Configure built-in retries across execution paths

internal/usecase/runner.go

@@ -43,15 +43,16 @@ var (
 
 // Runner executes the Promptkit core use case.
 type Runner struct {
-	promptDefs promptdef.Repository
-	profiles   profile.Repository
-	backends   BackendResolver
-	artifacts  artifact.Reader
-	renderer   prompt.Renderer
-	llm        llm.Client
-	validator  validate.Validator
-	repairer   OutputRepairer
-	admitter   RunAdmitter
+	promptDefs              promptdef.Repository
+	profiles                profile.Repository
+	backends                BackendResolver
+	artifacts               artifact.Reader
+	renderer                prompt.Renderer
+	llm                     llm.Client
+	validator               validate.Validator
+	repairer                OutputRepairer
+	admitter                RunAdmitter
+	defaultTransientRetries uint64
 }
 
 // BackendResolver resolves one normalized backend ID.
@@ -73,6 +74,7 @@ type preparationState struct {
 	targetPresence       domain.ExecutionTargetPresence
 	effectiveContract    domain.OutputContract
 	start                time.Time
+	transientRetries     uint64
 }
 
 type preparedOperation struct {
@@ -113,17 +115,23 @@ func NewRunnerWithRepairer(
 	validator validate.Validator,
 	repairer OutputRepairer,
 	admitter RunAdmitter,
+	defaultTransientRetries ...uint64,
 ) *Runner {
+	retries := uint64(0)
+	if len(defaultTransientRetries) > 0 {
+		retries = defaultTransientRetries[0]
+	}
 	return &Runner{
-		promptDefs: promptDefs,
-		profiles:   profiles,
-		backends:   backends,
-		artifacts:  artifacts,
-		renderer:   renderer,
-		llm:        llmClient,
-		validator:  validator,
-		repairer:   repairer,
-		admitter:   admitter,
+		promptDefs:              promptDefs,
+		profiles:                profiles,
+		backends:                backends,
+		artifacts:               artifacts,
+		renderer:                renderer,
+		llm:                     llmClient,
+		validator:               validator,
+		repairer:                repairer,
+		admitter:                admitter,
+		defaultTransientRetries: retries,
 	}
 }
 
@@ -152,7 +160,7 @@ func (r *Runner) Run(ctx context.Context, req domain.RunRequest) (*domain.RunRes
 	}
 	directAPIKey := state.effectiveModel.APIKey
 
-	return r.executePreparedRun(ctx, operation.run, directAPIKey, runID, start, func(
+	return r.executePreparedRun(ctx, operation.run, directAPIKey, state.transientRetries, runID, start, func(
 		ctx context.Context,
 		artifact *domain.Artifact,
 		attemptsUsed int,
@@ -173,6 +181,7 @@ func (r *Runner) executePreparedRun(
 	ctx context.Context,
 	prepared *domain.PreparedRun,
 	directAPIKey string,
+	transientRetries uint64,
 	runID string,
 	start time.Time,
 	validateArtifact preparedValidationFunc,
@@ -185,6 +194,7 @@ func (r *Runner) executePreparedRun(
 		executionTarget,
 		prepared.TargetPresence,
 		prepared.StructuredOutput,
+		transientRetries,
 	))
 	if err != nil {
 		return nil, wrapGenerationError(err)
@@ -213,6 +223,7 @@ func (r *Runner) executePreparedRun(
 				Target:           executionTarget,
 				TargetPresence:   prepared.TargetPresence,
 				StructuredOutput: prepared.StructuredOutput,
+				TransientRetries: transientRetries,
 				Attempt:          attemptsUsed,
 				MaxAttempts:      prepared.OutputContract.RepairAttempts,
 				Mode:             prepared.OutputContract.ValidationMode,
@@ -295,6 +306,10 @@ func (r *Runner) resolvePreparation(
 	req domain.RunRequest,
 	start time.Time,
 ) (*preparationState, error) {
+	transientRetries, err := domain.ResolveTransientRetries(r.defaultTransientRetries, req.TransientRetries)
+	if err != nil {
+		return nil, fmt.Errorf("%w: %w", ErrInvalidRequest, err)
+	}
 	if strings.TrimSpace(req.PromptID) == "" {
 		return nil, fmt.Errorf("%w: prompt id is required", ErrInvalidRequest)
 	}
@@ -346,6 +361,7 @@ func (r *Runner) resolvePreparation(
 		targetPresence:       targetPresence,
 		effectiveContract:    effectiveContract,
 		start:                start,
+		transientRetries:     transientRetries,
 	}, nil
 }