"""可插拔检索 provider + 网页抓取。

刻意避开 AGPL 许可的 SearXNG：通过商用/宽松许可的检索 API（默认 Tavily）实现，
未配置时返回空结果（流水线降级为无来源报告）。抓取正文用标准库 HTML→文本。
"""

import httpx

from config import config
from htmltext import html_to_text

_client = httpx.Client(timeout=30.0, follow_redirects=True,
                       headers={"User-Agent": "GovAI-Research/1.0"})


def search(query: str, k: int) -> list[dict]:
    """返回 [{title, url, snippet, content}]；未配置 provider 时返回 []。"""
    provider = (config.SEARCH_PROVIDER or "").lower()
    if provider == "tavily" and config.SEARCH_API_KEY:
        return _tavily(query, k)
    return []


def _tavily(query: str, k: int) -> list[dict]:
    try:
        resp = _client.post(
            f"{config.SEARCH_BASE_URL.rstrip('/')}/search",
            json={
                "api_key": config.SEARCH_API_KEY,
                "query": query,
                "max_results": k,
                "include_raw_content": True,
            },
        )
        resp.raise_for_status()
        data = resp.json()
    except Exception:
        return []

    out = []
    for r in data.get("results", []) or []:
        out.append({
            "title": r.get("title") or r.get("url", ""),
            "url": r.get("url", ""),
            "snippet": r.get("content", "") or "",
            "content": (r.get("raw_content") or r.get("content") or "")[:6000],
        })
    return out


def fetch_text(url: str) -> str:
    """抓取网页并抽取正文文本（失败返回空串）。"""
    try:
        resp = _client.get(url)
        resp.raise_for_status()
        ctype = resp.headers.get("content-type", "")
        if "html" in ctype or ctype == "":
            return html_to_text(resp.text)
        if "text" in ctype or "json" in ctype:
            return resp.text[:6000]
    except Exception:
        return ""
    return ""