app/tools/search_tools.py

"""
Search Tools for Product Discovery
- search_products is created via make_search_products_tool(session_id, registry).
- After search API, an LLM labels each result as Relevant / Partially Relevant / Irrelevant; we count and
  store the curated list in the registry, return [SEARCH_RESULTS_REF:ref_id] + quality counts + top10 titles.
"""
import base64
import json
import logging
import os
from pathlib import Path
from typing import Optional
import requests
from langchain_core.tools import tool
from openai import OpenAI
from app.config import settings
from app.search_registry import (
    ProductItem,
    SearchResult,
    SearchResultRegistry,
    global_registry,
)
logger = logging.getLogger(__name__)
_openai_client: Optional[OpenAI] = None
def _normalize_image_url(url: Optional[str]) -> Optional[str]:
    """Normalize image_url from API (e.g. ////cnres.appracle.com/... → https://cnres.appracle.com/...)."""
    if not url or not isinstance(url, str):
        return None
    url = url.strip()
    if not url:
        return None
    if url.startswith("https://") or url.startswith("http://"):
        return url
    # // or ////host/path → https://host/path (exactly one "//" after scheme)
    if url.startswith("/"):
        return "https://" + url.lstrip("/")
    return "https://" + url
def get_openai_client() -> OpenAI:
    global _openai_client
    if _openai_client is None:
        kwargs = {"api_key": settings.openai_api_key}
        if settings.openai_api_base_url:
            kwargs["base_url"] = settings.openai_api_base_url
        _openai_client = OpenAI(**kwargs)
    return _openai_client
# ── LLM quality assessment ─────────────────────────────────────────────────────
def _assess_search_quality(query: str, raw_products: list) -> tuple[list[str], str]:
    """
    Use LLM to label each search result and write a short quality_summary.
    Returns (labels, quality_summary). labels: one per product; quality_summary: 1–2 sentences.
    """
    n = len(raw_products)
    if n == 0:
        return [], ""
    lines = []
    for i, p in enumerate(raw_products, 1):
        title = (p.get("title") or "")[:60]
        lines.append(f"{i}. {title}")
    product_text = "\n".join(lines)
    prompt = f"""评估以下搜索结果与用户查询的匹配程度，完成两件事：
1. 为每条结果打一个等级：Relevant / Partially Relevant / Irrelevant。
2. 写一段 quality_summary（1–2 句话）：简要说明搜索结果主要包含哪些商品、是否基本满足搜索意图、整体匹配度如何。
用户查询：{query}
搜索结果（共 {n} 条）：
{product_text}
等级说明：
Relevant
The product generally satisfies the main shopping intent of the query. Minor missing, implicit, or unspecified attributes are acceptable as long as the product reasonably fits the intended use or scenario.
Partially Relevant
The product is related to the query and matches the general category or purpose, but shows weaker alignment with the specific intent or context.
Irrelevant
The product does not match the core intent or intended use implied by the query.
请严格按以下 JSON 输出，仅输出 JSON，无其他内容：
{{"labels": ["Relevant", "Partially Relevant", "Irrelevant", ...], "quality_summary": "你的1-2句总结"}}
labels 数组长度必须等于 {n}。"""
    try:
        client = get_openai_client()
        resp = client.chat.completions.create(
            model=settings.openai_model,
            messages=[{"role": "user", "content": prompt}],
            max_tokens=1200,
            temperature=0.1,
        )
        raw = resp.choices[0].message.content.strip()
        if raw.startswith("```"):
            raw = raw.split("```")[1]
            if raw.startswith("json"):
                raw = raw[4:]
        raw = raw.strip()
        data = json.loads(raw)
        labels = data.get("labels", [])
        valid = {"Relevant", "Partially Relevant", "Irrelevant"}
        labels = [l if l in valid else "Partially Relevant" for l in labels]
        while len(labels) < n:
            labels.append("Partially Relevant")
        quality_summary = (data.get("quality_summary") or "").strip() or ""
        return labels[:n], quality_summary
    except Exception as e:
        logger.warning(f"Quality assessment failed: {e}; using fallback.")
        return ["Partially Relevant"] * n, ""
# ── Shared search implementation ──────────────────────────────────────────────
def _call_search_api(query: str, size: int) -> Optional[tuple[list, int]]:
    """Call product search API. Returns (raw_results, total_hits) or None on failure."""
    if not query or not query.strip():
        return None
    try:
        url = f"{settings.search_api_base_url.rstrip('/')}/search/"
        headers = {
            "Content-Type": "application/json",
            "X-Tenant-ID": settings.search_api_tenant_id,
        }
        payload = {
            "query": query.strip(),
            "size": min(max(size, 1), 20),
            "from": 0,
            "language": "zh",
            "enable_rerank": True,
            "rerank_query_template": query.strip(),
            "rerank_doc_template": "{title}",
        }
        resp = requests.post(url, json=payload, headers=headers, timeout=60)
        if resp.status_code != 200:
            logger.warning(f"Search API {resp.status_code}: {resp.text[:200]}")
            return None
        data = resp.json()
        raw_results: list = data.get("results", [])
        total_hits: int = data.get("total", 0)
        return (raw_results, total_hits)
    except Exception as e:
        logger.warning(f"Search API error: {e}")
        return None
def _raw_to_product_items(
    raw_results: list, labels: Optional[list[str]] = None
) -> list[ProductItem]:
    """Build ProductItem list from API raw results. If labels given, filter by Relevant/Partially Relevant."""
    if labels is not None:
        valid = {"Relevant", "Partially Relevant"}
        return [
            ProductItem(
                spu_id=str(r.get("spu_id", "")),
                title=r.get("title") or "",
                price=r.get("price"),
                category_path=r.get("category_path") or r.get("category_name"),
                vendor=r.get("vendor"),
                image_url=_normalize_image_url(r.get("image_url")),
                relevance_score=r.get("relevance_score"),
                match_label=label,
                tags=r.get("tags") or [],
                specifications=r.get("specifications") or [],
            )
            for r, label in zip(raw_results, labels)
            if label in valid
        ]
    return [
        ProductItem(
            spu_id=str(r.get("spu_id", "")),
            title=r.get("title") or "",
            price=r.get("price"),
            category_path=r.get("category_path") or r.get("category_name"),
            vendor=r.get("vendor"),
            image_url=_normalize_image_url(r.get("image_url")),
            relevance_score=r.get("relevance_score"),
            match_label="Partially Relevant",
            tags=r.get("tags") or [],
            specifications=r.get("specifications") or [],
        )
        for r in raw_results
    ]
def search_products_impl(
    query: str,
    limit: int = 20,
    *,
    assess_quality: bool = True,
    session_id: Optional[str] = None,
    registry: Optional[SearchResultRegistry] = None,
) -> tuple[Optional[str], list[ProductItem], int]:
    """
    Single implementation: call API, optionally run LLM assessment, optionally register.
    Returns (ref_id_or_none, products, assessed_count). assessed_count = len(raw_results) when assess_quality else 0.
    """
    out = _call_search_api(query, limit)
    if not out:
        return (None, [], 0)
    raw_results, total_hits = out
    if not raw_results:
        return (None, [], 0)
    if assess_quality and session_id and registry:
        labels, quality_summary = _assess_search_quality(query, raw_results)
        products = _raw_to_product_items(raw_results, labels)
        perfect_count = sum(1 for l in labels if l == "Relevant")
        partial_count = sum(1 for l in labels if l == "Partially Relevant")
        ref_id = registry.next_ref_id(session_id)
        result = SearchResult(
            ref_id=ref_id,
            query=query,
            total_api_hits=total_hits,
            returned_count=len(raw_results),
            perfect_count=perfect_count,
            partial_count=partial_count,
            irrelevant_count=len(labels) - perfect_count - partial_count,
            quality_summary=quality_summary,
            products=products,
        )
        registry.register(session_id, result)
        logger.info(
            "[%s] Registered %s: query=%s perfect=%s partial=%s",
            session_id, ref_id, query, perfect_count, partial_count,
        )
        return (ref_id, products, len(raw_results))
    products = _raw_to_product_items(raw_results)
    return (None, products, 0)
# ── Tool factory ───────────────────────────────────────────────────────────────
def make_search_products_tool(
    session_id: str,
    registry: SearchResultRegistry,
):
    """
    Return a search_products tool bound to a specific session and registry.
    Uses LLM assessment and registers result; returns [SEARCH_RESULTS_REF:ref_id] string.
    """
    @tool
    def search_products(query: str) -> str:
        """搜索商品库并做质量评估：LLM 为每条结果打等级（Relevant / Partially Relevant / Irrelevant），返回引用与 top10 标题。
        Args:
            query: 自然语言商品描述
        Returns:
            【搜索完成】+ 结果引用 [SEARCH_RESULTS_REF:ref_id] + 质量情况 + results list（top10 标题）
        """
        try:
            limit = min(max(settings.search_products_limit, 1), 20)
            logger.info(f"[{session_id}] search_products: query={query!r} limit={limit}")
            ref_id, products, assessed_n = search_products_impl(
                query, limit,
                assess_quality=True,
                session_id=session_id,
                registry=registry,
            )
            if ref_id is None:
                if not products:
                    return (
                        f"【搜索完成】query='{query}'\n"
                        "未找到匹配商品，建议换用更宽泛或不同角度的关键词重新搜索。"
                    )
                return f"搜索失败：API 返回异常，请稍后重试。"
            perfect_count = sum(1 for p in products if p.match_label == "Relevant")
            partial_count = sum(1 for p in products if p.match_label == "Partially Relevant")
            top10_titles = [(p.title or "未知")[:80] for p in products[:10]]
            results_list = "\n".join(f"{i}. {t}" for i, t in enumerate(top10_titles, 1))
            return (
                f"【搜索完成】query='{query}'\n"
                f"结果引用：[SEARCH_RESULTS_REF:{ref_id}]\n"
                f"搜索结果质量情况：评估总条数{assessed_n}条，Relevant {perfect_count} 条，Partially Relevant {partial_count} 条。\n"
                f"results list:\n{results_list}"
            )
        except requests.exceptions.RequestException as e:
            logger.error(f"[{session_id}] Search network error: {e}", exc_info=True)
            return f"搜索失败（网络错误）：{e}"
        except Exception as e:
            logger.error(f"[{session_id}] Search error: {e}", exc_info=True)
            return f"搜索失败：{e}"
    return search_products
def search_products_api_only(query: str, limit: int = 12) -> list[ProductItem]:
    """API-only search (no LLM assessment). For 'Similar products' side panel."""
    _, products, _ = search_products_impl(query, limit, assess_quality=False)
    return products
# ── Standalone tools (no session binding needed) ───────────────────────────────
@tool
def web_search(query: str) -> str:
    """使用 Tavily 进行通用 Web 搜索，补充外部/实时知识。
    触发场景：
    - 需要**外部知识**：流行趋势、品牌、搭配文化、节日习俗等
    - 需要**实时/及时信息**：所有与天气相关的问题、当季流行元素、某地近期或者未来的事件、所有依赖当前时间相关的信息
    - 需要**宏观参考**：不同场合/国家的穿着建议、选购攻略
    Args:
        query: 要搜索的问题，自然语言描述
    Returns:
        总结后的回答 + 若干参考来源链接
    """
    try:
        api_key = os.getenv("TAVILY_API_KEY")
        if not api_key:
            return (
                "无法调用外部 Web 搜索：未检测到 TAVILY_API_KEY 环境变量。\n"
                "请在运行环境中配置 TAVILY_API_KEY 后再重试。"
            )
        logger.info(f"web_search: {query!r}")
        url = "https://api.tavily.com/search"
        headers = {
            "Authorization": f"Bearer {api_key}",
            "Content-Type": "application/json",
        }
        payload = {
            "query": query,
            "search_depth": "advanced",
            "include_answer": True,
        }
        response = requests.post(url, json=payload, headers=headers, timeout=60)
        if response.status_code != 200:
            return f"调用外部 Web 搜索失败：Tavily 返回状态码 {response.status_code}"
        data = response.json()
        answer = data.get("answer") or "（Tavily 未返回直接回答，仅返回了搜索结果。）"
        results = data.get("results") or []
        output_lines = [
            "【外部 Web 搜索结果（Tavily）】",
            "",
            "回答摘要：",
            answer.strip(),
        ]
        if results:
            output_lines.append("")
            output_lines.append("参考来源（部分）：")
            for idx, item in enumerate(results[:5], 1):
                title = item.get("title") or "无标题"
                link = item.get("url") or ""
                output_lines.append(f"{idx}. {title}")
                if link:
                    output_lines.append(f"   链接: {link}")
        return "\n".join(output_lines).strip()
    except requests.exceptions.RequestException as e:
        logger.error("web_search network error: %s", e, exc_info=True)
        return f"调用外部 Web 搜索失败（网络错误）：{e}"
    except Exception as e:
        logger.error("web_search error: %s", e, exc_info=True)
        return f"调用外部 Web 搜索失败：{e}"
@tool
def analyze_image_style(image_path: str) -> str:
    """分析用户上传的商品图片，提取视觉风格属性，用于后续商品搜索。
    适用场景：
    - 用户上传图片，想找相似商品
    - 需要理解图片中商品的风格、颜色、材质等属性
    Args:
        image_path: 图片文件路径
    Returns:
        商品视觉属性的详细文字描述，可直接作为 search_products 的 query
    """
    try:
        logger.info(f"analyze_image_style: {image_path!r}")
        img_path = Path(image_path)
        if not img_path.exists():
            return f"错误：图片文件不存在：{image_path}"
        with open(img_path, "rb") as f:
            image_data = base64.b64encode(f.read()).decode("utf-8")
        prompt = """请分析这张商品图片，提供详细的视觉属性描述，用于商品搜索。
请包含：
- 商品类型（如：连衣裙、运动鞋、双肩包、西装等）
- 主要颜色
- 风格定位（如：休闲、正式、运动、复古、现代简约等）
- 图案/纹理（如：纯色、条纹、格纹、碎花、几何图案等）
- 关键设计特征（如：领型、袖长、版型、材质外观等）
- 适用场合（如：办公、户外、度假、聚会、运动等）
输出格式：3-4句自然语言描述，可直接用作搜索关键词。"""
        client = get_openai_client()
        response = client.chat.completions.create(
            model=settings.openai_vision_model,
            messages=[
                {
                    "role": "user",
                    "content": [
                        {"type": "text", "text": prompt},
                        {
                            "type": "image_url",
                            "image_url": {
                                "url": f"data:image/jpeg;base64,{image_data}",
                                "detail": "high",
                            },
                        },
                    ],
                }
            ],
            max_tokens=800,
            temperature=0.3,
        )
        analysis = response.choices[0].message.content.strip()
        logger.info("Image analysis completed.")
        return analysis
    except Exception as e:
        logger.error(f"analyze_image_style error: {e}", exc_info=True)
        return f"图片分析失败：{e}"
# ── Tool list factory ──────────────────────────────────────────────────────────
def get_all_tools(
    session_id: str = "default",
    registry: Optional[SearchResultRegistry] = None,
) -> list:
    """
    Return all agent tools.
    search_products is session-bound (factory); other tools are stateless.
    """
    if registry is None:
        registry = global_registry
    return [
        make_search_products_tool(session_id, registry),
        analyze_image_style,
        web_search,
    ]