diff --git a/main.py b/main.py
index df048f4..f1aac9e 100644
--- a/main.py
+++ b/main.py
@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import asyncio
+import random
 from datetime import date, datetime, time, timedelta
 from typing import TypedDict, cast
 
@@ -9,7 +10,9 @@ from bs4 import BeautifulSoup
 from requests.exceptions import Timeout
 
 import queries_to_answers as q2a
-from db.models import Video, VideoHistory
+from db.models import Comment, Video, VideoHistory
+from nizika_ai.consts import Character, GPTModel, QueryType
+from nizika_ai.models import Query
 
 KIRIBAN_VIEWS_COUNTS: list[int] = sorted ({ *range (1_000, 10_000, 1_000),
                                             *range (10_000, 1_000_001, 10_000),
@@ -44,28 +47,16 @@ async def report_kiriban (
         (views_count, video_info, uploaded_at) = (
                 kiriban_list.pop (random.randint (0, len (kiriban_list) - 1)))
         since_posted = datetime.now () - uploaded_at
-        uri = f"https://www.nicovideo.jp/watch/{ video_info['contentId'] }"
-        (title, description, thumbnail) = fetch_embed_info (uri)
-        try:
-            upload = client.com.atproto.repo.upload_blob (
-                    io.BytesIO (requests.get (thumbnail,
-                                              timeout = 60).content))
-            thumb = upload.blob
-        except Timeout:
-            thumb = None
-        comments = nico.get_comments (video_info['contentId'])
+        video_code = video_info['contentId']
+        uri = f"https://www.nicovideo.jp/watch/{ video_code }"
+        (title, description, _) = fetch_embed_info (uri)
+        comments = fetch_comments (video_code)
         popular_comments = sorted (comments,
                                    key      = lambda c: c.nico_count,
                                    reverse  = True)[:10]
         latest_comments = sorted (comments,
                                   key       = lambda c: c.posted_at,
                                   reverse   = True)[:10]
-        embed_external = models.AppBskyEmbedExternal.Main (
-                external = models.AppBskyEmbedExternal.External (
-                    title       = title,
-                    description = description,
-                    thumb       = thumb,
-                    uri         = uri))
         prompt = f"{ since_posted.days }日と{ since_posted.seconds }秒前にニコニコに投稿された『{ video_info['title'] }』という動画が{ views_count }再生を突破しました。\n"
         prompt += f"コメント数は{ len (comments) }件です。\n"
         if video_info['tags']:
@@ -78,10 +69,19 @@ async def report_kiriban (
 ```html
 { video_info['description'] }
 ```
-このことについて、ニジカちゃんからのお祝いメッセージを下さい。
-ただし、そのメッセージ内には再生数の数値とその多さに応じたリアクションを添えてください。
-また、ぜひ投稿からこの再生数に至るまでにかかった時間や、つけられたタグ、コメントに対して思いを馳せてください。
-好きなコメントがあったら教えてね。"""
+このことについて、何かお祝いメッセージを下さい。
+ただし、そのメッセージ内には再生数の数値を添えてください。
+また、つけられたタグ、コメントからどのような動画か想像し、説明してください。"""
+        query = Query ()
+        query.user_id = None
+        query.target_character = Character.DEERJIKA.value
+        query.content = prompt
+        query.query_type = QueryType.KIRIBAN.value
+        query.model = GPTModel.GPT3_TURBO.value
+        query.sent_at = datetime.now ()
+        query.answered = False
+        query.transfer_data = { 'video_code': video_code }
+        query.save ()
         # 待ち時間計算
         dt = datetime.now ()
         d = dt.date ()
@@ -195,6 +195,53 @@ def create_bs_from_url (
     return BeautifulSoup (req.text, 'hecoml.parser')
 
 
+def fetch_embed_info (
+        url:    str,
+) -> tuple[str, str, str]:
+    title:          str = ''
+    description:    str = ''
+    thumbnail:      str = ''
+
+    try:
+        res = requests.get (url, timeout = 60)
+    except Timeout:
+        return ('', '', '')
+
+    if res.status_code != 200:
+        return ('', '', '')
+
+    soup = BeautifulSoup (res.text, 'html.parser')
+
+    tmp = soup.find ('title')
+    if tmp is not None:
+        title = tmp.text
+
+    tmp = soup.find ('meta', attrs = { 'name': 'description' })
+    if tmp is not None and hasattr (tmp, 'get'):
+        try:
+            description = cast (str, tmp.get ('content'))
+        except Exception:
+            pass
+
+    tmp = soup.find ('meta', attrs = { 'name': 'thumbnail' })
+    if tmp is not None and hasattr (tmp, 'get'):
+        try:
+            thumbnail = cast (str, tmp.get ('content'))
+        except Exception:
+            pass
+
+    return (title, description, thumbnail)
+
+
+def fetch_comments (
+        video_code: str,
+) -> list[Comment]:
+    video = Video.where ('code', video_code).first ()
+    if video is None:
+        return []
+    return video.comments
+
+
 async def report_nico (
 ) -> None:
     ...
diff --git a/nizika_ai b/nizika_ai
index 9e136a7..3fd16bd 160000
--- a/nizika_ai
+++ b/nizika_ai
@@ -1 +1 @@
-Subproject commit 9e136a7cb3baf97b384a07c4af31205d745e7200
+Subproject commit 3fd16bd1c242508e8a2f79618aa1e32953a6a840