Skip to content

loop

diapason.learning.spec_search.execute.loop

Per-edit execution loop for the spec-search execute phase.

Iterates over a plan's edits, handles tier routing, validates, and applies. Does NOT include the benchmark gate — that's wired in M5.

See spec §7.2.

execute_edits(*, edits, ctx, autonomy_mode, registry=None)

Execute a list of edits, returning outcomes for each.

This loop handles tier routing and validation but does NOT include the benchmark gate (wired in M5).

Parameters

edits : The edits to process. ctx : Shared context with config paths. autonomy_mode : How aggressively to apply edits. registry : Optional pre-built registry. If None, builds the default.