Feature/stt long voice messages

.gitignore

@@ -42,4 +42,7 @@ docker/traefik/acme.json
 docker/traefik/.env
 
 # my personal compose
-docker/docker-compose-local.yaml
+docker/docker-compose-local.yaml
+
+# codegraph
+.codegraph

cmd/service/main.go

@@ -7,13 +7,16 @@ import (
 	"noirbot/internal/domain/repository"
 	"noirbot/internal/domain/service"
 	"noirbot/internal/gateways/deepseek"
+	"noirbot/internal/gateways/groq"
 	httpgw "noirbot/internal/gateways/http"
 	"noirbot/internal/gateways/memory"
 	redisstore "noirbot/internal/gateways/redis"
+	"noirbot/internal/gateways/telegram/files"
 	"noirbot/internal/gateways/telegram/inbound"
 	"noirbot/internal/gateways/telegram/outbound"
-	"noirbot/internal/usecase/handle_business_connection"
-	"noirbot/internal/usecase/handle_business_message"
+	"noirbot/internal/usecase/businessconn"
+	"noirbot/internal/usecase/businessmsg"
+	"noirbot/internal/usecase/longvoice"
 	"noirbot/pkg/config"
 	"os"
 
@@ -32,6 +35,7 @@ func main() {
 			newGreetingDetector,
 			newFloodDetector,
 			newShortVoiceDetector,
+			newLongVoiceDetector,
 			newVoiceReplyWindowStore,
 
 			newOwnerWhitelist,
@@ -44,11 +48,17 @@ func main() {
 			newDeepseekConfig,
 			newLLMClient,
 
+			newGroqConfig,
+			newTranscriber,
+			newVoiceDownloader,
+			newHandleLongVoiceConfig,
+			longvoice.New,
+
 			newRedisClient,
 
 			newHandleBusinessMessageConfig,
-			handle_business_connection.New,
-			handle_business_message.New,
+			businessconn.New,
+			businessmsg.New,
 
 			inbound.NewLazyHandler,
 			inbound.NewUpdateMapper,
@@ -168,8 +178,8 @@ func newLLMClient(c deepseek.Config) repository.LLMClient {
 	return deepseek.NewClient(c)
 }
 
-func newHandleBusinessMessageConfig(cfg *config.Config) handle_business_message.Config {
-	return handle_business_message.Config{
+func newHandleBusinessMessageConfig(cfg *config.Config) businessmsg.Config {
+	return businessmsg.Config{
 		SystemPrompt:             cfg.Bot.SystemPrompt,
 		ShortVoicePrompt:         cfg.Bot.ShortVoicePrompt,
 		ShortVoiceResponseWindow: cfg.ShortVoice.ResponseWindow,
@@ -189,3 +199,34 @@ func newRedisClient(cfg *config.Config) *redis.Client {
 
 	return rdb
 }
+
+func newLongVoiceDetector(cfg *config.Config) *service.LongVoiceDetector {
+	return service.NewLongVoiceDetector(
+		service.LongVoiceDetectorConfig{
+			MaxDuration: cfg.LongVoice.MaxDuration,
+		},
+	)
+}
+
+func newGroqConfig(cfg *config.Config) groq.Config {
+	return groq.Config{
+		BaseURL: cfg.Groq.BaseURL,
+		APIKey:  cfg.Groq.APIKey,
+		Model:   cfg.Groq.Model,
+		Timeout: cfg.Groq.Timeout,
+	}
+}
+
+func newTranscriber(cfg groq.Config) repository.Transcriber {
+	return groq.NewClient(cfg)
+}
+
+func newVoiceDownloader(b *bot.Bot, cfg *config.Config) repository.VoiceDownloader {
+	return files.NewDownloader(b, cfg.Telegram.FileDownloadTimeout)
+}
+
+func newHandleLongVoiceConfig(cfg *config.Config) longvoice.Config {
+	return longvoice.Config{
+		LongVoicePrompt: cfg.Bot.LongVoicePrompt,
+	}
+}

internal/domain/model/message.go

@@ -17,6 +17,7 @@ type IncomingMessage struct {
 	Text                 string
 	VoiceDuration        time.Duration
 	ReceivedAt           time.Time
+	VoiceFileID          string
 }
 
 type ReplyDraft struct {

internal/domain/model/trigger.go

@@ -7,6 +7,7 @@ const (
 	TriggerKindGreeting   TriggerKind = "greeting"
 	TriggerKindFlood      TriggerKind = "flood"
 	TriggerKindShortVoice TriggerKind = "short_voice"
+	TriggerKindLongVoice  TriggerKind = "long_voice"
 )
 
 type TriggerDecision struct {

internal/domain/repository/transcriber.go

@@ -0,0 +1,12 @@
+package repository
+
+import (
+	"context"
+	"io"
+)
+
+//go:generate go tool mockgen -source=$GOFILE -destination=mock/$GOFILE -package=mock
+
+type Transcriber interface {
+	Transcribe(ctx context.Context, reader io.ReadCloser) (string, error)
+}

internal/domain/repository/voice_downloader.go

@@ -0,0 +1,12 @@
+package repository
+
+import (
+	"context"
+	"io"
+)
+
+//go:generate go tool mockgen -source=$GOFILE -destination=mock/$GOFILE -package=mock
+
+type VoiceDownloader interface {
+	Download(ctx context.Context, fileID string) (io.ReadCloser, error)
+}

internal/domain/service/long_voice_detector.go

@@ -0,0 +1,38 @@
+package service
+
+import (
+	"fmt"
+	"noirbot/internal/domain/model"
+	"time"
+)
+
+type LongVoiceDetectorConfig struct {
+	MinDuration time.Duration
+	MaxDuration time.Duration
+}
+
+type LongVoiceDetector struct {
+	cfg LongVoiceDetectorConfig
+}
+
+func NewLongVoiceDetector(cfg LongVoiceDetectorConfig) *LongVoiceDetector {
+	return &LongVoiceDetector{
+		cfg: cfg,
+	}
+}
+
+func (d *LongVoiceDetector) Detect(msg model.IncomingMessage) model.TriggerDecision {
+	switch {
+	case msg.Kind != model.MessageKindVoice:
+		return model.TriggerDecision{Kind: model.TriggerKindNone}
+	case msg.VoiceDuration <= d.cfg.MinDuration:
+		return model.TriggerDecision{Kind: model.TriggerKindNone}
+	case msg.VoiceDuration > d.cfg.MaxDuration:
+		return model.TriggerDecision{Kind: model.TriggerKindNone}
+	default:
+		return model.TriggerDecision{
+			Kind:   model.TriggerKindLongVoice,
+			Reason: fmt.Sprintf("voice %s > %s <= %s", msg.VoiceDuration.String(), d.cfg.MinDuration.String(), d.cfg.MaxDuration.String()),
+		}
+	}
+}