Settings

The app no longer has a settings page. Basic, Compact and Push are system-wide settings: one copy, shared by every account, changed by the platform operator in the admin console.

What this means for you:

• You neither need to nor can change them inside the app;
• Changes are picked up by the agent runtime immediately, with no restart;
• Ask the platform operator when something needs adjusting.

Basic (observability, agent behaviour and LLM retries)

• Log level: Debug / Info / Warning / Error.

• Metrics: whether agent execution statistics and performance metrics are collected.

• Audit log: whether key actions and configuration changes are recorded.

• Show context to the AI: when on, each turn reports token usage and related state to the agent so it notices earlier that the window is filling up. Useful on long sessions; turning it off saves tokens per request.

• LLM retries: connection failures and mid-stream drops are counted separately, with exponential backoff plus jitter. On unstable networks the operator will usually raise the stream retry count first.

Compact (context compaction)

Controls when long sessions are compacted and how work is batched: delay turns, minimum token threshold, reminder frequency, batch size and batch timeout. A higher threshold compacts later and keeps more raw text; a lower one compacts earlier and uses less context.

Push (notifications)

The master switch and device targeting (all / mobile only / desktop only) are system-wide. Device subscriptions always require the user’s own browser permission, but the in-app notification entry is currently closed; no browser notifications will arrive until the new delivery mechanism ships.