Configure PromptKit transient retries

internal/cli/catalog.go

@@ -157,13 +157,15 @@ func buildProductionLLMClient(ctx context.Context, cfg config.Config, profileID
 		return nil, nil, fmt.Errorf("production asset registry must not be nil")
 	}
 	recorder := llm.NewLLMProfileRecorder()
+	transientRetries := cfg.PromptKit.TransientRetries
 	client, err := llm.NewPromptKitClient(llm.PromptKitClientConfig{
-		ProfileDir:      cfg.PromptKit.ProfileDir,
-		ProfileFile:     cfg.PromptKit.ProfileFile,
-		LocalBackend:    mapPromptKitLocalBackend(cfg.PromptKit.LocalBackend),
-		Assets:          assets,
-		Recorder:        recorder,
-		ReasoningEffort: overrides.ReasoningEffort,
+		ProfileDir:       cfg.PromptKit.ProfileDir,
+		ProfileFile:      cfg.PromptKit.ProfileFile,
+		TransientRetries: &transientRetries,
+		LocalBackend:     mapPromptKitLocalBackend(cfg.PromptKit.LocalBackend),
+		Assets:           assets,
+		Recorder:         recorder,
+		ReasoningEffort:  overrides.ReasoningEffort,
 	})
 	if err != nil {
 		return nil, nil, fmt.Errorf("create PromptKit-backed LLM client: %w", err)

internal/cli/production_contract_test.go

@@ -557,11 +557,16 @@ func TestNormalizeOptionsSharesProductionProfileAssetsWithDefaultRuntime(t *test
 func TestProductionLLMClientFactoryUsesConfiguredLocalBackend(t *testing.T) {
 	var providerCalls atomic.Int32
 	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
-		providerCalls.Add(1)
+		call := providerCalls.Add(1)
 		if r.URL.Path != "/v1/chat/completions" {
 			t.Errorf("provider path = %q, want /v1/chat/completions", r.URL.Path)
 		}
 		w.Header().Set("Content-Type", "application/json")
+		if call == 1 {
+			w.WriteHeader(http.StatusServiceUnavailable)
+			_, _ = w.Write([]byte(`{"error":{"message":"transient"}}`))
+			return
+		}
 		_, _ = w.Write([]byte(`{
 			"choices": [{"message": {"role": "assistant", "content": "{\"ok\":true}"}}],
 			"usage": {"prompt_tokens": 3, "completion_tokens": 4, "total_tokens": 7}
@@ -596,6 +601,7 @@ output:
 
 	cfg := config.Default()
 	cfg.PromptKit.ProfileFile = profilePath
+	cfg.PromptKit.TransientRetries = 1
 	cfg.PromptKit.LocalBackend = &config.PromptKitLocalBackendConfig{
 		Endpoint:         server.URL + "/v1",
 		ConcurrencyLimit: 2,
@@ -624,8 +630,8 @@ output:
 	if err != nil {
 		t.Fatalf("CompleteStructured() error = %v, want nil", err)
 	}
-	if providerCalls.Load() != 1 {
-		t.Fatalf("provider calls = %d, want 1", providerCalls.Load())
+	if providerCalls.Load() != 2 {
+		t.Fatalf("provider calls = %d, want 2", providerCalls.Load())
 	}
 	provider, ok := client.(contracts.LLMProfileManifestProvider)
 	if !ok {