[Docs] Explain model selection from chat

[roomote-roomote]

​Opened on behalf of @mrubens. Follow up by mentioning @roomote-roomote, in the web UI, or in Slack.

What changed

• Clarified that users can ask a Roomote agent to switch the model for an existing task from the web or a connected chat.
• Added a direct-message and channel example for choosing the model and reasoning level of work delegated through Fast, while distinguishing that task choice from Fast's orchestration model and deployment defaults.

Why this change was made

A recent support question showed that users could not tell whether model selection worked from a direct message. Roomote already supports conversational per-task switching and model-aware Fast delegation, so documenting the two paths is the smallest way to make the shipped behavior discoverable without adding another command or control.

Impact

Users can now tell how a chat request selects a model, when the choice applies, and which model settings remain unchanged. The Mintlify docs build and broken-link checks pass; no screenshot is attached because this run's visual-proof runtime was unavailable.

[roomote-community]

1 issue outstanding. See task

• Fast chat requests cannot currently set a delegated task's reasoning level (apps/docs/models.mdx:297).

_{Reviewed cffb29d}

[roomote-community]

The Fast launch_task interface only accepts model; it has no reasoningEffort field, and the launch route therefore stamps the delegated task with the deployment coding reasoning level. This new guidance promises that a chat request can choose both values, but max reasoning is currently ignored. Either document model selection only, or add reasoning-level support to the Fast launch path.