tags: cofacts, meeting note GA: UA-98468513-3 —
20231004 會議記錄
:::info
- 所有會議記錄
- 線上出席:bil, orz, nonumpa, 4000, cai
- https://gather.town/app/z3x18KQFgZCX8MeZ/cofacts :::
:potable_water: Release pipeline
:star: Released to production
:electric_plug: API
GetYdocfix https://github.com/cofacts/rumors-api/releases/tag/release%2F20230929
:robot_face: rumors-line-bot
- Group bug https://github.com/cofacts/rumors-line-bot/releases/tag/release%2F20230928
:rocket: Staging
:electric_plug: API
- https://github.com/cofacts/rumors-api/pull/318
Testing checklist
https://lin.ee/1QUzEX4nI
- 1:1 可送出全新影片、圖片、聲音
- 點到網站,可以看到 AI 逐字稿自動填入
:globe_with_meridians: Site
Testing checklist
主攻上週未竟項目
- 有歷史紀錄
- 時間、作者資訊均有
http://dev.cofacts.tw/
登入自有帳號後檢測:
- AI 逐字稿填入後,edit history 有呈現送出訊息的時間,作者為「AI Transcript」
⛔️ Release Blockers
未竟項目
- 有影片 (7/1 不禮讓行人)送入 chatbot always 會 busy
:eye: Under review
:electric_plug: API
- Upgrade jest https://github.com/cofacts/rumors-api/pull/319
- Fix 上週提到的 lib0 問題
- https://github.com/cofacts/rumors-api/pull/318
CCPRIP
[Comm] Transcript
https://g0v.hackmd.io/@cofacts/rd/%2FOhGIQzoxR5eF6audQuS_FQ [name=nonumpa]
計畫:補 test、弄 monorepo 網站補提示
AI: API 寫入 ydoc
Design doc https://g0v.hackmd.io/wkx286lmTDaFUpgRhnUawQ#CreateMediaArticle
Migration plan
Summary from https://g0v.hackmd.io/wkx286lmTDaFUpgRhnUawQ#Migration-of-exisitng-media-articles :
- Use open data to put articles to transcribe in Google sheet
- Run Whisper / OCR using Colab and store back to Google sheet
- Call
writeAITranscript(articleId, text)for all transcribed items
Note: No airesponses doc is created in this process
Execution

Estimated 20,000 articles to transcribe.
- Sort 20,000 articles to transcribe by creation date, latest first
- Process first 1,000 articles, calculate time needed
- Plan for the rest 19,000 migration
Other issues
OCR的偵測敏感度(?)可以調比較低嗎 傳疾管署那種宣導圖卡,OCR都會把沒有文字的變成文字 [name=cai]
[Comm] Article group / cooccurrence implementation
我在思考 article group 處理多則訊息 時,跟現有 chatbot context 裡面的某些欄位(尤其是
selectedArticleId)會有衝突的問題 正在思考「先整理 chatbot context 把東西移到 postback action」的方向時,發現現在把selectedArticleId記在 context 會有這個 bug: https://github.com/cofacts/rumors-line-bot/issues/327 ` 我想要先把selectedArticleId移動到 postback action 來修好這個 bug 再來逐步移除不需要的 context 最後再回頭思考 article group 與 chatbot context 的問題,說不定清完之後發現就沒有那些不 compatible 的 context field 了 [name=mrorz]
postback.data有長度限制 [name=nonumpa]
[Infra] Migrate to Cloudrun
https://g0v.hackmd.io/BRsJOevWSbyUMBSZEVVWrA#Phase-1-rumors-site-amp-rumors-line-bot-%E4%B8%8A-Google-cloud-run
Website
Previous TODO: https://g0v.hackmd.io/ulv1SGtBRWmhxnE9Lpv9TQ?both#TODO
- Github image build triggers deploy: https://github.com/cofacts/rumors-site/pull/549
- Service account:
rumors-siteservice account runs container, has no permissionrumors-site-deployis used on Github actions- Shared across
rumors-siteandrumors-line-bot(see below) - Went through Workload Identity Federation setup
- Authenticates to Google via Github’s token
- No need to generate and maintain JSON keys
- It has enough permission to create Cloud run deployments and impersonate as individual service accounts
- Shared across
Chatbot
Original
- Chatbot on Linode
- Notification job in Crontab
- Redis on Linode
- MongoDB on MongoDB Atlas
Now (Staging only)
- Chatbot on Cloud Run as a service
- Notification job on Cloud Run as a job
- Automatic deploy after Github action build https://github.com/cofacts/rumors-line-bot/pull/364
- Reuse service account
line-botto run the cloud run containers- Already used in Linode to access Dialogflow
- Will remove its key after all line-bot is running on Cloud Run
- Redis on Redis Cloud
- Free for now
- Production grade: 8usd / mo for 100MB

- MongoDB on MongoDB Atlas
- Still free XD
TODO
- Apply to en, ja production, observe 1 month
- Apply to tw production
Hugging Face
Hugging Face dataset page: https://huggingface.co/datasets/Cofacts/line-msg-fact-check-tw Example Colab (含有如何 join table 做一個 classifier 需要的 train data) https://colab.research.google.com/drive/1qdE-OMJTi6ZO68J6KdzGdxNdheW4ct6T [name=mrorz] 相較多數資料集 是詳細了很多 但我覺得詳細說明還蠻不錯的 example 的話算是蠻簡單的,可以很直接地讀進資料(不過這應該是用 HF 這套蠻常見的就是),能想到的就是有個圖輔助你要 join tables 之間的概念可以錦上添花一些 [name=pai]
https://github.com/cofacts/opendata
- 關閉 google form
- Github 上只會把使用者導向到 HuggingFace
- 如何生檔案的部分換寫在
CONTRIBUTING.md
- 如何生檔案的部分換寫在
MozTW 服務學習
徵求「交大資工服務學習課程自由軟體組」的參與專案 大家好,我跟軟自協的 @frank 、前教育部自由軟體中心的 Eric Sun 及維基社群的 @s8321414 每學期都會去交大資工協助服務學習自由軟體組。由於今年系上想要擴大舉辦,因此想要徵求一些有興趣帶領大一同學:beginner::beginner::beginner:的坑主來加入。 有興趣提供貢獻機會,請把專案資料填寫到此份投影片:https://docs.google.com/presentation/d/1ZnbJy5_qpjOYPtkoH88v3X8IGVQ6yNVEloRipg_MVzc/edit?usp=sharing (參與項目不限寫 code) 我們將會在 10/13 去學校舉辦說明會,到時會將投影片提供給同學參考,我也會快速介紹一輪。會後供他們自由選擇有興趣的項目,再直接按照投影片上的資訊聯絡各專案坑主。 (另外,系上也希望額外邀請參與者來系上演講,分享自己的開源及程式貢獻社會的經驗等(此項有演講費)。時間可安排在十一月或十二月下午或晚上,有興趣也請舉手一下。) [name=irvin]
- Code: Good first issue
- 闢謠
:::success Johnson 寫 :::
大松籌備
報名
2023/10/15
- https://github.com/cofacts/rumors-site/issues
- https://github.com/cofacts/rumors-line-bot/issues
:::success Johnson 報 :::
來信處理
- 喬治亞
- 直銷
回溯性投資
月底出計算的 spreadsheet 12 月底結案 需要錢包的解釋 Zoey 線上說明會 / 另外辦一個實體的