ChineseBQB: A 5,871-Image Chinese Sticker Archive With an Open JSON Index

What ChineseBQB Solves, and Who It Is Actually For

Chinese chat stickers circulate as loose images inside WeChat groups, screenshots and reposts. There is no canonical index of them, and searching for a specific one usually means scrolling a group chat. ChineseBQB addresses that by collecting stickers into numbered folders and publishing an index. The README states the repository now holds 5,871 stickers across 114 categories, and the top-level directory listing confirms the naming convention: 000Contribution_贡献BQB/, 001Funny_滑稽大佬BQB/, 010Cat_是喵星人啦BQB/, and so on. Each folder name encodes an index number, an English label and a Chinese label.

The audience is narrower than the repository's tone suggests. The README explicitly invites developers: the sticker images and annotation data are described as fully open, and it suggests building a WeChat Mini Program, a crawler demo, or using the set as a machine learning data source. That is the real use case. A developer who wants a labelled set of Chinese reaction images, or a search front end over stickers, gets a ready-made corpus. Someone who wants to send stickers to friends is pointed at the hosted web tool instead, not at the repository.

The Data Layout: Numbered Folders, Two JSON Indexes, Per-Category ZIPs

The mechanism is deliberately plain. Stickers live in top-level directories named with a three-digit prefix, an English word and a Chinese word, such as 024Programmer_程序员BQB/ or 039YaoMing表情包三巨头_姚明BQB/. The README's directory table is generated from those folders, which is why it carries a BQB-DIRECTORY marker and a note that it updates automatically.

Two JSON files form the machine-readable surface. The catalog index is published at https://zhaoolee.com/ChineseBQB/catalog/index.json, and the full image index at https://zhaoolee.com/ChineseBQB/catalog/search.json. The README is specific about resolution: relative paths inside those indexes are resolved against https://zhaoolee.com/ChineseBQB/. That detail matters more than it looks. If you fetch search.json and treat the image fields as absolute URLs, you will get broken links; you have to join them onto that base.

The third piece is distribution. Each category has a direct ZIP download hosted as a release asset under the bqb-downloads tag, for example bqb-111-63f207ae27c17569.zip for the 黑袍纠察队 category. The most recent release listed is bqb-downloads, dated 2026-09-10. That gives you two consumption paths: pull the JSON and fetch images individually, or download a whole category as one archive.

Getting the Index and One Category Without Installing Anything

There is no package to install. The README documents no pip package, no npm module and no CLI. What it documents is a data source, and the directory table itself is the only place that names concrete download URLs, so the first real use is opening one of those URLs. The README gives this example for the 111 category, 13 stickers:

https://github.com/zhaoolee/ChineseBQB/releases/download/bqb-downloads/bqb-111-63f207ae27c17569.zip

That is the direct download the README's table lists for that folder, and fetching it gives you the category as a single archive. The README does not state the archive's internal directory structure, and it does not document checksums for these ZIPs, so inspect the contents before you extract them.

The second entry point is the index. The README names both files and the base URL they resolve against:

https://zhaoolee.com/ChineseBQB/catalog/index.json
https://zhaoolee.com/ChineseBQB/catalog/search.json

index.json is the category catalogue and search.json is the full image index, per the README. It does not publish a schema for either file, so open one and look at its shape before you write parsing code. Do not assume field names.

Where ChineseBQB Breaks Down

The licence is the first problem. The repository metadata carries no licence identifier, and the README does not name one. The README calls the sticker images and annotation data open, but that is a statement of intent, not a licence grant. Stickers are derivative works of films, games, anime and photographs of real people; the categories in the directory listing include 黑神话悟空, 甄嬛传, 老友记, 黄仁勋 and 罗永浩. A permissive licence over the collection would not settle the rights in the underlying frames. If you plan to redistribute the images or ship them inside a product, that is a question for a lawyer, not for the README.

Second, the JSON indexes are described as updating on each release, but their schema is not documented. Field names, whether categories are nested or flat, and whether search.json covers all 5,871 images are all unstated. You will be reverse-engineering the format, and it can change between releases without a version number to pin.

The third failure mode is subtler: this is a mirror, not an API. The site is the delivery mechanism, and the README gives no rate limits, no uptime commitment and no deprecation policy. A production feature that hotlinks sticker URLs from zhaoolee.com is depending on a personal site. Download the assets you need and host them yourself, subject to the licence question above.

Finally, it is the wrong tool for anything requiring semantic search. The categories are human-chosen labels, and the README does not describe tags, captions or embeddings per image. If your requirement is finding a sticker that expresses mild disappointment, the folder names will not get you there.

How It Compares With a General Image Dataset

The obvious alternative for a machine learning use case is a general image dataset such as LAION or a scraped meme corpus. The difference is in what is labelled. A general dataset gives you scale and, in some variants, CLIP-style captions, but no guarantee that the images are Chinese chat stickers with a usable category structure. ChineseBQB gives you the opposite trade: a small, hand-organised set where every image sits under a named category, and where the folder name itself is the label.

Against a meme API such as a hosted giphy-style service, the difference is control. An API gives you search, pagination and a stable contract, and it can revoke your access or change its terms. ChineseBQB gives you raw files and a JSON index with no contract at all. For a one-off dataset build, the second is better. For a product that needs search-as-a-service, the first is better and ChineseBQB is not a substitute.

The one thing ChineseBQB has that neither alternative replicates is the Chinese-language category naming. 滑稽大佬, 莲蓬头男孩, 反港独: these are labels produced by someone inside the meme culture, not translations applied afterwards.

Maintenance, Releases and What the Licence Silence Costs You

The repository is not archived, and the last push was on 2026-09-12, so the collection is being extended. The release channel is bqb-downloads, with the most recent release dated 2026-09-10. The pattern the README describes is a regenerated directory table plus refreshed JSON indexes on each release, which means the practical upgrade cost is low: re-fetch index.json, re-fetch search.json, and re-download only the category ZIPs whose contents changed. There is no migration step and no version pinning, because there is no package.

The cost that is not low is legal. With no licence identifier in the repository metadata and no licence section in the README, you have no stated permission to redistribute the images, only a README sentence calling the data open. That is enough for local experimentation and internal tooling. It is not enough to justify shipping the sticker files inside a commercial product without your own review of the underlying works. Treat the absence of a licence file as the single most important fact on this page.

Editorial conclusion

Adopt ChineseBQB if you need a Chinese sticker corpus for a bot, a search front end or a machine learning sample set, and you are willing to resolve the licence question before redistribution. Do not adopt it if you need an actively documented API, stable asset URLs or a maintained SDK; the README describes a data source and a web tool, not a library. Verify three things first: whether the repository carries a licence file, whether the category ZIPs you need are still attached to the bqb-downloads release, and whether the sticker images you plan to redistribute are ones you have the right to redistribute.

Frequently asked questions

Official sources

1. Issues
2. Project website
3. README
4. Releases
5. zhaoolee/ChineseBQB on GitHub