AI

AI Overview

exceed's AI features at a glance, all powered by Google Gemini with a shared per-server usage budget.

AI in exceed

exceed ships a suite of AI features built on Google Gemini. They cover the boring-but-useful jobs that eat up your time: catching people up on a busy channel, drafting embeds, planning a fresh server, helping staff write fair mod reasons, and pointing members at the right command.

Every AI feature runs through one place internally, which means two things for you:

  • One on/off switch and one budget. AI is governed per server from the AI Usage dashboard page. Turn it off there and every AI command stops immediately.
  • A shared monthly allowance with per-user cooldowns. Because Gemini's free tier has hard daily request caps shared across the whole bot, each server gets its own monthly request and token allowance so one busy server can't starve everyone else. See Usage & Limits for the full explanation.

Note: AI features require the bot owner to have configured a Gemini API key. If AI is not configured on your instance of the bot, AI commands reply that "AI features are not configured" and the dashboard pages say the same.

What you get

FeatureWhat it doesWhere to enable
Setup AssistantDescribe a server, get a roles/channels/modules plan, then apply itAI Setup Assistant page
Summaries & Recaps,summarize, ,recap, ,tldr/,eli5, ,ask for catching up fastAI Summaries (for ,summarize)
Persona ChatA named AI personality that chats with memory in one channelAI Persona Chat page
Embed Generator,aiembed — describe an embed and AI builds itOn by default; tune budget on AI Usage
Mod Copilot,copilot — staff case briefings and polished mod reasons (advisory only),copilot enable
Smart Help,findcommand / ,howdo — plain English to the real commandSmart help toggle
Image AutoModAI checks posted images against your rules and acts (image mute, delete)Image AutoMod page

How the AI works

  • Model: Google Gemini, defaulting to gemini-2.5-flash with automatic fallback to gemini-2.5-flash-lite and gemini-flash-lite-latest if the primary model is rate-limited or unavailable.
  • Grounded, not open-ended where it matters. Smart Help can only return real commands from exceed's catalog, the ticket knowledge-base assistant answers only from your own entries, and Mod Copilot works only from the actual case history. The AI is constrained so it can't invent commands, policies, or incidents.
  • Safe output. Generated embeds and messages have @everyone, @here, and role mentions stripped before they post, so AI output can never mass-ping your server.

Next steps