20210512 會議記錄

--- tags: cofacts, meeting note GA: UA-98468513-3 --- # 20210512 會議記錄 :::info - [所有會議記錄](https://g0v.hackmd.io/@mrorz/cofacts-meeting-notes/) - 出席：bil, mrorz ::: ## :potable_water: Release pipeline ### :star: Released to production https://github.com/cofacts/rumors-site/releases/tag/release%2F20210507 - Tico integration - Japanese version https://ja.cofacts.tw ### :rocket: Staging #### :robot_face: rumors-line-bot - [ ] [Ask user not to trust the message just yet when no reply is available #252](https://github.com/cofacts/rumors-line-bot/pull/252) - [ ] [Dockerize LINE bot](https://github.com/cofacts/rumors-line-bot/pull/253) - 缺 README, automatic build 與 log drain ## [LIFF article page](https://g0v.hackmd.io/0RX4MsjRRJmBqJSKVilWMA#article-amp-reply-LIFF) 1. [x] Article LIFF 設計 https://www.figma.com/file/DvmAQjMJCncuPORWKnljM1/?node-id=3004%3A224 2. [ ] 實作 article & reply 3. [ ] 實作 feedback form 4. [ ] 實作 reply request list & form 5. [ ] 實作 related article > 調整了一下 Article LIFF feedback form upvote / downvote 與 reason text input 會放一起填 > 送出後會進入 summary，可回到編輯模式或展開其他人的 feedback > Summary 高度最矮，僅用刷淡底色標注使用者的選擇。這是希望使用者繼續往下捲到其他回應 or 下面的內容，以不干擾使用者繼續往下讀為主。 > 之前討論的「分頁」形式我覺得還是太複雜，會不知道點下去會是編輯還是看其他人回應還怎樣，而且會干擾使用者繼續往下讀。 > 其他人的 feedback 是 upvote / downvote 人混排 > https://www.figma.com/file/DvmAQjMJCncuPORWKnljM1/Cofacts-website-MrOrz?node-id=3042%3A2 > [name=mrorz] ## LINE bot to Docker > 未來打算搬進跟 API、site、elasticsearch 同樣的一台主機。如果試跑起來沒問題的話，之後就能擺脫 Heroku 的 512MB / 1024MB RAM、每天抱怨 memory quota exceed、以及每個月約 45 鎂的使用費囉。 > > 現在 Docker 版 LINE bot 已經放到 staging https://lin.ee/1QUzEX4nI 大家可以測測。 > [name=mrorz] 預計將 rumors-line-bot 從 Heroku 轉移到現有 Linode。 ![](https://s3-ap-northeast-1.amazonaws.com/g0v-hackmd-images/uploads/upload_fd9e3b1ab508e316232719d1be5186dc.png) ### Chatbot error rate baseline 2021/4/26 - 30 202 errors, mostly `Request Timeout` ![](https://s3-ap-northeast-1.amazonaws.com/g0v-hackmd-images/uploads/upload_492ca147a71a2c1102cd5c64612b282e.png) 2021/5/8 - 12: 37 errors mostly still `Request Timeout` ![](https://s3-ap-northeast-1.amazonaws.com/g0v-hackmd-images/uploads/upload_9e2bf31db1706a91c7a984dcec1d07fc.png) - 綠盾牌 Response time > 1 sec 就算 request timeout [name=mrorz] ### Linode server load baseline Peak: 5/3, 使用者較多時 ![](https://s3-ap-northeast-1.amazonaws.com/g0v-hackmd-images/uploads/upload_5d1800e09cbc9667f776164d1131b62e.png) Sampled date: 5/12 20:07 ![](https://s3-ap-northeast-1.amazonaws.com/g0v-hackmd-images/uploads/upload_fdc82759194b6f31853a73ef18a5cce6.png) ![](https://s3-ap-northeast-1.amazonaws.com/g0v-hackmd-images/uploads/upload_f9ad3c9dc5913c0238817bbde7e78837.png) RAM ups & down probably triggered by site-zh periodic restart, which happens every hour at :00 ![](https://s3-ap-northeast-1.amazonaws.com/g0v-hackmd-images/uploads/upload_62bce97a2a86bfd33eb6ed666e745ecb.png) ### Heroku LINE bot load baseline ![](https://s3-ap-northeast-1.amazonaws.com/g0v-hackmd-images/uploads/upload_a411f8fb70f62bad33fc2e64b1bc5a81.png) ![](https://s3-ap-northeast-1.amazonaws.com/g0v-hackmd-images/uploads/upload_d63c8aa6670823ed7fde773ccf7b5cee.png) ### Target 0 errors after migration! ### Migration check :::info Moved to [20210526 會議記錄](/8VGFKx-cR8yEdKlAQ3vv5A) ::: - [ ] `linebot.cofacts.tw` up and running - [ ] AWS credential for cloudwatch is working; `awslogs` is sending output to cloudwatch - [ ] `/` returns - [ ] `/liff/index.html?p=article&articleId=fq3hn0gn4z4e` works - [ ] log drain setup complete - [ ] notification cronjob added - [ ] `lastScannedAt` should be in Redis and available - [ ] long term token and all other ENV vars are done - [ ] LINE developer - [ ] Switch webhook URL to `linebot.cofacts.tw` - [ ] Switch LIFF endpoint URL - [ ] LINE business - [ ] Update rich menu cache busting ## 新的 Dialogflow 自動回覆 https://www.facebook.com/groups/cofacts/permalink/2992380564327086/ Input: 「資料庫裡有幾篇訊息，跟你傳給我的有些接近」「不實訊息常常會被人修改重發。請選擇比較接近的版本」 Output: 「不要學我說話。」 ![](https://s3-ap-northeast-1.amazonaws.com/g0v-hackmd-images/uploads/upload_cccccdda75283582ab83e822474d2a5c.png) - Repeat chatbot phrases - 超過 10 個字，confidence 要是 1 才行 QQ [name=mrorz] - 下面這個就可以 [name=mrorz] ![](https://s3-ap-northeast-1.amazonaws.com/g0v-hackmd-images/uploads/upload_cffcc309fb28e2ca0f2bdb65ad3b1541.png) ---- Input: 「不要學你說話？」 Output: 「你是笨蛋嗎？」 ----- Input: 「我是老虎」 Output: https://www.youtube.com/watch?v=AvsId84Guyc ----- ### Question intent 使用者問真的假的 https://cofacts.tw/search?type=messages&q=真的還假的 https://cofacts.tw/search?type=messages&q=真的嗎？ https://cofacts.tw/search?type=messages&q=幫我查 https://cofacts.tw/search?type=messages&q=請問 ![](https://s3-ap-northeast-1.amazonaws.com/g0v-hackmd-images/uploads/upload_e65982aff93ed0a4fdc404c14a7ca333.png) ## 確診訊息太極鍵譜 > hackmd: https://hackmd.io/@cofacts/r1EUpWTdu 壱ノ型「足跡無用」比對時間後發現根本不可能感染 https://cofacts.tw/reply/S6l-T3kB9w1KR1Ikb5fA 可遇不可求，但效果顯著（27 upvote, 0 downvote）弐ノ型「不用知道」指出足跡太囉唆，給出更簡單的圖請他看這個就好 https://cofacts.tw/reply/ralSLXkB9w1KR1IkkYnq 参ノ型「不如基本功」通常會有 1/3 的人不滿意，但至少有 2/3 會滿意 (?) https://cofacts.tw/reply/UKljOnkB9w1KR1IkD5DR 跟公告足跡完全反白 - 當成假的、來亂的 https://cofacts.g0v.tw/article/159qco3doxrxe 「南方澳1人.冬山1人.羅東2人.壯圍1人」 https://cofacts.g0v.tw/article/2z6shkjujo932 OOO 不要去 - 如果是真的也消毒了啦 https://cofacts.tw/article/1rttomwy5ttsx （調整語氣：因為不是說「不要去」而是說「去過了要小心」，觀念正確） https://cofacts.tw/article/156v6hkwdstg0 ## Good design award https://www.g-mark.org/about/?locale=zh_TW https://www.g-mark.org/guide/2021/guide2.html#guide2-2 ![](https://s3-ap-northeast-1.amazonaws.com/g0v-hackmd-images/uploads/upload_1a5165855e2efe17d174b942daec521a.png) ※ 報名開始至獲獎為止所需費用的估算參考：一次評審費 11,000日圓＋二次評審費 58,300日圓＋獲獎推廣費88,000日圓＝合計 157,300日圓（以小型且不使用電源的展示作品為例。）

Syntax	Example	Reference
# Header	Header	基本排版
- Unordered List	Unordered List
1. Ordered List	Ordered List
- [ ] Todo List	Todo List
> Blockquote	Blockquote
Bold font	Bold font
Italics font	Italics font
~~Strikethrough~~	~~Strikethrough~~
19^th^	19^th
H~2~O	H₂O
++Inserted text++	Inserted text
==Marked text==	Marked text
[link text](https:// "title")	Link
![image alt](https:// "title")	Image
`Code`	`Code`	在筆記中貼入程式碼
```javascript var i = 0; ```	`var i = 0;`
:smile:		Emoji list
{%youtube youtube_id %}	Externals
$L^aT_eX$	L^aT_eX
:::info This is a alert area. :::	This is a alert area.