0.8.3 — Analyses see the whole picture

Released 2026-08-01.

What's new

Mission and ancestor context in analysis (FR-21.7). Every requirement analysis now includes the application's mission statement and the full chain of parent requirements above the requirement being analysed. Analysers can therefore resolve upwards for information defined higher in the tree instead of treating it as absent.

The practical effect: fewer false "missing" findings. Where a business goal, actor, or constraint is stated in a parent requirement or in the mission, a child requirement is no longer reported as lacking it. This applies to story analyses too, walking up from the story's linked requirement.

The structural context block is assembled once per analysis job and cached across the whole analyser sweep, so a run of several analysers over the same requirement pays for it once.
Analyses are substantially cheaper. Two changes to how the glossary is supplied: the nine analysers that never reference the glossary no longer receive it at all, and the two that do — undefined-term detection and feature generation — now receive only the terms that actually appear in the requirement and its ancestors, rather than the whole organisation glossary. Measured on a live ten-step sweep, analysis input dropped by roughly 85%.
Note this is a cost improvement rather than a speed one. Analysis wall-clock time is governed by how much the model writes, not how much it reads, so a large reduction in input has only a modest effect on how long a run takes.
The glossary term index is now refreshed automatically before it is used for undefined-term detection, so a recently added definition is not reported as undefined.

None. No schema change and no configuration change.