Integrate transcription checkpoints across lifecycle

internal/stage/merge.go

@@ -306,6 +306,12 @@ func normalizeMergeInputs(
 }
 
 func discoverRawTranscripts(m *manifest.Manifest, paths artifacts.SessionPaths) ([]string, error) {
+	if record, versioned, err := transcribeManifestState(m); err != nil {
+		return nil, err
+	} else if versioned {
+		return discoverVersionedRawTranscripts(record, paths)
+	}
+
 	fromManifest := make([]string, 0)
 	if m != nil && m.Stages != nil {
 		if tr := m.Stages["transcribe"]; tr != nil {
@@ -350,6 +356,64 @@ func discoverRawTranscripts(m *manifest.Manifest, paths artifacts.SessionPaths)
 	return dedupeAndSortPaths(out), nil
 }
 
+func discoverVersionedRawTranscripts(
+	record *manifest.StageRecord,
+	paths artifacts.SessionPaths,
+) ([]string, error) {
+	if record.Status != manifest.StatusSucceeded {
+		return nil, fmt.Errorf("versioned transcribe state is not aggregate-successful (status %q)", record.Status)
+	}
+	if err := manifest.ValidateTranscribeSpeakerCollection(
+		record.TranscribeStateVersion,
+		record.TranscribeSpeakers,
+	); err != nil {
+		return nil, fmt.Errorf("validate versioned transcribe state: %w", err)
+	}
+	if len(record.TranscribeSpeakers) == 0 {
+		return nil, fmt.Errorf("versioned transcribe state has no completed speakers")
+	}
+
+	expected := make(map[string]string, len(record.TranscribeSpeakers))
+	for speaker, speakerRecord := range record.TranscribeSpeakers {
+		path := filepath.Clean(artifacts.ResolveSessionLocalPathForRead(paths, speakerRecord.Output.LocalPath))
+		if other, exists := expected[path]; exists {
+			return nil, fmt.Errorf("versioned transcribe speakers %q and %q declare the same output", other, speaker)
+		}
+		expected[path] = speaker
+	}
+	inputs := make([]string, 0, len(expected))
+	for _, output := range record.Outputs {
+		if output.Kind != "transcript_raw" {
+			continue
+		}
+		path := strings.TrimSpace(output.LocalPath)
+		if path == "" {
+			return nil, fmt.Errorf("versioned transcribe aggregate output has an empty local path")
+		}
+		path = filepath.Clean(artifacts.ResolveSessionLocalPathForRead(paths, path))
+		speaker, declared := expected[path]
+		if !declared {
+			return nil, fmt.Errorf("versioned transcribe aggregate output %q is not declared by current speaker state", output.LocalPath)
+		}
+		if speaker == "" {
+			return nil, fmt.Errorf("versioned transcribe aggregate output %q is duplicated", output.LocalPath)
+		}
+		expected[path] = ""
+		inputs = append(inputs, path)
+	}
+	missing := make([]string, 0)
+	for _, speaker := range expected {
+		if speaker != "" {
+			missing = append(missing, speaker)
+		}
+	}
+	if len(missing) > 0 {
+		sort.Strings(missing)
+		return nil, fmt.Errorf("versioned transcribe aggregate outputs are incomplete for speakers: %s", strings.Join(missing, ", "))
+	}
+	return dedupeAndSortPaths(inputs), nil
+}
+
 func dedupeAndSortPaths(paths []string) []string {
 	if len(paths) == 0 {
 		return nil