Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Suggested replies for email #8505

Closed
nimishavijay opened this issue May 31, 2023 · 7 comments · Fixed by #8863, nextcloud/documentation#11507 or nextcloud/documentation#11512
Closed

Suggested replies for email #8505

nimishavijay opened this issue May 31, 2023 · 7 comments · Fixed by #8863, nextcloud/documentation#11507 or nextcloud/documentation#11512

Comments

@nimishavijay
Copy link
Member

nimishavijay commented May 31, 2023

Is your feature request related to a problem? Please describe.

It could be useful to have some suggested replies for a message to quickly send an email, especially simple replies like "Thank you", "That works for me", or something similar

Describe the solution you'd like

There could be a few suggested replies at the bottom of the last message in a thread, with some relevant suggestions based on the contents of the message. Clicking on it could send the message immediately (with the option to undo of course). This would save 2 clicks and some typing.

Rough designs and some variations in placement and button style, cc @jancborchardt and @szaimen for feedback
Note that in the mockups the action is reply, we could also do reply all, what do you think?

Implementation

  • Synchronous mode, like the thread summary. Deployment needs fast LLM. Slow LLM not supported for the first version.
  • Use free prompt API if possible, or ask Nextcloud integration team for help with a new backend/API.

Work packages

  • MVP
  • Admin docs
  • User docs

Describe alternatives you've considered

No response

Additional context

No response

@ChristophWurst
Copy link
Member

This requires some sort of machine learning to process and "understand" the text. I guess classification into types of emails could be one approach, where you then suggest an "Ok thank you" for informational messages and "Yes"/"No" for yes-no questions. The more complex approach is to provide replies with actual context information, e.g. confirming/declining a date suggestion.

The ML has to also be able to work with all supported languages, not just english.

The training of this data will be hard to do unsupervised because there is no high quality reference data to train. So I'm afraid we have to ship a pretrained model.

@ChristophWurst
Copy link
Member

ChristophWurst commented May 31, 2023

Smart replies on Android are a similar feature: https://www.techmesto.com/enable-disable-smart-replies-android-10/. Interesting enough it only works with English text: https://developers.google.com/ml-kit/language/smart-reply

@jancborchardt jancborchardt moved this to 🧭 Planning evaluation / ideas in 🖍 Design team May 31, 2023
@ChristophWurst
Copy link
Member

ChristophWurst commented May 31, 2023

I'm not familiar with it but if there is a way to use the smart reply model together with https://www.tensorflow.org/js it could be possible to process the messages in the browser and therefore have more options than PHP+rubix/ml.

Edit: the model is not available for tfjs.

@jancborchardt jancborchardt moved this from 🧭 Planning evaluation / ideas to 🏗️ At engineering in 🖍 Design team Aug 22, 2023
@hamza221 hamza221 mentioned this issue Sep 13, 2023
6 tasks
@jancborchardt jancborchardt moved this from 🏗️ At engineering to 🧭 Planning evaluation / ideas in 🖍 Design team Oct 19, 2023
@ChristophWurst ChristophWurst moved this to 🧭 Planning evaluation (dont pick) in 💌 📅 👥 Groupware team Dec 21, 2023
@jancborchardt jancborchardt moved this from 🧭 Planning evaluation / ideas to 🏗️ At engineering in 🖍 Design team Dec 21, 2023
@ChristophWurst ChristophWurst moved this from 🧭 Planning evaluation (dont pick) to 📄 To do (~10 entries) in 💌 📅 👥 Groupware team Dec 21, 2023
@hamza221 hamza221 moved this from 📄 To do to 🏗️ In progress in 💌 📅 👥 Groupware team Dec 28, 2023
@ChristophWurst
Copy link
Member

Work packages

* MVP

* Admin docs

* User docs

Just added. @hamza221 I think this feature should be documented

@github-project-automation github-project-automation bot moved this from 🏗️ At engineering to 🎉 Done in 🖍 Design team Jan 12, 2024
@github-project-automation github-project-automation bot moved this from 🏗️ In progress to ☑️ Done in 💌 📅 👥 Groupware team Jan 12, 2024
@hamza221
Copy link
Contributor

Reopened to track docs

  • Admin
  • User

@hamza221 hamza221 reopened this Jan 12, 2024
@github-project-automation github-project-automation bot moved this from ☑️ Done to 📄 To do in 💌 📅 👥 Groupware team Jan 12, 2024
@hamza221
Copy link
Contributor

hamza221 commented Feb 6, 2024

@ChristophWurst what should be documented on the user side ?
I couldn't think of anything 😅

@ChristophWurst
Copy link
Member

@ChristophWurst what should be documented on the user side ?
I couldn't think of anything 😅

Imagine explaining the feature to a friend. For example, tell that it provides a suggested reply based on the thread the user sees and that clicking the button prefills the composer of the reply. Mention that the feature might not be available because admins have to enable it.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment