Consultation via Whatsapp

# Consultation via Whatsapp - Chatbot - `Tanya` -> flow qiscus ## RG's Proposal ```plantuml wabiz -> chatbot: Kelas 1 SMP chatbot -> chatbot: <record siswa = kelas 1 SMP> chatbot -> wabiz: Mau tanya apa? wabiz -> chatbot: Tanya chatbot -> qiscus: {Message: "Tanya", command-extra: <biologi+smp>} qiscus -> teacher: <Assign ke teacher yang jago Biologi> teacher -> qiscus: <ambil conversation> qiscus -> chatbot: chatbot -> chatbot: record start conversation wabiz -> chatbot: Berapa luas lingkaran? chatbot -> qiscus: Berapa luas lingkaran? teacher -> qiscus: Jari2 dikali r kuadrat qiscus --> wabiz: ini dipotong, webhook Qiscus diarahkan ke chatbot qiscus -> chatbot: <proposal> chatbot -> chatbot: <record teacherID> chatbot -> wabiz: <proposal> teacher -> qiscus: <end conversation> qiscus -> chatbot: <end conversation (agentEmail)> chatbot -> chatbot: <record end conversation> ``` **Qiscus Note** - Proposal RG possible untuk dilakukan - Bisa mendapatkan TeacherID (agent_id) - Mengimplementasikan Custom channel (https://documentation.qiscus.com/multichannel-customer-service/customization#custom-channel) - Ada intermediary service diantara Chatbot & Qiscus - Tidak memerlukan integrasi Whatsapp channel - Drawbacks - Whatsapp feature tidak dapat digunakan (misal: Outbound message, HSM 24 hours) - Perlu handle terkait HSM 24 Hours - Mengimplementasikan Custom Agent allocation untuk assgin agent based on Division (https://documentation.qiscus.com/multichannel-customer-service/customization#custom-agent-allocation) - Custom channel yang akan allocate Agent dengan data yang didapat dari Chatbot (comment extra: mapel+kelas, room_id) ## Qiscus' Proposal ```plantuml wabiz -> qiscus qiscus -> chatbot: webhook (can have agentID payload) chatbot --> qiscus: reply (in HTTP response) qiscus --> wabiz ``` **Qiscus Note** - Common implementation dari Qiscus - Mungkin akan cukup banyak perubahan di sisi Chatbot - Perubahannya misal parsing data yang sebelumnya dari WA disini dari Qiscus - Keuntungannya Agent bisa outbound message, follow up ke customer. - Mengimplementasikan Custom Agent allocation untuk assgin agent based on Division (https://documentation.qiscus.com/multichannel-customer-service/customization#custom-agent-allocation) - Custom channel yang akan allocate Agent dengan data yang didapat dari Chatbot (comment extra: mapel+kelas, room_id) ## Qiscus send another data to our logger ```plantuml wabiz -> qiscus qiscus -> logger: webhook: log incoming qiscus -> logger: webhook: log outgoing qiscus -> wabiz ``` ## Catch resolve event and track the first comment in room ```plantuml wabiz -> qiscus qiscus -> chatbot: resolve event (contain agent's email and room id) chatbot -> qiscus: get room detail by room id qiscus --> chatbot: return room detail chatbot -> chatbot: get the first and last agent comment timestamp chatbot -> rlo: deduct coin based on first and last timestamp qiscus --> wabiz ``` ### Using inapp chat for teacher ```plantuml actor user entity "rg-chatbot-roboguru" as chatbot entity "rlo-api" as rlo actor teacher as teacher user -> chatbot: Kelas 1 SMP Biologi chatbot -> chatbot: <grade=kelas 1 SMP, subject=Biologi> chatbot --> user: Silakan ketik pertanyaan yg kurang jelas? user -> chatbot: Tanya tentang biologi chatbot -> rlo: Match making user & teacher rlo -> rlo: Lookup guru yg sesuai subject rlo -> teacher: Notify teacher via push notif teacher --> rlo: Confirm availability di app rlo -> chatbot: Notify student for teacher availability chatbot -> user: Notify student teacher -> user: Halo murid, ada pertanyaan apa? user -> teacher: Ini ada pertanyaan .... teacher -> user: Sesi habis, mau lanjut lagi ga? Kalo mau lanjut ketik "LANJUT SESI" user -> chatbot: LANJUT SESI chatbot -> rlo: Check koin rlo --> chatbot: Koin available chatbot -> rlo: Deduct koin teacher -> rlo: End conversation rlo -> chatbot: End session chatbot -> chatbot: Change state to "input rating guru" user -> chatbot: End conversation chatbot -> rlo: End session chatbot -> chatbot: Change state to "input rating guru" ``` - [x] Record semua 2 way conversation - [x] Agent ID - [x] Bisa matching ke teacher untuk subject tertentu - [x] Record start chat & end chat -> untuk penggajian - [x] Multiple phone number - [x] Only one agent can handle a single conversation Conversation - Realtime* - Non realtime Conclusion - Use new account, new dashboard, different agents - Multiple phone number can be ingested to the same dashboard - Qiscus to discuss internally how to handle 3 unchecked requirements above ## Internal ```plantuml actor user entity "rlo-api" as rlo entity "rg-chatbot-roboguru" as chat entity "rg-user" as rguser entity qiscus as qiscus entity logger user -> chat: Tanya soal chat -> rguser: Lookup userID by phone rguser --> chat: userID chat -> rlo: check coin rlo --> chat: coin available chat -> user: Please wait chat -> qiscus: looking for teacher qiscus --> chat: response teacher chat -> rlo: start session rlo -> rlo: deduct coin user -> chat: Berapa luas lingkaran? chat -> qiscus: Berapa luas lingkaran? qiscus -> chat: Tergantung jari2nya chat --> user qiscus -> logger: send webhook event logger -> rlo: store chat qiscus -> chat: end chat chat -> rlo: end session rlo -> rlo: decide to rollback or not ``` ## Questions - Student ask to different Roboguru number at the same time, different subject - Decision: Use the same state across multiple number - Bisa ngeban - Decision: gak bisa - given koin cuma 1, kalau nanya di app terus nanya di WA, bisa data race? - Decision: sekarang bisa, need discus with team rlo - Proposal: Deduct coin when start looking for teacher & will be refund if condition not meet ## notes - messaging team handle things related to chatbot-qiscus communication - roboguru team handle things related to chatbot-rlo communication - roboguru team need to discuss how to handle rlo-coin in app with rlo team - messaging team will push our recently discussed idea to qiscus - need to find where to save conversation log ## feedback and questions for WA flow Link: [spreadsheet](https://docs.google.com/spreadsheets/d/10nNXM0qwiraLcepHpvqa8dSrw2J6XjSH9nvTQzslBns/edit#gid=0) - for all states, what to do given invalid input? repeat the mesage with same wording or no? (there is exist a state that give different wording when given invalid input by user) - need confirm, which one is the source of truth? `Message` or`Message (Edited)`? - should we inform user about which RG account that this number is associated with - it seems column `Next State` is not yet completed. - - row 10: RootRLO - cara kembali ke start roboguru pakai `ROBO` sedangkan banyak yang lain pakai keyword `belajar`, apa bisa dibuat uniform aja? - row 11: PurchaseCoin - pilihan beli koin yang mana ga ditaro di web/deeplink aja kah? - row 21, 22: Question, TutorMatchmaking - immediately change state from Question or waiting for input? - can we make put this two things in one state? - row 25, 27: LearningSession, ContinueLearning - how can student terminate session? - row 26: FinishLearning - need confirmation: it is now written `Lanjut = LearningSession`. perhaps it was supposed to be ContinueLearning instead of LearningSession? - is there a timeout for user so that he can't continue learning with previous tutor? - row 27: ContinueLearning - should we inform user about their remaining coin? - row 29: ContinueLearningNoCoin - what state and what input makes the user go into this state ## APIs ### RLO api

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.