Voxta docs

Chat commands

Slash-prefixed commands you can type in chat to inject events, notes, trigger events, capture vision, and more.

You can send slash commands in chat to do things you can't do with a normal message — inject narrated events, hidden notes, trigger scenario events, regenerate the last reply, capture a webcam frame, set flags.

Commands are case-insensitive and prefixed with /.

Message injection

CommandWhat it does
/event textInject a narrated event message. Visible to all, read by the narrator.
/story descriptionSame as /event but the story writer generates the actual text from your description.
/note textSilent note — visible to both AI and user, not read aloud.
/instructions textHidden from the AI, visible to you.
/secret textHidden from you, visible to the AI.
/char textSend a message as the main character (impersonating the AI).
/message --args -- textFull control over role, character, generation caps and expiry. See Custom messages.

Expiring an injected message

/note, /secret and /instructions accept an expiry, which drops the message from the prompt after N conversation turns instead of leaving it there for the life of the chat. Put the options first, then --, then the text:

/secret --expires-after-turns 2 -- {{ char }} is sitting on the couch, watching TV.

Arguments are only read when the text starts with --; anything else is the message verbatim. See Expiring messages.

Chat flow control

CommandWhat it does
(empty message) or *Trigger a reply without injecting anything. Doesn't interrupt currently-playing audio.
/noreply textSend a user message that doesn't trigger an AI reply.
/repeatSpeak the last character reply again.
/summarize allForce summarization of all current chat messages.

Scenario integration

CommandWhat it does
/trigger eventNameFire a scenario event by name.
/setflags flag1 flag2 !flag3Set or unset flags. Prefix with ! to unset.
/context textSet a context string for the current turn.

Vision

CommandWhat it does
/captureGrab a frame from the active Vision Capture source (webcam or screen) and include it in the next reply. Requires the Vision augmentation and a Computer Vision-capable LLM.

See also

On this page