AI Features

Lunova integrates AI in several places to reduce staff workload — from auto-handling routine ticket messages to Q&A commands and content moderation assistance.

ℹ AI features run through the bot's local Ollama service. Server admins control access with the per-server AI toggle.

Enabling AI

AI features are off by default per server. To turn them on:

Go to Dashboard → AI Assistant and toggle AI features.

Ticket AI Auto-Handle

When a member sends a message in a support ticket, Lunova AI reads the conversation and sends a suggested reply — visible to staff, or optionally sent directly to the member. This reduces time-to-first-response for common questions.

Powered by the local Ollama qwen2.5:7b model.
Fails gracefully if the local Ollama service is unavailable.
Enable per-server in Dashboard → Tickets → AI ticket support.
Works independently from the global AI toggle — a server can have ticket AI on without enabling general AI commands.

✦ Ticket AI works best when your system prompt is customized. Set server-specific AI instructions in Dashboard → AI Assistant to reflect your rules and tone.

Commands

/ask

/ask question: How do I appeal a ban?

Ask Lunova AI any question. The response is ephemeral (only the user sees it) and uses the server's general AI system prompt if one is configured.

Parameter	Required	Description
question	required	The question or prompt to send to the AI.

/ai

/ai prompt: Summarise our server rules in bullet points

A more general AI prompt command. Useful for staff to quickly generate text, summaries, or draft announcements. Response is ephemeral.

Parameter	Required	Description
prompt	required	The message to send to the AI.

/summarize

/summarize [channel:] [limit:]

Fetches recent messages from a channel and returns an AI-written summary. Useful for catching up on a busy channel or logging key decisions from a discussion.

Parameter	Required	Description
channel	optional	Channel to summarize. Defaults to current channel.
limit	optional	Number of messages to fetch (default 30, max 100).

AI Moderation

Lunova can optionally route suspicious messages through an AI content check before taking automod action. This is separate from the rule-based AutoMod filters and is intended for edge cases that pattern matching misses.

Analyzes message content for toxicity, hate speech, scams, and rule violations.
Returns a structured decision: none, warn, mute, or delete.
Configure in Dashboard → AutoMod → AI content check.

⚠ AI moderation should be used alongside — not instead of — rule-based AutoMod filters. AI responses can occasionally be wrong; always review mod logs for AI-triggered actions.

Custom System Prompts

Each AI feature can be given a custom system prompt per server, letting you tailor the AI's personality and knowledge to your community.

Prompt type	Affects	Where to set
General	/ask, /ai	Dashboard → General → AI system prompt (general)
Summarize	/summarize	Dashboard → General → AI system prompt (summarize)
Moderation	AI content check	Dashboard → General → AI system prompt (moderation)

Rate limits & caching

AI calls are rate-limited to 8 per minute per server to protect local CPU/RAM headroom.
Identical prompts are cached for 5 minutes — repeated questions return quickly without another generation.
Ticket AI uses the same local Ollama service and bounded conversation history.

Providers

Feature	Provider	Model
Ticket auto-handle	Ollama	qwen2.5:7b
/ask, /ai, /summarize	Ollama	qwen2.5:7b
AI moderation	Ollama	qwen2.5:7b

Configure the local service with OLLAMA_HOST and OLLAMA_MODEL. Defaults are http://localhost:11434 and qwen2.5:7b.