fix(run): address gadfly review of the phases PR
executus CI / test (pull_request) Successful in 48s

Real findings from the consensus review (37 raw; many devstral dups/noise):

- Optional/budget-salvage branches no longer swallow a context
  cancellation / deadline / critic-kill: such errors return immediately so
  the run is classified cancelled/timeout/killed, not "ok" with a fallback.
  (the most serious finding — an Optional final phase could mask a killed run)
- IsRunFunc bare phase now feeds the SHARED step observer (not just the
  audit recorder), so the critic's activity clock + Result.Steps see it —
  a long synthesize phase no longer looks idle to the critic.
- phaseModel returns the resolver's enriched (usage-attribution) context and
  the phase's calls use it, mirroring the single-loop path (non-base-tier
  phases were mis-attributed).
- salvagePhaseTranscript trims the tail on a rune boundary (was a raw byte
  slice that could split a UTF-8 rune); maxSalvage is now a named const with
  rationale.
- expandPhaseTemplate logs a WARN on parse/execute failure instead of
  silently returning the unexpanded template; documented the phase-name
  identifier requirement + the "Query" shadow.
- removed the dead phaseDeps.baseTier field.
- extracted multimodalUserMessage, shared by runAgent + the phase runner
  (was duplicated image-folding).
- aggregated phase usage is stamped onto the result even on a hard-error
  return; TrimSpace computed once; filterToolbox returns the base toolbox
  as-is for the empty-names (full-palette) case instead of copying;
  phaseModel WARN no longer prints error=<nil>.

New test: Optional phase does not swallow a cancellation. Full ./... green.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-06-29 15:44:04 -04:00
parent 30b79a330f
commit 0dd2ced717
3 changed files with 151 additions and 87 deletions
+25
View File
@@ -105,6 +105,31 @@ func TestPhases_OptionalFailureSubstitutesFallback(t *testing.T) {
}
}
// TestPhases_OptionalDoesNotSwallowCancellation: an Optional phase that fails
// with a context cancellation must NOT be swallowed into its FallbackMessage —
// the run genuinely ended (cancel/deadline/critic-kill) and must surface the
// error so the run is classified cancelled/timeout/killed, not "ok".
func TestPhases_OptionalDoesNotSwallowCancellation(t *testing.T) {
models, _ := phaseProvider(t, fake.Fail(context.Canceled))
ex := New(Config{Registry: tool.NewRegistry(), Models: models})
ra := RunnableAgent{
Name: "pipeline",
ModelTier: "test-model",
Phases: []Phase{
// IsRunFunc so the cancellation surfaces directly wrapped (%w).
{Name: "a", SystemPrompt: "Phase A", IsRunFunc: true, Optional: true, FallbackMessage: "FB"},
},
}
res := ex.Run(context.Background(), ra, tool.Invocation{RunID: "r", CallerID: "c"}, "Q")
if !errors.Is(res.Err, context.Canceled) {
t.Fatalf("Optional phase must NOT swallow a cancellation; res.Err = %v", res.Err)
}
if res.Output == "FB" {
t.Error("a cancelled run must not report the fallback message as output")
}
}
// TestPhases_HardErrorAborts: a NON-optional phase that hits a hard error (not a
// budget/step exhaustion) aborts the pipeline; later phases do not run.
func TestPhases_HardErrorAborts(t *testing.T) {