clessy/mods/stt.go

package main

import (
	"context"
	"fmt"
	"log"

	"encoding/base64"

	speech "cloud.google.com/go/speech/apiv1"
	"git.fromouter.space/hamcha/tg"
	"google.golang.org/api/option"
	speechpb "google.golang.org/genproto/googleapis/cloud/speech/v1"
)

var sttClient *speech.Client
var sttCtx context.Context

func stt_init() {
	sttCtx = context.Background()
	var err error
	sttClient, err = speech.NewClient(sttCtx, option.WithServiceAccountFile(*gapifile))
	if err != nil {
		panic(fmt.Errorf("Could not initialize Google Speech API client: %s", err.Error()))
	}
}

func stt_message(broker *tg.Broker, update tg.APIMessage) {
	if isCommand(update, "stt") {
		// Make replies work
		if update.ReplyTo != nil && update.ReplyTo.Voice != nil {
			update.Voice = update.ReplyTo.Voice
		}

		// Make sure it's a voice message
		if update.Voice == nil {
			broker.SendTextMessage(update.Chat, "Non vedo un messaggio vocale da ascoltare", &update.MessageID)
			return
		}

		if update.Voice.Duration > 60 {
			broker.SendTextMessage(update.Chat, "L'audio dura un bel po' (> 1 minuto), al momento non posso farci nulla :S", &update.MessageID)
			return
		}

		broker.GetFile(update.Voice.FileID, func(broker *tg.Broker, data tg.BrokerUpdate) {
			if data.Type == tg.BError {
				log.Printf("[stt] Received error from broker: %s\n", *data.Error)
				broker.SendTextMessage(update.Chat, "<b>ERRORE!</b> @hamcha controlla la console!", &update.MessageID)
				return
			}

			bytes, err := base64.StdEncoding.DecodeString(*data.Bytes)
			if err != nil {
				log.Printf("[stt] Base64 decode error: %s\n", err.Error())
				broker.SendTextMessage(update.Chat, "<b>ERRORE!</b> @hamcha controlla la console!", &update.MessageID)
				return
			}

			broker.SendChatAction(update.Chat, tg.ActionTyping)
			resp, err := sttClient.Recognize(sttCtx, &speechpb.RecognizeRequest{
				Config: &speechpb.RecognitionConfig{
					Encoding:        speechpb.RecognitionConfig_OGG_OPUS,
					SampleRateHertz: 16000,
					LanguageCode:    "it-IT",
				},
				Audio: &speechpb.RecognitionAudio{
					AudioSource: &speechpb.RecognitionAudio_Content{Content: bytes},
				},
			})
			if err != nil {
				log.Printf("[stt] STT request error: %s\n", err.Error())
				broker.SendTextMessage(update.Chat, "<b>ERRORE!</b> @hamcha controlla la console!", &update.MessageID)
				return
			}

			out := "Questo è quello che ho capito:\n"
			for _, result := range resp.Results {
				for _, alt := range result.Alternatives {
					out += fmt.Sprintf("\"%v\"\n", alt.Transcript)
				}
			}

			broker.SendTextMessage(update.Chat, out, &update.MessageID)
		})
	}
}