# AI Forecast Narrative

## Objective

Turn the numbers a forecast already produced into a plain explanation of the outlook.

A written interpretation of an existing forecast, covering the horizon, the confidence range, and the assumptions behind it.

## 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. Take the forecast values, the confidence interval, and the horizon from the app's existing forecasting calculation and pass them to the model as fixed inputs. The model interprets numbers; it never produces or adjusts them.
2. Mark historical values and projected values distinctly in what you send, and require the narrative to keep them distinct in the prose. A sentence that blends last quarter's actuals with next quarter's projection is misinformation.
3. State the horizon, the confidence range, and the assumptions the forecast rests on in the narrative itself, in the metric's own units and the workspace's time zone.
4. Invalidate the narrative whenever the forecast inputs change — new actuals, a changed model, a corrected historical series — and regenerate rather than leaving prose that no longer matches the chart above it.
5. Describing past movement is owned by AI Trend Analysis and explaining a specific outlier is owned by AI Anomaly Explanation. This feature covers the projected portion only; do not let it re-narrate the history the sibling features already describe.

## 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 number in the narrative that did not come from the forecasting calculation is a defect. Verify the figures in the generated text against the source values and discard the narrative on a mismatch.
- The horizon, the confidence interval, and the assumptions must be named. A projection stated without its uncertainty reads as a promise and will be quoted as one.
- Projected values must never be described in the past tense or alongside actuals without a label. Enforce the distinction in the output format, not only in the instruction.
- A narrative generated from a superseded forecast must be marked stale or removed. Tie it to the forecast run that produced it and drop it when a new run lands.
- A wide confidence interval or a short history means the forecast is weak. Suppress the narrative or lead with the caveat rather than writing a confident paragraph about a guess.
- If the model refuses, times out, or returns a truncated paragraph, show the forecast chart and its interval alone. The numeric forecast is the product; the prose is a convenience.
- Forecasts often cover revenue or headcount. Decide what may leave the app, and do not send customer names, account identifiers, or other detail the narrative does not require.
- Narratives should be generated once per forecast run and stored, with a ceiling on regeneration, so a dashboard refresh does not trigger a new call each time.

## 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

- [ ] Every figure in the narrative traces back to the forecasting calculation and is verified against it.
- [ ] Historical and projected values are labelled distinctly in the output.
- [ ] The horizon, confidence range, and assumptions are stated in every narrative.
- [ ] A narrative is invalidated and regenerated when its forecast inputs change.
- [ ] Low-confidence forecasts are either suppressed or carry an explicit caveat rather than reading as commitments.
- [ ] The forecast chart and its interval remain available when the model is unavailable.
- [ ] Narratives are stored per forecast run and are not regenerated on each page view.
- [ ] 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.