Why
Without visible cost, users cannot tell whether a preset is actually working or whether a turn over-spent. We will get bug reports that are really preset complaints. Closing the feedback loop is what makes the cost-tier system trustworthy.
Depends on
- Per-role model routing foundation
Scope
- Per-turn footer: model used, role, input/output tokens, $ estimate
- Channel header: rolling session cost
- Settings → Cost tab: per-day / per-week breakdown by role and model
- Cost data sourced from existing token accounting (do not re-invent metering)
Out of scope
- Hard budget caps / kill switches (track separately if demand emerges)
Why
Without visible cost, users cannot tell whether a preset is actually working or whether a turn over-spent. We will get bug reports that are really preset complaints. Closing the feedback loop is what makes the cost-tier system trustworthy.
Depends on
Scope
Out of scope