Separate OOC popup for messages (both user and AI)

[3-MAMA-3]

Dzień dobry.

The Marinara Engine has been game changer for me for the last few weeks. Really enjoying it, and I'm not planning on returning to the original SillyTavern.

Problem:
In my experience, some models occasionally get confused mid-scene—mixing up timelines, contradicting earlier details, etc. This sometimes shows up in Model Thoughts as the AI trying to ask for clarification, but there's no clean way to answer that without embedding [[OOC: ...]] directly in the in-character message text, and no way to see if the AI itself included OOC content without it cluttering the reply.

Proposed feature:
A separate OOC input/popup, mirroring the existing "Model Thoughts" popup, available in two places:

1. While composing a message—an OOC field, ideally right above the message input bar, available BEFORE sending (since the current action buttons only appear on messages that already exist). Whatever's typed there gets sent alongside the message as a distinct block, separate from the in-character text the AI responds to.
2. After sending, and on AI messages—same OOC content becomes viewable/editable via a button in the existing action row (next to Edit, Branch, Delete, Model Thoughts). For AI messages this is read-only—lets me check if the AI included OOC (e.g. "you selected object A earlier and object B this time—which one?") without it needing to be part of the in-character reply.

Optional addition: a toggle to get notified when an AI message contains OOC content, so it doesn't need manual checking every time.

[SpicyMarinara]

Hey pookie, thank you!

You can use Author's Notes for that exact purpose (pencil icon in the top right corner); there you can add OOC instructions and choose where they will be injected. As for models responding OOC, or including your OOC messages inside the reply but not showing them in the actual message, you can create a regex catching them and hiding them from display alone.

Example, for the (OOC: …) format: