Implemented an initial transcript merge stage

internal/builtin/preprocess.go

@@ -3,11 +3,13 @@ package builtin
 import (
 	"context"
 	"fmt"
+	"strings"
 
 	"gitea.maximumdirect.net/eric/seriatim/internal/config"
 	"gitea.maximumdirect.net/eric/seriatim/internal/model"
 	"gitea.maximumdirect.net/eric/seriatim/internal/pipeline"
 	"gitea.maximumdirect.net/eric/seriatim/internal/report"
+	"gitea.maximumdirect.net/eric/seriatim/internal/speaker"
 )
 
 type noopPreprocessor struct {
@@ -42,6 +44,39 @@ func (p noopPreprocessor) Process(ctx context.Context, in pipeline.PreprocessSta
 	}, nil
 }
 
+type trimText struct{}
+
+func (trimText) Name() string {
+	return "trim-text"
+}
+
+func (trimText) Requires() pipeline.ModelState {
+	return pipeline.StateCanonical
+}
+
+func (trimText) Produces() pipeline.ModelState {
+	return pipeline.StateCanonical
+}
+
+func (trimText) Process(ctx context.Context, in pipeline.PreprocessState, cfg config.Config) (pipeline.PreprocessState, []report.Event, error) {
+	if err := ctx.Err(); err != nil {
+		return pipeline.PreprocessState{}, nil, err
+	}
+	if in.State != pipeline.StateCanonical {
+		return pipeline.PreprocessState{}, nil, fmt.Errorf("preprocessing module %q requires state %q but received %q", "trim-text", pipeline.StateCanonical, in.State)
+	}
+
+	for transcriptIndex := range in.Canonical {
+		for segmentIndex := range in.Canonical[transcriptIndex].Segments {
+			in.Canonical[transcriptIndex].Segments[segmentIndex].Text = strings.TrimSpace(in.Canonical[transcriptIndex].Segments[segmentIndex].Text)
+		}
+	}
+
+	return in, []report.Event{
+		report.Info("preprocessing", "trim-text", "trimmed canonical segment text"),
+	}, nil
+}
+
 type normalizeSpeakers struct{}
 
 func (normalizeSpeakers) Name() string {
@@ -64,11 +99,33 @@ func (normalizeSpeakers) Process(ctx context.Context, in pipeline.PreprocessStat
 		return pipeline.PreprocessState{}, nil, fmt.Errorf("preprocessing module %q requires state %q but received %q", "normalize-speakers", pipeline.StateRaw, in.State)
 	}
 
+	speakers, err := speaker.LoadMap(cfg.SpeakersFile)
+	if err != nil {
+		return pipeline.PreprocessState{}, nil, err
+	}
+
 	canonical := make([]model.CanonicalTranscript, 0, len(in.Raw))
 	for _, raw := range in.Raw {
+		canonicalSpeaker, err := speakers.SpeakerForSource(raw.Source)
+		if err != nil {
+			return pipeline.PreprocessState{}, nil, err
+		}
+
+		segments := make([]model.Segment, 0, len(raw.Segments))
+		for index, rawSegment := range raw.Segments {
+			segments = append(segments, model.Segment{
+				Source:             raw.Source,
+				SourceSegmentIndex: index,
+				Speaker:            canonicalSpeaker,
+				Start:              rawSegment.Start,
+				End:                rawSegment.End,
+				Text:               rawSegment.Text,
+			})
+		}
+
 		canonical = append(canonical, model.CanonicalTranscript{
 			Source:   raw.Source,
-			Segments: nil,
+			Segments: segments,
 		})
 	}
 
@@ -77,6 +134,6 @@ func (normalizeSpeakers) Process(ctx context.Context, in pipeline.PreprocessStat
 			Raw:       append([]model.RawTranscript(nil), in.Raw...),
 			Canonical: canonical,
 		}, []report.Event{
-			report.Info("preprocessing", "normalize-speakers", "created placeholder canonical transcript(s)"),
+			report.Info("preprocessing", "normalize-speakers", "created canonical transcript(s) from raw input"),
 		}, nil
 }