LLM Commands

Chat and image generation via an OpenAI-compatible API, with Langfuse observability.

All LLM commands are grouped under /llm.

`/llm chat`

Chat with a language model. Supports text prompts and image attachments for vision models.

Parameter	Description	Required
`prompt`	Your text prompt	Yes (or `image`)
`image`	Image attachment for vision models	No
`model`	Model to use (owner only, autocomplete)	No

Responses are streamed in real-time and displayed with a live-updating embed.
Non-owner users always use the configured LLM_DEFAULT_MODEL.
The bot owner can select from all available models via autocomplete.
All interactions are traced in Langfuse (if configured) with user ID, channel, and guild metadata.
A cancel button is available during streaming.

List all available models from the configured LLM provider.

Info

This command is available to all users, but model selection in /llm chat is restricted to the bot owner.

Generate an image from a text prompt.

Parameter	Description	Default
`prompt`	Description of the image to generate	—
`size`	Image size (`720x480`, `1280x720`, `1024x1024`, `1920x1080`)	`1280x720`
`num_inference_steps`	Number of inference steps	`9`
`true_cfg_scale`	CFG scale	`4.0`
`seed`	Random seed for reproducibility	Random

Warning

This command is currently restricted to the bot owner.

All /llm commands support both guild and user install contexts, meaning they work in: