Revert "Feat/speech support (#392)" (#393)

This reverts commit c9dcfe287f.

cmd/agent/app.go

@@ -23,7 +23,6 @@ import (
 	agentpkg "github.com/memohai/memoh/internal/agent"
 	"github.com/memohai/memoh/internal/agent/background"
 	agenttools "github.com/memohai/memoh/internal/agent/tools"
-	audiopkg "github.com/memohai/memoh/internal/audio"
 	"github.com/memohai/memoh/internal/bind"
 	"github.com/memohai/memoh/internal/boot"
 	"github.com/memohai/memoh/internal/bots"
@@ -88,6 +87,7 @@ import (
 	"github.com/memohai/memoh/internal/storage/providers/containerfs"
 	"github.com/memohai/memoh/internal/storage/providers/fallback"
 	"github.com/memohai/memoh/internal/storage/providers/localfs"
+	ttspkg "github.com/memohai/memoh/internal/tts"
 	"github.com/memohai/memoh/internal/version"
 	"github.com/memohai/memoh/internal/workspace"
 )
@@ -331,7 +331,7 @@ func provideChannelRouter(
 	policyService *policy.Service,
 	bindService *bind.Service,
 	mediaService *media.Service,
-	audioService *audiopkg.Service,
+	ttsService *ttspkg.Service,
 	settingsService *settings.Service,
 	scheduleService *schedule.Service,
 	mcpConnService *mcp.ConnectionService,
@@ -372,8 +372,7 @@ func provideChannelRouter(
 	processor.SetMediaService(mediaService)
 	processor.SetStreamObserver(local.NewRouteHubBroadcaster(hub))
 	processor.SetDispatcher(inbound.NewRouteDispatcher(log))
-	processor.SetSpeechService(audioService, &settingsSpeechModelResolver{settings: settingsService})
-	processor.SetTranscriptionService(&settingsTranscriptionAdapter{audio: audioService}, &settingsTranscriptionModelResolver{settings: settingsService})
+	processor.SetTtsService(ttsService, &settingsTtsModelResolver{settings: settingsService})
 	cmdHandler := command.NewHandler(
 		log,
 		&command.BotMemberRoleAdapter{BotService: botService},
@@ -450,7 +449,7 @@ func provideBackgroundManager(log *slog.Logger) *background.Manager {
 	return background.New(log)
 }
 
-func provideToolProviders(log *slog.Logger, cfg config.Config, channelManager *channel.Manager, registry *channel.Registry, routeService *route.DBService, scheduleService *schedule.Service, settingsService *settings.Service, searchProviderService *searchproviders.Service, manager *workspace.Manager, mediaService *media.Service, memoryRegistry *memprovider.Registry, emailService *emailpkg.Service, emailManager *emailpkg.Manager, fedGateway *handlers.MCPFederationGateway, mcpConnService *mcp.ConnectionService, modelsService *models.Service, browserContextService *browsercontexts.Service, queries *dbsqlc.Queries, audioService *audiopkg.Service, sessionService *sessionpkg.Service, bgManager *background.Manager) []agenttools.ToolProvider {
+func provideToolProviders(log *slog.Logger, cfg config.Config, channelManager *channel.Manager, registry *channel.Registry, routeService *route.DBService, scheduleService *schedule.Service, settingsService *settings.Service, searchProviderService *searchproviders.Service, manager *workspace.Manager, mediaService *media.Service, memoryRegistry *memprovider.Registry, emailService *emailpkg.Service, emailManager *emailpkg.Manager, fedGateway *handlers.MCPFederationGateway, mcpConnService *mcp.ConnectionService, modelsService *models.Service, browserContextService *browsercontexts.Service, queries *dbsqlc.Queries, ttsService *ttspkg.Service, sessionService *sessionpkg.Service, bgManager *background.Manager) []agenttools.ToolProvider {
 	var assetResolver messaging.AssetResolver
 	if mediaService != nil {
 		assetResolver = &mediaAssetResolverAdapter{media: mediaService}
@@ -468,8 +467,7 @@ func provideToolProviders(log *slog.Logger, cfg config.Config, channelManager *c
 		agenttools.NewSpawnProvider(log, settingsService, modelsService, queries, sessionService),
 		agenttools.NewSkillProvider(log),
 		agenttools.NewBrowserProvider(log, settingsService, browserContextService, manager, cfg.BrowserGateway),
-		agenttools.NewTTSProvider(log, settingsService, audioService, channelManager, registry),
-		agenttools.NewTranscriptionProvider(log, settingsService, audioService, mediaService),
+		agenttools.NewTTSProvider(log, settingsService, ttsService, channelManager, registry),
 		agenttools.NewImageGenProvider(log, settingsService, modelsService, queries, manager, config.DefaultDataMount),
 		agenttools.NewFederationProvider(log, fedSource),
 		agenttools.NewHistoryProvider(log, sessionService, queries),
@@ -513,23 +511,23 @@ func provideUsersHandler(log *slog.Logger, accountService *accounts.Service, ide
 	return handlers.NewUsersHandler(log, accountService, identityService, botService, routeService, channelStore, channelLifecycle, channelManager, registry)
 }
 
-func provideWebHandler(channelManager *channel.Manager, channelStore *channel.Store, chatService *conversation.Service, hub *local.RouteHub, botService *bots.Service, accountService *accounts.Service, resolver *flow.Resolver, mediaService *media.Service, audioService *audiopkg.Service, settingsService *settings.Service) *handlers.LocalChannelHandler {
+func provideWebHandler(channelManager *channel.Manager, channelStore *channel.Store, chatService *conversation.Service, hub *local.RouteHub, botService *bots.Service, accountService *accounts.Service, resolver *flow.Resolver, mediaService *media.Service, ttsService *ttspkg.Service, settingsService *settings.Service) *handlers.LocalChannelHandler {
 	h := handlers.NewLocalChannelHandler(local.WebType, channelManager, channelStore, chatService, hub, botService, accountService)
 	h.SetResolver(resolver)
 	h.SetMediaService(mediaService)
-	h.SetSpeechService(audioService, &settingsSpeechModelResolver{settings: settingsService})
+	h.SetTtsService(ttsService, &settingsTtsModelResolver{settings: settingsService})
 	return h
 }
 
-func provideAudioRegistry() *audiopkg.Registry {
-	return audiopkg.NewRegistry()
+func provideTtsRegistry() *ttspkg.Registry {
+	return ttspkg.NewRegistry()
 }
 
-func provideAudioTempStore() (*audiopkg.TempStore, error) {
-	return audiopkg.NewTempStore(os.TempDir())
+func provideTtsTempStore() (*ttspkg.TempStore, error) {
+	return ttspkg.NewTempStore(os.TempDir())
 }
 
-func startAudioTempStoreCleanup(lc fx.Lifecycle, store *audiopkg.TempStore) {
+func startTtsTempStoreCleanup(lc fx.Lifecycle, store *ttspkg.TempStore) {
 	done := make(chan struct{})
 	lc.Append(fx.Hook{
 		OnStart: func(_ context.Context) error {
@@ -585,11 +583,11 @@ func (a *sessionEnsurerAdapter) CreateNewSession(ctx context.Context, botID, rou
 	return inbound.SessionResult{ID: sess.ID, Type: sess.Type}, nil
 }
 
-type settingsSpeechModelResolver struct {
+type settingsTtsModelResolver struct {
 	settings *settings.Service
 }
 
-func (r *settingsSpeechModelResolver) ResolveSpeechModelID(ctx context.Context, botID string) (string, error) {
+func (r *settingsTtsModelResolver) ResolveTtsModelID(ctx context.Context, botID string) (string, error) {
 	s, err := r.settings.GetBot(ctx, botID)
 	if err != nil {
 		return "", err
@@ -597,36 +595,6 @@ func (r *settingsSpeechModelResolver) ResolveSpeechModelID(ctx context.Context,
 	return s.TtsModelID, nil
 }
 
-type settingsTranscriptionModelResolver struct {
-	settings *settings.Service
-}
-
-func (r *settingsTranscriptionModelResolver) ResolveTranscriptionModelID(ctx context.Context, botID string) (string, error) {
-	s, err := r.settings.GetBot(ctx, botID)
-	if err != nil {
-		return "", err
-	}
-	return s.TranscriptionModelID, nil
-}
-
-type settingsTranscriptionAdapter struct {
-	audio *audiopkg.Service
-}
-
-type inboundTranscriptionResult struct {
-	text string
-}
-
-func (r inboundTranscriptionResult) GetText() string { return r.text }
-
-func (a *settingsTranscriptionAdapter) Transcribe(ctx context.Context, modelID string, audio []byte, filename string, contentType string, overrideCfg map[string]any) (inbound.TranscriptionResult, error) {
-	result, err := a.audio.Transcribe(ctx, modelID, audio, filename, contentType, overrideCfg)
-	if err != nil {
-		return nil, err
-	}
-	return inboundTranscriptionResult{text: result.Text}, nil
-}
-
 func provideEmailRegistry(log *slog.Logger, tokenStore *emailpkg.DBOAuthTokenStore) *emailpkg.Registry {
 	reg := emailpkg.NewRegistry()
 	reg.Register(emailgeneric.New(log))
@@ -716,11 +684,11 @@ func startRegistrySync(lc fx.Lifecycle, log *slog.Logger, cfg config.Config, que
 	})
 }
 
-func startAudioProviderBootstrap(lc fx.Lifecycle, log *slog.Logger, queries *dbsqlc.Queries, registry *audiopkg.Registry) {
+func startSpeechProviderBootstrap(lc fx.Lifecycle, log *slog.Logger, queries *dbsqlc.Queries, registry *ttspkg.Registry) {
 	lc.Append(fx.Hook{
 		OnStart: func(ctx context.Context) error {
-			if err := audiopkg.SyncRegistry(ctx, log, queries, registry); err != nil {
-				log.Warn("audio registry bootstrap failed", slog.Any("error", err))
+			if err := ttspkg.SyncRegistry(ctx, log, queries, registry); err != nil {
+				log.Warn("speech registry bootstrap failed", slog.Any("error", err))
 			}
 			return nil
 		},

cmd/agent/module.go

@@ -8,7 +8,6 @@ import (
 
 	"github.com/memohai/memoh/internal/accounts"
 	"github.com/memohai/memoh/internal/acl"
-	audiopkg "github.com/memohai/memoh/internal/audio"
 	"github.com/memohai/memoh/internal/bind"
 	"github.com/memohai/memoh/internal/boot"
 	"github.com/memohai/memoh/internal/bots"
@@ -30,6 +29,7 @@ import (
 	"github.com/memohai/memoh/internal/schedule"
 	"github.com/memohai/memoh/internal/searchproviders"
 	"github.com/memohai/memoh/internal/settings"
+	ttspkg "github.com/memohai/memoh/internal/tts"
 )
 
 func runServe() {
@@ -63,9 +63,9 @@ func options() fx.Option {
 			identities.NewService,
 			bind.NewService,
 			event.NewHub,
-			provideAudioRegistry,
-			audiopkg.NewService,
-			provideAudioTempStore,
+			provideTtsRegistry,
+			ttspkg.NewService,
+			provideTtsTempStore,
 			emailpkg.NewDBOAuthTokenStore,
 			provideEmailRegistry,
 			emailpkg.NewService,
@@ -121,8 +121,8 @@ func options() fx.Option {
 			provideServerHandler(weixin.NewQRServerHandler),
 			provideServerHandler(provideUsersHandler),
 			provideServerHandler(handlers.NewMemoryProvidersHandler),
-			provideServerHandler(handlers.NewAudioHandler),
-			provideServerHandler(handlers.NewBotAudioHandler),
+			provideServerHandler(handlers.NewSpeechHandler),
+			provideServerHandler(handlers.NewBotTtsHandler),
 			provideServerHandler(handlers.NewEmailProvidersHandler),
 			provideServerHandler(handlers.NewEmailBindingsHandler),
 			provideServerHandler(handlers.NewEmailOutboxHandler),
@@ -141,7 +141,7 @@ func options() fx.Option {
 		fx.Invoke(
 			injectToolProviders,
 			startRegistrySync,
-			startAudioProviderBootstrap,
+			startSpeechProviderBootstrap,
 			startMemoryProviderBootstrap,
 			startSearchProviderBootstrap,
 			startScheduleService,
@@ -151,7 +151,7 @@ func options() fx.Option {
 			startEmailManager,
 			startContainerReconciliation,
 			startBackgroundTaskCleanup,
-			startAudioTempStoreCleanup,
+			startTtsTempStoreCleanup,
 			startServer,
 		),
 		fx.WithLogger(func(logger *slog.Logger) fxevent.Logger {