feat: generate transcript and event-driven particle processing

This generates the transcript and shows the caption experience on the
client side for media particles. It also simplifies other side effects
that we must perform such as updating the `last_child_created_at` field
for stream and container particles.

go/internal/particle/firestore_types.go

@@ -0,0 +1,53 @@
+package particle
+
+import "time"
+
+type FirestoreMediaParticle struct {
+	CreatedByHumanId string                           `firestore:"created_by_human_id"`
+	Type             string                           `firestore:"type"`
+	Properties       FirestoreMediaParticleProperties `firestore:"properties"`
+	CreatedAt        time.Time                        `firestore:"created_at,serverTimestamp"`
+	UpdatedAt        *time.Time                       `firestore:"updated_at,omitempty"`
+}
+
+type FirestoreTranscriptWord struct {
+	Word  string  `firestore:"word"`
+	Start float64 `firestore:"start"`
+	End   float64 `firestore:"end"`
+}
+
+type FirestoreTranscriptSentence struct {
+	Text  string  `firestore:"text"`
+	Start float64 `firestore:"start"`
+	End   float64 `firestore:"end"`
+}
+
+type FirestoreTranscriptParagraph struct {
+	Sentences []FirestoreTranscriptSentence `firestore:"sentences"`
+	Start     float64                       `firestore:"start"`
+	End       float64                       `firestore:"end"`
+}
+
+type FirestoreTranscript struct {
+	Transcript string                         `firestore:"transcript"`
+	Words      []FirestoreTranscriptWord      `firestore:"words"`
+	Paragraphs []FirestoreTranscriptParagraph `firestore:"paragraphs"`
+}
+
+type FirestoreMediaParticleProperties struct {
+	ObjectId   string               `firestore:"object_id"`
+	MimeType   string               `firestore:"mime_type"`
+	DurationMs int                  `firestore:"duration_ms"`
+	SizeBytes  int                  `firestore:"size_bytes"`
+	Transcript *FirestoreTranscript `firestore:"transcript,omitempty"`
+}
+
+type FirestoreStreamParticle struct {
+	CreatedByHumanId string `firestore:"created_by_human_id"`
+	Type             string `firestore:"type"`
+	// Properties       FirestoreStreamParticleProperties `firestore:"properties"`
+	CreatedAt          time.Time  `firestore:"created_at,serverTimestamp"`
+	LastChildCreatedAt *time.Time `firestore:"last_child_created_at,omitempty"`
+	VisibleTo          []string   `firestore:"visible_to"`
+	UpdatedAt          *time.Time `firestore:"updated_at,omitempty"`
+}

go/internal/particle/processing_repository.go

@@ -0,0 +1,37 @@
+package particle
+
+import (
+	"context"
+
+	"github.com/jackc/pgx/v5/pgxpool"
+)
+
+type ProcessingRepository interface {
+	IsProcessed(ctx context.Context, particleID string) (bool, error)
+	MarkProcessed(ctx context.Context, particleID string) error
+}
+
+type processingRepositoryImpl struct {
+	pool *pgxpool.Pool
+}
+
+func NewProcessingRepository(pool *pgxpool.Pool) ProcessingRepository {
+	return &processingRepositoryImpl{pool: pool}
+}
+
+func (r *processingRepositoryImpl) IsProcessed(ctx context.Context, particleID string) (bool, error) {
+	var exists bool
+	err := r.pool.QueryRow(ctx,
+		`SELECT EXISTS(SELECT 1 FROM processed_particles WHERE particle_id = $1)`,
+		particleID,
+	).Scan(&exists)
+	return exists, err
+}
+
+func (r *processingRepositoryImpl) MarkProcessed(ctx context.Context, particleID string) error {
+	_, err := r.pool.Exec(ctx,
+		`INSERT INTO processed_particles (particle_id) VALUES ($1) ON CONFLICT (particle_id) DO NOTHING`,
+		particleID,
+	)
+	return err
+}

go/internal/speech/service.go

@@ -0,0 +1,99 @@
+package speech
+
+import (
+	"context"
+	"log/slog"
+
+	dgapi "github.com/deepgram/deepgram-go-sdk/v3/pkg/api/listen/v1/rest"
+	interfaces "github.com/deepgram/deepgram-go-sdk/v3/pkg/client/interfaces"
+	client "github.com/deepgram/deepgram-go-sdk/v3/pkg/client/listen"
+)
+
+type TranscriptWord struct {
+	Word  string
+	Start float64
+	End   float64
+}
+
+type TranscriptSentence struct {
+	Text  string
+	Start float64
+	End   float64
+}
+
+type TranscriptParagraph struct {
+	Sentences []TranscriptSentence
+	Start     float64
+	End       float64
+}
+
+type TranscriptResult struct {
+	Transcript string
+	Words      []TranscriptWord
+	Paragraphs []TranscriptParagraph
+}
+
+type SpeechService interface {
+	Transcribe(ctx context.Context, mediaUrl string) (*TranscriptResult, error)
+}
+
+type speechServiceImpl struct {
+	deepgramClient *dgapi.Client
+}
+
+func NewSpeechService(ctx context.Context) SpeechService {
+	return &speechServiceImpl{
+		deepgramClient: dgapi.New(client.NewRESTWithDefaults()),
+	}
+}
+
+func (s *speechServiceImpl) Transcribe(ctx context.Context, mediaUrl string) (*TranscriptResult, error) {
+	options := &interfaces.PreRecordedTranscriptionOptions{
+		Model:      "nova-3",
+		SmartFormat: true,
+		Paragraphs: true,
+	}
+
+	response, err := s.deepgramClient.FromURL(ctx, mediaUrl, options)
+	if err != nil {
+		slog.Error("failed to transcribe prerecorded media", "error", err)
+		return nil, err
+	}
+
+	alt := response.Results.Channels[0].Alternatives[0]
+
+	words := make([]TranscriptWord, len(alt.Words))
+	for i, w := range alt.Words {
+		words[i] = TranscriptWord{
+			Word:  w.PunctuatedWord,
+			Start: w.Start,
+			End:   w.End,
+		}
+	}
+
+	var paragraphs []TranscriptParagraph
+	if alt.Paragraphs != nil {
+		paragraphs = make([]TranscriptParagraph, len(alt.Paragraphs.Paragraphs))
+		for i, p := range alt.Paragraphs.Paragraphs {
+			sentences := make([]TranscriptSentence, len(p.Sentences))
+			for j, s := range p.Sentences {
+				sentences[j] = TranscriptSentence{
+					Text:  s.Text,
+					Start: s.Start,
+					End:   s.End,
+				}
+			}
+			paragraphs[i] = TranscriptParagraph{
+				Sentences: sentences,
+				Start:     p.Start,
+				End:       p.End,
+			}
+		}
+	}
+
+	return &TranscriptResult{
+		Transcript: alt.Transcript,
+		Words:      words,
+		Paragraphs: paragraphs,
+	}, nil
+}