tags: cofacts, meeting note GA: UA-98468513-3 —

20230426 會議記錄

:::info

• 所有會議記錄
• 線上出席：bil,nonumpa
• https://gather.town/app/z3x18KQFgZCX8MeZ/cofacts :::

:potable_water: Release pipeline

:star: Released to production

:globe_with_meridians: Site

• https://github.com/cofacts/rumors-site/releases/tag/release%2F20230425

:rocket: Staging

:robot_face: rumors-line-bot

Testing checklist

https://lin.ee/1QUzEX4nI

• Typescript https://github.com/cofacts/rumors-line-bot/pull/346

• 應可送出「全新訊息」
  • 問訊息來源時選擇「我自己打的」會被擋下。
  • 選擇「整篇轉傳」後會詢問是否要送出訊息。
  • 不同意送出訊息後可以收到感謝。
  • 同意送出訊息後就會送出訊息，並得到：
    • Cofacts article page 按鈕
    • 寫理由的按鈕
    • （若沒開啟推播）應該要看到「開啟小鈴鐺」泡泡，且可打開 setting 頁面
    • 「分享到 Facebook」、「分享到 LINE」且可以正常運作
  • 可從聊天視窗內打開理由視窗，繼續填寫理由送出。查看 article page 看理由是否有被送出。
  • 可以再打開理由視窗，此時會載入上次填寫的理由。修改理由送出後，查看 article page 看理由是否有被送出。
• 送出「沒回應」的舊訊息，應可送出新理由
  • 文章的「N 人回報」應該仍然要 + 1（除非測試者已經針對該篇送過 reply request）。
  • 可從聊天視窗內打開理由視窗，繼續填寫理由送出。查看 article page 看理由是否有被送出。
  • 可以修改理由送出。查看 article page 看理由是否有被送出。
  • （若沒開啟推播）應該要看到「開啟小鈴鐺」泡泡，且可打開 setting 頁面
• 送出「有回應」的舊訊息，應自動回傳回應
  • 應列出訊息所有的回應
  • 選擇回應之後可以幫回應 upvote
  • 可以再次選擇 downvote
  • 選完回應之後，還可以捲回去選其他回應
  • （若沒開啟推播）應該要看到「開啟小鈴鐺」泡泡，且可打開 setting 頁面
• Rich menu 測試
  • 「設定」更改後再次打開，應該會保留原本設定
  • 「教學」可以觸發教學流程

⛔️ Release Blockers

未竟項目

:eye: Under review

https://github.com/cofacts/rumors-line-bot/pull/346

CCPRIP

[Comm] Crowd-source transcript

https://g0v.hackmd.io/@cofacts/rd/%2FOhGIQzoxR5eF6audQuS_FQ [name=nonumpa]

整理 code 先發 PR，UI 可以之後再慢慢改

• hocuspocus server and extension-elasticsearch
• rumors-site

[Document] 開發時用到登入相關的東西 https://github.com/cofacts/rumors-site/issues/242

[Comm] AI transcript

https://g0v.hackmd.io/@cofacts/rd/%2Fwkx286lmTDaFUpgRhnUawQ

[Comm] Article group / co-occurrence

LINE bot tech detail: input batching https://g0v.hackmd.io/f_Ze19PhQuqx_fzOAOkohQ#Technical-detail

GA4 <> UA

承 3/1 時的比較 https://g0v.hackmd.io/zVyuEX5RS5uONLnucsxiYA#GA4-ltgt-UA-%E5%B7%AE%E7%95%B0

LIFF

Both UA (ViewArticle events) and GA4 (view_item events) have exactly the same trigger.

以下以 article 2q2xw134jyblq 為例。

UA:

https://lookerstudio.google.com/u/0/reporting/18J8jZYumsoaCPBk9bdRd97GKvi_W5v-r/page/p_44cex908mc?params=%7B%22df320%22:%22include%25EE%2580%25800%25EE%2580%2580IN%25EE%2580%25802q2xw134jyblq%22%7D

SQL:

SELECT event_date, count(distinct user_id) as user_count, count(*) as event_count
FROM `<table>.events_*`, UNNEST(items) as items
WHERE stream_id = '<LIFF stream ID>' AND event_name = 'view_item' AND items.item_id='2q2xw134jyblq'
GROUP BY event_date
ORDER BY event_date ASC

Result: https://docs.google.com/spreadsheets/d/18jX_N5shuJHibWHJlH9YXfB9Mo-th8d7SdZVFbIQpBI/edit#gid=741917313

• GA4 gets 1.45% less event
• GA4 gets 0.5% more unique users (?)
  • 可能是因為 GA4 是 unique user count 每日相加，UA 是一整個區段的 unique user count

Website

以下以 article 11kmpkou7le3c 為例。

UA uses “page path level 2” of the pageview event to represent article ID.

GA uses dataLoaded event in detail pages instead, which is fired from React.JS code.

UA:

https://lookerstudio.google.com/u/0/reporting/18J8jZYumsoaCPBk9bdRd97GKvi_W5v-r/page/orBY?params=%7B%22df239%22:%22include%25EE%2580%25800%25EE%2580%2580IN%25EE%2580%2580%252F11kmpkou7le3c%22%7D

SQL:

SELECT event_date, count(distinct user_id) as user_count, count(*) as event_count
FROM `<table>.events_*`, UNNEST(items) as items
WHERE stream_id = '<LIFF stream ID>' AND event_name = 'view_item' AND items.item_id='11kmpkou7le3c'
GROUP BY event_date
ORDER BY event_date ASC

Result: https://docs.google.com/spreadsheets/d/18jX_N5shuJHibWHJlH9YXfB9Mo-th8d7SdZVFbIQpBI/edit#gid=741917313

• GA4 gets 0.77% less event
• GA4 gets 0.6% more unique users (?)

TODO

https://g0v.hackmd.io/cJFFXVzgT4OFT6bBLtulGg#-rumors-line-bot