自分用：言語指定込み（en, ja, 無指定）でハッシュタグ付きの投稿を取得。
ランダムに近くするように、1日をN（24の約数）分割して、その中で取得時間幅（1分）を生成。

動くことは確認したがコードの適切性は入念に検証していない。あと改行が適当すぎる。

事前に開発者プラットフォーム
https://developer.x.com/
でログインして、APP作成＆プロジェクト作成を済ませる

import random
import requests

# 日付系。calenderなどはルーティーン処理したかったので追加した
import datetime
from time import sleep
from dateutil.relativedelta import relativedelta
import calendar

# 保存用
# import csv
import json

import tweepy

# Twitter Developer Platformの使いたいAPPのページから取得する
API_Key     = ''
API_Sec     = ''
Token       = ''
Token_Sec   = ''
BEARER = ''

# 指定した時刻から1分前までの最大10投稿を取得（※APIのmax_resultsは10以上制約があるっぽい）
def main_search(time, filter_lang=None):
    client = tweepy.Client(bearer_token=BEARER, consumer_key=API_Key, consumer_secret=API_Sec, access_token=Token, access_token_secret=Token_Sec)

    count=10
    results=[]

    search_query = '"#" -is:retweet has:hashtags'
    if filter_lang is not None:
        search_query = search_query + ' lang:' + filter_lang

    response = client.search_recent_tweets(
        query=search_query,
        tweet_fields=["created_at", "lang", "entities","attachments","conversation_id","geo","edit_history_tweet_ids","public_metrics"], 
        expansions=["attachments.media_keys", "author_id", "in_reply_to_user_id"],
        max_results = count,
        end_time = time.isoformat(timespec='seconds') + 'Z',
        start_time = (time - datetime.timedelta(minutes=1)).isoformat(timespec='seconds') + 'Z',
    )

    tweets = response.data

    includes = response.includes
    users = includes["users"]


    sleep(1)


    if tweets != None:
        for tweet in tweets:
            obj = {}
            obj["tweet_id"] = tweet.id
            created_at = tweet.created_at
            created_at_str = created_at.isoformat()
            obj['created_at'] = created_at_str
            obj["text"] = tweet.text
            obj['author_id'] = tweet.author_id
            obj['conversation_id'] = tweet.conversation_id
            if tweet.in_reply_to_user_id is not None:
                obj['in_reply_to_user_id'] = tweet.in_reply_to_user_id
            if tweet.attachments is not None:
                obj['attachments'] = tweet.attachments
            if tweet.entities is not None:
                obj['entities'] =tweet.entities
            if tweet.geo is not None:
                obj['geo'] = tweet.geo
            
            results.append(obj)
    else:
        results.append({})

    print(results[0])

    return results

# ランダム時刻生成
def get_random_time_within_hour(day, hour_range=6, hour_add=0):
    # hour_range: 24を分割。
    """ 指定された日のランダムな時間、分、秒を返す """
    hour = random.randint(0, hour_range-1) + hour_add
    minute = random.randint(0, 59)
    second = random.randint(0, 59)
    return datetime.datetime(day.year, day.month, day.day, hour, minute, second)

def make_loop_in_a_day(target_day, separate_num = 6, filter_lang="ja"):
    separate_num = int(separate_num)
    hour_range = 24 // separate_num

    results_total = []

    for loop_count in range(separate_num):
        hour_add = (loop_count) * hour_range
        results_now = []
        time = get_random_time_within_hour(target_day, hour_range=hour_range, hour_add=hour_add)
        try:
            results_now = main_search(time=time, filter_lang=filter_lang)
        except Exception as e:
            print(e)
        results_total = results_total + results_now

        sleep(5)

    return results_total

時刻リストと言語リストを与えて検索してもらう。
保存してもらう。
そのあたりのまとめ関数。

# 時刻リストと言語リスト→検索する。

def make_search_in_time_lang_list(random_time_list, filter_lang_list=["ja", "en", None]):

    labels = []
    for lang in filter_lang_list:
        if lang is None:
            labels.append("unspecified")
        else:
            labels.append(lang)
    
    results_dict = {label: [] for label in labels}


    for label, filter_lang in zip(labels, filter_lang_list):
        results_total = []
        for time in random_time_list:
            results_now = []
            try:
                results_now = main_search(time=time, filter_lang=filter_lang)
            except Exception as e:
                print(e)
            results_total = results_total + results_now
            sleep(5)
        results_dict[label] = results_total

    return results_dict


# 保存用

def makejson_from_twlists(filename: str, twlist):
    if '.json' not in filename:
        filename = filename + ".json"
    
    with open(filename, 'w') as f:
        json.dump(twlist, f, indent=2, ensure_ascii=False)


# 日付与えて検索と保存をしてもらう。

def search_and_save_jaens_on(year=2024, month=9, day=3):
    # ランダムに時間決めるよ
    # 言語は中にコーディングして決めてるよ

    target_day = datetime.date(year, month, day)
    str_target_day = target_day.strftime('%Y%m%d')

    print(str_target_day)

    # 2言語ならseparate_numが12でも10000投稿のリミットは大丈夫なはず
    separate_num=6
    hour_range = 24 // separate_num

    random_time_list = []

    for loop_count in range(separate_num):
        hour_add = (loop_count) * hour_range
        time = get_random_time_within_hour(target_day, hour_range=hour_range, hour_add=hour_add)
        print(time)
        random_time_list.append(time)

    filter_lang_list = ["ja", "en", None]

    results_dict = make_search_in_time_lang_list(random_time_list=random_time_list, filter_lang_list=filter_lang_list)

    for label in results_dict.keys():
        makejson_from_twlists(filename="tw_ht_rand_6_" + str_target_day +"_" + label, twlist=results_dict[label])

    return True

ここからは日付横断用


def search_and_save_jaens_matome(year=2024, month=9, day1=4, day2=6):

    # day2はとりたい日付+1！
    # 1ヶ月の中で指定する用の関数

    for day in range(day1, day2):
        search_and_save_jaens_on(year=year, month=month, day=day)
        sleep(10)

    return True


# 今日の前の日までのデータを取る。

def easy_routine(days_width=5):

    today = datetime.date.today()
    print(today)
    year = today.year
    month = today.month
    day = today.day
    
    # days_width = 5

    if days_width < day:
        # 5日間隔で今日が5th日ならやばいので分ける
        day1 = day - days_width
        day2 = day
        search_and_save_jaens_matome(year=year, month=month, day1=day1, day2=day2)
    else: 
        #2つに分割
        day2 = day
        search_and_save_jaens_matome(year=year, month=month, day1=1, day2=day2)
        # 差分 開始が9/2で2日とるなら 8/31のを取る　lm_delda=0
        lm_delta = day - days_width
        lm_end_day = None
        if month == 1:
            # まあ12月は閏じゃないからyear-1とかしなくてもいいんだが
            lm_end_day =  (calendar.monthrange(year-1, 12)[1])
            day1 = lm_end_day-lm_delta
            day2 = lm_end_day+1
            search_and_save_jaens_matome(year=year-1, month=12, day1=day1, day2=day2)
        else:
            lm_end_day = (calendar.monthrange(year, month-1)[1])
            day1 = lm_end_day-lm_delta
            day2 = lm_end_day+1
            search_and_save_jaens_matome(year=year, month=month-1, day1=day1, day2=day2)

    return True