brave search

date: 2026-01-18 excerpt: Brave Search APIの使い方

brave search

概要

Brave Search APIでWeb検索結果を取得する
APIトークンは環境変数で渡す
requests_cacheを使うと繰り返し実行が速くなる

事前準備

環境変数にAPIトークンを設定する

export BRAVE_SEARCH_API="YOUR_TOKEN"

最小の実装

import os
import requests


def search_brave(query: str, count: int = 10, country: str = "US"):
    endpoint = "https://api.search.brave.com/res/v1/web/search"

    headers = {
        "Accept": "application/json",
        "Accept-Encoding": "gzip",
        "X-Subscription-Token": os.environ.get("BRAVE_SEARCH_API"),
    }
    params = {
        "q": query,
        "count": count,
        "country": country,
        "search_lang": "jp",
    }

    response = requests.get(endpoint, headers=headers, params=params, timeout=10)
    response.raise_for_status()
    return response.json()

キャッシュ付きの実装

import os
import requests_cache


def search_brave_cached(query: str, count: int = 10, country: str = "US"):
    endpoint = "https://api.search.brave.com/res/v1/web/search"

    headers = {
        "Accept": "application/json",
        "Accept-Encoding": "gzip",
        "X-Subscription-Token": os.environ.get("BRAVE_SEARCH_API"),
    }
    params = {
        "q": query,
        "count": count,
        "country": country,
        "search_lang": "jp",
    }

    session = requests_cache.CachedSession(
        cache_name="data/http_cache_search_brave",
        backend="sqlite",
        expire_after=None,
        allowable_methods=("GET",),
    )
    response = session.get(endpoint, headers=headers, params=params, timeout=10)
    response.raise_for_status()
    return response.json()

取得結果の取り出し方

Web検索の結果は web -> results に入る

result = search_brave_cached("NVDAの株価", count=5, country="JP")

for item in result.get("web", {}).get("results", []):
    title = item.get("title")
    url = item.get("url")
    age = item.get("age")
    print(title, url, age)

パラメータのメモ

count は最大20
country はJPやUSなどの国コード
search_lang を en にすると英語の結果を優先できる