feat: chatty ui

[julien-nc]

Additional requirement: Mention in the README the length of the assistant memory (number of messages used to generate a new one). Mention it can be configured with an app config value.

[kyteinsky]

There's no admin docs though, should I create one or just add it to the README in the root dir?

[kyteinsky]

It would be nice to have a visual feedback when the LLM answer is being computed. How about a simple loading spinner (NcLoadingIcon for example) where the next message will appear?

The input box shows the "Thinking" text and there is a breathing animation as well. Maybe it needs to be more prominent. Do you suggest a loading icon in addition to this?

Something goes wrong when regenerating. The last user message (before the assistant message to regenerate) is deleted (from the UI only). It comes back after switching sessions or reloading. This does not happen if one asks to regenerate a message right after loading the session. I can give you more details if needed.

Fixed.

The instruct prefix is lost when chat_last_n_messages is exceeded. See inline comment in the controller.

Fixed.

[marcelklehr]

There's no admin docs though, should I create one or just add it to the README in the root dir?

Could be done somewhere on this page? https://docs.nextcloud.com/server/latest/admin_manual/ai/app_assistant.html

[marcelklehr]

I guess/hope it's the same for most LLMs.

It isn't, I think.

[kyteinsky]

There are a few more tweaks to go, proposed by the design team.

[marcelklehr]

Super cool :tada: