
                client, llm.DEFAULT_MODEL, prompt=prompt,
                response_schema=social.RESPONSE_SCHEMA, temperature=0.2, max_retries=0,
            )
            try:
                parsed = result.json if result.json is not None else json.loads(result.text)
                body = unicodedata.normalize('NFC', parsed['body'].strip())
                keys = parsed['fact_keys']
                text = body + '\n\n' + link_url
                # ★ 지역·업종은 fact 가 아니라 검증된 place 값이다. 근거에 얹지 않으면
                #   본문에 쓴 순간 "근거 없는 주장" 으로 반려되고 3회 재시도를 태우고 실패한다 —
                #   상호명을 이렇게 다루는 그 방식 그대로다.
                grounds = facts + [FactInput(key='name', label='상호명', value=place_name)]
                if region:
                    grounds += [FactInput(key='region', label='지역', value=region)]
                if category:
                    grounds += [FactInput(key='category', label='업종', value=category)]
                ok, _ = ground_check(body, grounds)
                if (body and isinstance(keys, list) and keys and all(k in allowed for k in keys)
                        and ok and not URL.search(body) and weighted_length(text, provider) <= limit):
                    return text
            except (ValueError, KeyError, TypeError):
                pass
            feedback = '이전 응답은 길이 또는 근거 검증에 실패했다. 더 짧게, 제공된 사실만으로 다시 써라.'
        raise GeminiInvalidOutput('SOCIAL_INVALID_OUTPUT')
    finally:
        if owns:
            await client.aclose()