# AI Usage Quotas

## Objective

Limit how much AI work a plan or workspace can do, and show what is left.

Countable allowances for AI messages, runs, documents, or tokens, enforced per plan and workspace with a defined reset period.

## Before You Begin

This feature is being added to an application that already exists and already
works. Do not scaffold a new project, and do not assume a blank slate.

Inspect the codebase first and establish:

- The existing application structure and where code of this kind already lives.
- The framework and version in use.
- The existing design system — colours, spacing, typography, and component conventions.
- Existing UI components you can reuse instead of writing new ones.
- The existing database structure, if this feature needs to persist anything.
- The existing authentication and authorization system, if this feature is user-scoped.
- Dependencies already installed, so you don't add a library that duplicates one.
- The existing test setup and conventions.

Only start writing code once you understand the above. If the application
already implements part of this feature, extend it rather than replacing it.

## Implementation Instructions

1. Identify the units the product actually sells on: messages, runs, processed documents, or tokens. Pick as few as possible and define each one precisely before writing any enforcement.
2. Enforce every quota on the server at the point the work is dispatched. Client-side counters and hidden buttons are presentation, not enforcement.
3. Decide and document whether a run that failed, was cancelled, or returned a refusal consumes allowance, and apply that decision consistently across every AI surface.
4. Support a shared workspace pool with optional per-member caps, so one member cannot drain the whole allowance in an afternoon.
5. AI Cost Budgets owns money and provider spend. This entry owns countable units and plan entitlements. Reuse the same dispatch interception point but do not derive one limit from the other.

## UI and UX Requirements

Match the application's existing design system exactly. Reuse its components,
spacing, and typography. This feature should look like it was always there.

## Responsive Requirements

Works on mobile, tablet, and desktop. Touch targets are large enough to hit on a
phone, and nothing overflows horizontally at 320px.

## Accessibility Requirements

- Fully keyboard navigable.
- Correct semantic elements and ARIA roles.
- Visible focus states.
- Meets WCAG AA contrast.
- Dynamic changes are announced to screen readers.
- Respects prefers-reduced-motion.

## Edge Cases

- Any check performed only in the interface can be bypassed by calling the endpoint directly. The dispatch path must re-check the quota server-side every time, including for background and scheduled runs.
- Failed, cancelled, and refused runs need an explicit rule. Charging for a run the model refused feels like theft; charging for nothing at all invites retry loops as a way around the limit.
- A shared workspace pool without a per-member cap lets one person exhaust the month for everyone. Support both, and make the interaction between them explicit in the interface.
- Reset periods must be anchored to a stated time zone and to the billing anniversary rather than a server-local midnight, or usage will reset at a time that makes no sense to the customer.
- Token counts are estimates until the provider reports actuals. Show remaining usage in the honest unit and never present an estimate with a precision it does not have.
- Concurrent requests near the limit must not all be admitted. Decrement atomically at dispatch rather than reading a count and writing it back.
- A plan downgrade mid-period must not produce a negative balance or retroactively invalidate work already done.
- When a quota is exhausted, the user needs to know which unit ran out, when it resets, and who in the workspace can raise it. A generic refusal produces a support ticket.

## Testing

Exercise the feature end to end in the running application. Cover every edge case
above, then run the existing test suite and confirm nothing regressed.

## Acceptance Criteria

- [ ] Quotas are enforced server-side on every dispatch path, including background and scheduled runs.
- [ ] The treatment of failed, cancelled, and refused runs is defined and applied uniformly.
- [ ] Workspace pools and per-member caps both exist and their interaction is visible to the user.
- [ ] Reset periods use a stated time zone and align to the billing anniversary.
- [ ] Remaining usage is displayed in honest units with no false precision on estimates.
- [ ] Concurrent requests at the limit cannot exceed the allowance.
- [ ] An exhausted quota tells the user what ran out, when it resets, and who can raise it.
- [ ] The feature matches the existing design system.
- [ ] No existing functionality is broken.

## Adaptation Rules

- Match the existing design system. Do not introduce a new colour palette,
  spacing scale, or component library.
- Reuse existing components and utilities wherever they fit.
- Follow the naming, file layout, and code style already present.
- Do not upgrade, replace, or remove existing dependencies to make this
  feature fit. Adapt the feature to the app, not the app to the feature.
- Do not break existing functionality. If a change is genuinely required in
  existing code, make the smallest one that works and say so.
- If something in these instructions conflicts with how the application is
  built, follow the application and explain the deviation.

## Final Verification

Before you report the work as done:

1. Re-read the acceptance criteria above and check each one against what you
   actually built.
2. Run the application and exercise the feature end to end.
3. Run the existing test suite and confirm you have broken nothing.
4. Check the feature on mobile, tablet, and desktop widths.
5. Check keyboard navigation and focus handling.
6. Summarize what changed: files added, files modified, and anything you
   deliberately did differently because of how this application is built.

If any acceptance criterion is unmet, fix it before reporting completion.