16466: clarify role of Orchestrator agent in Theia AI chat interactions

[CamilleLetavernier]

What it does

• improve the Chat UX for first time Theia AI users (when Theia AI is not fully configured)

How to test

• start from a fresh workspace
  • alternatively: unset all your API Keys (Anthropic, OpenAI, Gemini) and Ollama Configs (models, ...)
  • unset the default agent setting
• start a new conversation in AI Chat
  • the welcome page should indicate a configuration error, indicating that Models are not configured
  • the page also indicates the error messages reported by each model (Typically "missing API Key")
  • there is no error for Ollama, because Ollama no longer provides any model out of the box (we check "Language Models" rather than "Providers").
• add an API Key (can be a mock key) for OpenAI, Google or Anthropic (or any other provider)
  • should also work when adding a model for Ollama
• the Welcome Page changes: it now requests the user to pick a default Agent
  • this can be done either via one of the recommendations (single click) or by navigating to the settings via the suggested link and entering a custom value

Screen 1: AI Features are disabled (Not change for this screen)

Screen 2: AI Models are not configured (New screen 🌟 )

Screen 3: Default Agent has not been selected (New screen 🌟 )

Screen 4: Everything is (probably) fine, happy chatting! (No change for this screen)

Follow-ups

• Remove Orchestrator agent and related setting
• Align styling of all different chat views
• Update the preference change listener for the model entries for example in the general follow Preference change listeners should handle undefined values for newValue properly #16645

Breaking changes

• This PR introduces breaking changes and requires careful review. If yes, the breaking changes section in the changelog has been updated.

Attribution

Review checklist

• As an author, I have thoroughly tested my changes and carefully followed the review guidelines
• User-facing text is internationalized using the nls service (for details, please see the Internationalization/Localization section in the Coding Guidelines)

Reminder for reviewers

• As a reviewer, I agree to behave in accordance with the review guidelines

[CamilleLetavernier]

I've rebased the PR on the current master, which included a refactoring of the Welcome Page from regular React to (Localized)Markdown: issues/16470

[ndoschek]

Hi @CamilleLetavernier, thanks for this improvement, the flow works great and will guide new users much better now! 🙌

I added a few minor comments inline and have one general thought:
The views could be aligned in a follow-up so they all use consistent headings and layout. The first and default view weren't touched by this PR, so handling it separately should be fine imo. When testing the new user flow, the views switch between different styles, fonts, sizes, and alignments, which feels a bit jumpy to the eye.

[ndoschek]

In general I would like to double check the font sizes, margins and so on and use theia css variables where possible, but as mentioned in my general comment, it would also be fine for me to rework the styling in a follow up together with the other views we have.

[ndoschek]

@JonasHelming could you also take a look at the flow in general and let us know if you’re fine with it or have any additional feedback? TIA!

[JonasHelming]

Thanks, this is a nice improvement for initial users!

1. We still have an orchestrator related setting
2. I was not able to get the "error" page, eve though I removed all AI related settings, so I could not test this
3. Feedback to the screenshot above:

I think we should make the message feel less like an exception, as users will very likely run into it.

Suggestion:

Please configure at least one language model provider

If you want to use OpenAI, Anthropic or GoogleAI hosted models, please enter an API key in the settings. (maybe link the provider names to their settings page)
If you want to use another provider such as Ollama, please configure it in the settings and adapt agents or a model alias to use your custom model.
Note: Some agents, such as Claude Code do not need a provider to be configure, just continue in this case.
See the documentation for more details.

Current Configuration State

message from providers

[CamilleLetavernier]

I've rebased the PR to main and addressed most of the review comments. In the meantime, the AI Configuration view was rewritten quite significantly, and I believe it introduced a regression, as the Welcome Message Provider was no longer notified when agents were enabled/disabled. I've patched the agent-service accordingly (it's not directly related to this PR; but I need this notification mechanism to track configuration changes and properly refresh the Welcome message).

We still have an orchestrator related setting

This PR only removed the default agent, but didn't change anything about the Orchestrator itself. The Agent (and its settings) still exist.

I was not able to get the "error" page, eve though I removed all AI related settings, so I could not test this

The Config warning message will go away if at least one provider is "properly" configured; and some providers are fully (or almost fully) configured out of the box (e.g. Ollama with its default values). So you also need to remove e.g. all Ollama models to see this configuration screen.

Feedback to the screenshot above:
I think we should make the message feel less like an exception, as users will very likely run into it.

👍

[ndoschek]

Looks great, thanks a lot for the updates @CamilleLetavernier!

I added a few notes to the follow-ups as discussed offline.
I also noticed a small issue with updating the language models: if all Google models are removed and then the defaults are restored, no update is triggered, but this is a known problem outside the scope of this PR, and we already planned to address it next month, so no action needed here (see also #16645)

Works great for me, approved from my side 🎉