AI quota

Every AI command spends from a shared per-user budget called points. The budget refills automatically on a 30-minute tumbling window — fair, predictable, and impossible to burn through accidentally.

How points work

The window starts the moment a user runs their first AI command in a server. From there:

• Each call deducts that command's cost from the user's bucket.
• When the bucket would go negative, the call is refused with a friendly “come back in N minutes” message.
• The full bucket refills exactly 30 minutes after the first call — not gradually. (Tumbling, not sliding.)

The bot owner is exempt by default. Additional owner IDs can be set via the BOT_ADDITIONAL_OWNER_IDS environment variable.

Per-command costs

Costs are fully configurable per-server via /ai-limits (owner only). The defaults are tuned so a typical user can run a mixed handful of commands per window:

Command          Cost  Notes
─────────────────────────────────
/ask             1     short Q&A, multi-turn
/think           3     deep reasoning, Gemini Pro
/summarise       2     channel summaries
/fallacy         2     message-history scan
/modadvice       3     full channel context
/imagine         4     image generation
/edit            4     image editing

Default budget per user, per 30 min: 12 points.

Boosts and bonuses

There are two ways to extend a user's budget beyond the default:

• Permanent boosts.Add a flat multiplier to a specific user's bucket. Useful for staff or boosters.
• One-time bonus pools.Hand out extra points that sit alongside the regular bucket and don't reset on the 30-minute timer. They drain first.

Tip. Bonus pools are great for events — top-up every Nitro booster with 50 extra points for the weekend, then forget about it. The pool drains naturally and nothing rolls over into the regular budget.