Johnson Liang

Impact: many replies Mitigation: handle scraping error and fill in empty arrays to `hyperinks` As for unit tests, we can mock `scrapUrl` to make it throw exceptions. It should insert...

As discussed in https://g0v-slack-archive.g0v.ronny.tw/index/channel/C2PPMRQGP/2020-07#ts-1594556835.293700 The return type of the API can leverage union type, [as Medium suggested](https://medium.com/@sachee/200-ok-error-handling-in-graphql-7ec869aec9bc).

As discussed in https://g0v.hackmd.io/OC7BneJ3TEGJHge1hWV-0w?both#%E8%A8%8E%E8%AB%96%EF%BC%9A%E8%AE%93%E4%BD%9C%E8%80%85%E8%83%BD%E7%B7%A8%E8%BC%AF%E5%9B%9E%E6%87%89 the priority of this item is set to `low`.

要把 `ListArticle` 與 `SearchArticle` 兩個 API 合在一起。

Tech review: https://hackmd.io/dfAkGHLCShacnOoruGxr0w

Regarding the "Very long string" error emitted by elasticsearch and triggers OOM, I highly suspect that it is because rumors-api is trying to access [`html` field](https://github.com/cofacts/rumors-db/blob/master/schema/urls.js#L7) in `urls` index, and...

I have one thing to add: Should we allow website users inputting `aiModel` and `aiConfidence` in `CreateArticleCategory`? I think in the resolver we should restrict the `appId` that can provide...

3/18 結論 1. `CreateCategory`: 蓋一個新的 util function 決定一個人是不是 admin (另外有個檔案？)，非 admin 的人呼叫的話就 error 2. `CreateArticleCategory`： `checkSecret` 認出新的 shared secret 並且設定 graphql context 的 `appId`。 不同 AI model 可以使用不同 `query.userId`，或使用不同 `aiModel`...

Conclusion: According to [0804 discussion](https://g0v.hackmd.io/Nhc9MnskSsGvUqntz0z7Dg#iOS12-%E7%99%BB%E5%85%A5%E5%95%8F%E9%A1%8C) We will detect iOS

https://cofacts.g0v.tw/article/3ehvvrsuaf7bg Has no related articles, but it's content is highly related with: https://cofacts.g0v.tw/article/AWCHK3peyCdS-nWhult4