Autosearch

Skill Attribution

Source: self-written, plan autosearch-0418-channels-and-skills.md § F002c.

name: bilibili description: Chinese tech video platform with tutorials, conference recordings, and uploader-authored articles, via TikHub. version: 1 languages: [zh, mixed] methods:

• id: via_tikhub

impl: methods/via_tikhub.py requires: [env:TIKHUB_API_KEY] rate_limit: {per_min: 60, per_hour: 1000}

• id: api_search

impl: methods/api_search.py requires: [] rate_limit: {per_min: 30, per_hour: 500} fallback_chain: [api_search, via_tikhub] when_to_use: query_languages: [zh, mixed] query_types: [tutorial-video, comparison, breakdown, tech-opinion] avoid_for: [text-only-query, academic-papers] quality_hint: typical_yield: medium chinese_native: true layer: leaf domains: [chinese-ugc] scenarios: [chinese-native, video-content, tutorial, tech-opinion] model_tier: Fast experience_digest: experience.md

Overview

Bilibili is a Chinese video platform with strong coverage in tutorials, hardware comparisons, software explainers, gaming, and creator-led technical commentary. It is useful when the user wants Chinese-language visual content rather than text-first references.

For autosearch coverage, this channel fills a gap between global video search and Chinese-native creator ecosystems. It matters because many Chinese tutorials, teardown videos, and side-by-side product breakdowns are published on Bilibili long before they appear elsewhere.

When to Choose It

• Choose it for Chinese tutorial-video and breakdown queries.
• Choose it for device comparisons, creator explainers, and tech-opinion in video form.
• Choose it when visual demonstration matters more than short text description.
• Prefer it over YouTube when the target audience, terminology, or creator ecosystem is Chinese-native.
• Avoid it for pure text lookup or formal academic paper search.

How To Search (Planned)

• via_tikhub - Use TikHub's paid Bilibili general search API to retrieve mixed result groups, then map only video and article results into normalized evidence.
• via_tikhub - Strip Bilibili search-hit HTML markers, normalize uploader identity, and derive canonical video or article URLs from bvid / article ids when needed.
• api_search - Call Bilibili search endpoints for videos and creators using Chinese or mixed query text, then rank by topical relevance and engagement.
• api_video_detail - Fetch richer metadata for shortlisted videos with authenticated detail access when cookie:bilibili is available.
• api_video_detail - Normalize title, uploader, publish time, duration, play stats, and canonical video URL.

Known Quirks

• TikHub billing applies per request, so this route should avoid wasteful exploratory fan-out.
• Search titles and descriptions can include <em class="keyword"> hit markers that must be stripped before ranking or display.
• Only video and article result types are mapped today; user, live, and topic-style results are intentionally ignored.
• Basic search works without auth, but video detail and richer metadata are more stable with a valid cookie.
• Many results are entertainment-adjacent, so ranking must distinguish tech and tutorial intent carefully.
• Some high-signal tutorials use slang or fandom terminology that generic keyword matching may miss.
• Video popularity can dominate search ordering even when a smaller creator has the better explanation.

Quality Bar

• Evidence items have non-empty title and url.
• No crash on empty or malformed API response.
• Source channel field matches the channel name.