Protecting your fixes
Hand-made changes — a surgical Fix, an edited entity field, a moved capability — are stored as authored elements in your model. Understanding where they live tells you when they're safe and when they're at risk.
Where authored changes live
Authored changes are part of model.json — the complete, versionable document holding every layer of
your business. That's the durable source of truth for export and code generation. They survive a
re-review, and they're included when you Export model or generate code.
They are not written into the Business Narrative, and the modeling flow is one-way, top-down (narrative → capabilities → … → agents). So a fresh top-down pass doesn't know about them.
When a regeneration would discard them
Regenerating a layer replaces it wholesale. That means:
- Generate on a layer that holds hand-made fixes replaces them;
- editing the narrative resets every downstream layer;
- Generate Capabilities resets everything below it.
The regeneration guard
To stop this happening silently, Kiln warns first. If a Generate (or a narrative edit) would discard authored fixes, you get:
This layer has N hand-made fix(es). Regenerating replaces the whole layer and discards them. They live in
model.json(⬇ Export model) — the narrative doesn't hold them, so a top-down regenerate loses them. Continue?
Cancel and nothing is lost. If you do want to regenerate, Export model first to keep a copy.
Keeping the narrative honest
Because fixes don't flow back into the narrative, the prose can fall behind the model. Sync narrative closes that gap — see Narrative sync.
Resetting the model
There's no single "reset" button, but you can start over in a few ways: edit the narrative (clears every derived layer), Import model (replace wholesale), delete the project and start a new one, or Versions → restore an earlier snapshot. The regeneration guard applies to the narrative-edit route, so you won't wipe fixes by accident.