internal/plugin,internal/telegram,internal/tts: refactor hub router ,…

… switch between different configs

internal/plugin/hub/hub.go

@@ -7,6 +7,7 @@ import (
 	"github.com/scbizu/Astral/internal/plugin/py"
 	"github.com/scbizu/Astral/internal/plugin/sayhi"
 	anime "github.com/scbizu/Astral/internal/plugin/today-anime"
+	"github.com/scbizu/Astral/internal/plugin/tts"
 	"github.com/sirupsen/logrus"
 )
 
@@ -20,6 +21,7 @@ var plugins = []plugin.IPlugin{
 	&sayhi.Handler{},
 	&anime.Handler{},
 	&ai.AICommands{},
+	&tts.TTSCommand{},
 	// py plugin must be put in the end
 	&py.Handler{},
 }
@@ -62,17 +64,20 @@ func (ph *TGPluginHub) AddTGPlugin(
 }
 
 // Do iters telegram plugin
-func (ph *TGPluginHub) Do(rawmsg *tgbotapi.Message) (msg tgbotapi.Chattable) {
+func (ph *TGPluginHub) Do(rawmsg *tgbotapi.Message) tgbotapi.Chattable {
 	for _, p := range ph.GetEnabledTelegramPlugins() {
-		msg, _ = p.Run(rawmsg)
-		if _, ok := msg.(tgbotapi.MessageConfig); ok {
-			msgConf := msg.(tgbotapi.MessageConfig)
+		msg, _ := p.Run(rawmsg)
+		switch msgConf := msg.(type) {
+		case tgbotapi.VoiceConfig:
+			logrus.Infof("[chatID:%d]", msgConf.ChatID)
+			return msgConf
+		case tgbotapi.MessageConfig:
 			logrus.Infof("[chatID:%d,msg:%s]", msgConf.ChatID, msgConf.Text)
 			if p.Validate(msgConf) {
-				return
+				return msgConf
 			}
 		}
 	}
 
-	return
+	return tgbotapi.NewMessage(rawmsg.Chat.ID, "hello ?")
 }

internal/plugin/tts/tts.go

@@ -1,8 +1,10 @@
 package tts
 
 import (
+	"bytes"
 	"context"
 	"fmt"
+	"strings"
 
 	tgbotapi "github.com/go-telegram-bot-api/telegram-bot-api/v5"
 	"github.com/scbizu/Astral/internal/plugin"
@@ -28,13 +30,14 @@ func (t *TTSCommand) Process(msg *tgbotapi.Message) tgbotapi.Chattable {
 		"tts",
 		func(msg *tgbotapi.Message) tgbotapi.Chattable {
 			c := itts.NewElevenLabClient()
-			bs, err := c.ToSpeech(context.Background(), msg.Text)
+			text := strings.TrimPrefix(msg.Text, "/tts")
+			bs, err := c.ToSpeech(context.Background(), text)
 			if err != nil {
 				return tgbotapi.NewMessage(msg.Chat.ID, err.Error())
 			}
-			return tgbotapi.NewVoice(msg.Chat.ID, tgbotapi.FileBytes{
-				Name:  fmt.Sprintf("%s.mp3", msg.Text),
-				Bytes: bs,
+			return tgbotapi.NewVoice(msg.Chat.ID, tgbotapi.FileReader{
+				Name:   fmt.Sprintf("%s.ogg", msg.Text),
+				Reader: bytes.NewBuffer(bs),
 			})
 		},
 	)

internal/telegram/webhook.go

@@ -69,7 +69,14 @@ func (b *Bot) ServeBotUpdateMessage() error {
 		}
 		pluginHub := hub.NewTGPluginHub(update.Message)
 		msg := pluginHub.Do(update.Message)
-		if msgConf, ok := msg.(tgbotapi.MessageConfig); ok {
+		switch msgConf := msg.(type) {
+		case tgbotapi.VoiceConfig:
+			msgConf.ReplyToMessageID = update.Message.MessageID
+			logrus.Debugf("[raw msg][sent]:%#v\n", msg)
+			if _, err := b.bot.Send(msg); err != nil {
+				logrus.Errorf("send msg failed:%q", err)
+			}
+		case tgbotapi.MessageConfig:
 			if isMsgBadRequest(msgConf) {
 				continue
 			}
@@ -83,9 +90,11 @@ func (b *Bot) ServeBotUpdateMessage() error {
 				continue
 			}
 			msgConf.ReplyToMessageID = update.Message.MessageID
-			b.bot.Send(msgConf)
-		} else {
-			b.bot.Send(msg)
+			if _, err := b.bot.Send(msgConf); err != nil {
+				logrus.Errorf("send msg failed:%q", err)
+			}
+		default:
+			logrus.Errorf("unknown msg type:%T", msgConf)
 		}
 	}
 	return nil

internal/tts/eleven_lab.go

@@ -4,6 +4,7 @@ import (
 	"bytes"
 	"context"
 	"encoding/json"
+	"fmt"
 	"io"
 	"net/http"
 	"net/url"
@@ -17,12 +18,19 @@ type Client struct {
 	model   string
 }
 
+type BadResponse struct {
+	Detail struct {
+		Status  string `json:"status"`
+		Message string `json:"message"`
+	} `json:"detail"`
+}
+
 func NewElevenLabClient() *Client {
 	return &Client{
 		domain:  "https://api.elevenlabs.io/v1/text-to-speech/",
 		voiceID: "21m00Tcm4TlvDq8ikWAM",
 		apiKey:  os.Getenv("ELEVEN_LAB_API_KEY"),
-		model:   "eleven_monolingual_v2",
+		model:   "eleven_monolingual_v1",
 	}
 }
 
@@ -46,6 +54,9 @@ func (c *Client) ToSpeech(ctx context.Context, text string) ([]byte, error) {
 	if err != nil {
 		return nil, err
 	}
+	req.Header.Set("Content-Type", "application/json")
+	req.Header.Set("xi-api-key", c.apiKey)
+
 	resp, err := http.DefaultClient.Do(req.WithContext(ctx))
 	if err != nil {
 		return nil, err
@@ -57,5 +68,13 @@ func (c *Client) ToSpeech(ctx context.Context, text string) ([]byte, error) {
 		return nil, err
 	}
 
+	if resp.StatusCode != http.StatusOK {
+		details := &BadResponse{}
+		if err := json.NewDecoder(bytes.NewBuffer(respBody)).Decode(&details); err != nil {
+			return nil, err
+		}
+		return nil, fmt.Errorf("[11lab] bad response: %s", details.Detail.Message)
+	}
+
 	return respBody, nil
 }