First version of the STT module (only synchronous)

mods/main.go

@@ -48,6 +48,7 @@ var mods = map[string]Mod{
 		OnMessage: talk,
 	},
 	"stt": {
+		OnInit:    initstt,
 		OnMessage: stt,
 	},
 }
@@ -90,6 +91,7 @@ var gillmt *string
 var sourcesans *string
 var proverbi *string
 var wittoken *string
+var gapifile *string
 
 func main() {
 	brokerAddr := flag.String("broker", "localhost:7314", "Broker address:port")
@@ -100,6 +102,7 @@ func main() {
 	macropath = flag.String("macropath", "macros.json", "Path to macros db (JSON)")
 	proverbi = flag.String("proverbi", "proverbi.txt", "Path to proverbi pairs (separated by /)")
 	wittoken = flag.String("wit", "", "Wit.ai token")
+	gapifile = flag.String("gapi", "", "Google API Service Credentials file")
 	disable := flag.String("disable", "", "Blacklist mods (separated by comma)")
 	enable := flag.String("enable", "", "Whitelist mods (separated by comma)")
 	flag.Parse()

mods/stt.go

@@ -1,16 +1,30 @@
 package main
 
 import (
+	"context"
 	"fmt"
 	"log"
 
 	"encoding/base64"
 
-	"io/ioutil"
+	speech "cloud.google.com/go/speech/apiv1"
-
 	"github.com/hamcha/clessy/tg"
+	"google.golang.org/api/option"
+	speechpb "google.golang.org/genproto/googleapis/cloud/speech/v1"
 )
 
+var sttClient *speech.Client
+var sttCtx context.Context
+
+func initstt() {
+	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(broker *tg.Broker, update tg.APIMessage) {
 	if isCommand(update, "stt") {
 		// Make replies work
@@ -18,7 +32,10 @@ func stt(broker *tg.Broker, update tg.APIMessage) {
 			update.Voice = update.ReplyTo.Voice
 		}
 
-		broker.SendTextMessage(update.Chat, fmt.Sprintf("Ok! Eccoti intanto delle informazioni sul file: \n<b>Durata: </b> %ds\n<b>Dimensione:</b> %d\n<b>Tipo:</b> %s", update.Voice.Duration, update.Voice.FileSize, *update.Voice.MimeType), &update.MessageID)
+		if update.Voice.Duration > 20 {
+			broker.SendTextMessage(update.Chat, "L'audio dura un bel po' (>20s), 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 {
@@ -34,14 +51,30 @@ func stt(broker *tg.Broker, update tg.APIMessage) {
 				return
 			}
 
-			err = ioutil.WriteFile(update.Voice.FileID, bytes, 0755)
+			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] Save to file error: %s\n", err.Error())
+				log.Printf("[stt] STT request error: %s\n", err.Error())
 				broker.SendTextMessage(update.Chat, "<b>ERRORE!</b> @hamcha controlla la console!", &update.MessageID)
 				return
 			}
 
-			broker.SendTextMessage(update.Chat, fmt.Sprintf("Salvata registrazione su disco! (fname %s)", update.Voice.FileID), &update.MessageID)
+			out := "Questo è quello che ho capito:\n"
+			for _, result := range resp.Results {
+				for _, alt := range result.Alternatives {
+					out += fmt.Sprintf("\"%v\"", alt.Transcript, alt.Confidence)
+				}
+			}
+
+			broker.SendTextMessage(update.Chat, out, &update.MessageID)
 		})
 	}
 }