bilibili-comments

name	bilibili-comments
description	Fetch and export all comments from a Bilibili video, including first-level comments and second-level replies, through the user's already logged-in Chrome browser. Use when the user asks to crawl, download, export, save, or analyze comments for a Bilibili/B站/哔哩哔哩 video URL, BV id, or av id.

Bilibili Comments

Workflow

Use the bundled script for comment export. Do not manually copy cookies.

1. Ensure Chrome is open and remote debugging is enabled at chrome://inspect/#remote-debugging.
2. Run:

node E:\projects\codexProjects\AI博主爬取\.agents\skills\bilibili-comments\scripts\fetch_comments.mjs --video "https://www.bilibili.com/video/BVxxxx" --output ".\bili-comments.jsonl"

3. For a quick smoke test, limit root comments:

node E:\projects\codexProjects\AI博主爬取\.agents\skills\bilibili-comments\scripts\fetch_comments.mjs --video "BVxxxx" --max-root 20 --no-replies --output ".\sample.jsonl"

4. For multiple videos, always use batch mode so one Node process keeps one CDP websocket connection open across all videos:

node E:\projects\codexProjects\AI博主爬取\.agents\skills\bilibili-comments\scripts\fetch_comments.mjs --videos-file ".\videos.json" --output-dir ".\comments"

videos.json must be a JSON array. Entries can be strings or objects:

[
  "BVxxxx",
  { "video": "https://www.bilibili.com/video/BVyyyy", "bvid": "BVyyyy", "output": "comments/BVyyyy.jsonl" }
]

5. Report the output path and counts: root comments, reply comments, total rows, and duplicate skips.

Script Options

• --video <url|BV|av>: Required. Bilibili video URL, BV id, or av id.
• --videos-file <file>: Optional batch mode. Use this for more than one video to avoid reconnecting to Chrome for every video.
• --output-dir <dir>: Optional batch output directory. Used with --videos-file.
• --output <file>: Optional. Defaults to bilibili_comments_<BV-or-aid>_<timestamp>.jsonl in the current directory.
• --max-root <n>: Optional. Limit first-level comments for testing.
• --no-replies: Optional. Export only first-level comments.
• --delay-ms <n>: Optional. Delay between root-comment API pages. Default 500.
• --reply-delay-ms <n>: Optional. Delay between reply API pages. Default 300.
• --keep-tab: Optional. Keep the temporary Bilibili tab open for debugging.

Output Schema

Each JSONL row uses this shape:

{
  "comment_id": "303482460912",
  "parent_comment_id": "0",
  "root_comment_id": "303482460912",
  "level": 1,
  "create_time": 1779613639,
  "video_id": "116618478750430",
  "bvid": "BV...",
  "content": "评论正文",
  "user_id": "1655181908",
  "nickname": "用户名",
  "sex": "保密",
  "sign": "",
  "avatar": "https://...",
  "sub_comment_count": "13",
  "like_count": 60,
  "last_modify_ts": 1781857457856
}

parent_comment_id is 0 for first-level comments. For replies, it comes from Bilibili's parent field; root_comment_id identifies the first-level thread.

Reliability Notes

• The script fetches from inside the logged-in Chrome page context, so it uses the browser's current Bilibili login state without printing or exporting cookies.
• Chrome remote debugging is a server, but each Node run is still a new CDP client websocket connection. Batch mode avoids repeated client connections inside a multi-video crawl.
• If connection times out, ask the user to keep Chrome open, confirm any browser authorization dialog, and retry.
• If chrome://inspect/#remote-debugging shows no running server, ask the user to toggle "Allow remote debugging for this browser instance" off and on.
• For very large videos, increase delays instead of running tight loops.