Basic for whisper support (via jarvis-server and whisper-asr)

.idea/misc.xml

@@ -45,6 +45,7 @@
         <entry key="../../../../../layout/compose-model-1641396094843.xml" value="0.33" />
         <entry key="../../../../../layout/compose-model-1641399721181.xml" value="2.0" />
         <entry key="../../../../../layout/compose-model-1641463307897.xml" value="2.0" />
+        <entry key="../../../../../layout/compose-model-1641566263645.xml" value="0.33" />
         <entry key="app/src/main/res/drawable-v24/ic_launcher_foreground.xml" value="0.5307291666666667" />
         <entry key="app/src/main/res/drawable/ic_baseline_arrow_back_24.xml" value="0.38981481481481484" />
         <entry key="app/src/main/res/drawable/ic_baseline_keyboard_24.xml" value="0.38981481481481484" />

app/src/main/java/ch/mathieubroillet/jarvis/android/pages/MainPage.kt

@@ -7,21 +7,25 @@ import androidx.compose.ui.Alignment
 import androidx.compose.ui.Modifier
 import androidx.compose.ui.res.painterResource
 import androidx.compose.ui.res.stringResource
+import androidx.compose.ui.tooling.preview.Preview
 import androidx.compose.ui.unit.DpOffset
 import androidx.compose.ui.unit.dp
 import androidx.compose.ui.unit.sp
 import androidx.navigation.NavController
+import androidx.navigation.compose.rememberNavController
 import ch.mathieubroillet.jarvis.android.R
 import ch.mathieubroillet.jarvis.android.audio.AudioRecorder
 import ch.mathieubroillet.jarvis.android.chat.ConversationUiState
 import ch.mathieubroillet.jarvis.android.chat.Message
 import ch.mathieubroillet.jarvis.android.chat.Messages
 import ch.mathieubroillet.jarvis.android.nav.Screen
+import ch.mathieubroillet.jarvis.android.ui.theme.JarvisComposeTheme
 import ch.mathieubroillet.jarvis.android.ui.theme.productSansFont
 import ch.mathieubroillet.jarvis.android.utils.DefaultBox
 import ch.mathieubroillet.jarvis.android.utils.IconAlertDialogTextField
 import ch.mathieubroillet.jarvis.android.utils.contactServerWithFileAudioRecording
 import com.github.squti.androidwaverecorder.RecorderState
+import com.github.squti.androidwaverecorder.WaveRecorder
 import org.json.JSONObject
 import kotlin.concurrent.thread
 
@@ -150,13 +154,12 @@ fun DisplayMainPage(
                             val requestOutput =
                                 contactServerWithFileAudioRecording(audioRecorder.getOutputFile())
 
-                            val json: JSONObject = JSONObject(requestOutput)
+                            val json = JSONObject(requestOutput)
-                            val sent = JSONObject(requestOutput).getString("sent").replace("\"", "")
+                            val sent = json.getString("transcription")
-                                .replace("[", "").replace("]", "").replace(",", " ")
+
-                            sent.replaceFirstChar { sent.first().uppercase() }
                             uiState.addMessage(Message(false, sent))
                             Thread.sleep(1000)
-                            uiState.addMessage(Message(true, json.getString("response")))
+                            uiState.addMessage(Message(true, json.getString("answer")))
                             audioRecorder.getOutputFile().delete()
                         }
                     }
@@ -172,14 +175,14 @@ fun DisplayMainPage(
     }
 }
 
-/*@Preview(showBackground = true)
+@Preview(showBackground = true)
 @Composable
 fun MainPagePreview() {
     JarvisComposeTheme {
         DisplayMainPage(
             rememberNavController(), ConversationUiState(
                 listOf(Message(true, stringResource(id = R.string.demo_message_1)))
-            ), null
+            ), audioRecorder = AudioRecorder("", WaveRecorder(""))
         )
     }
-}*/
+}

app/src/main/java/ch/mathieubroillet/jarvis/android/utils/RequestsUtils.kt

@@ -1,9 +1,11 @@
 package ch.mathieubroillet.jarvis.android.utils
 
+import okhttp3.Headers
 import okhttp3.MediaType.Companion.toMediaType
 import okhttp3.OkHttpClient
 import okhttp3.Request
 import okhttp3.RequestBody.Companion.asRequestBody
+import okhttp3.internal.http1.HeadersReader
 import java.io.File
 import java.io.IOException
 
@@ -11,12 +13,38 @@ fun contactServerWithFileAudioRecording(file: File): String {
     val client = OkHttpClient()
 
     val request = Request.Builder()
-        .url("http://192.168.1.130:5000/process_audio_request_file")
+        .url("https://whisper.broillet.ch/process_audio_request_file")
-        .post(file.asRequestBody("audio/mpeg; charset=utf-8".toMediaType()))
+        .post(file.asRequestBody("audio/x-wav; charset=utf-8".toMediaType()))
         .build()
 
+
     client.newCall(request).execute().use { response ->
         if (!response.isSuccessful) throw IOException("Unexpected code $response")
         return response.body!!.string()
     }
+
+    /* # SEND STT REQUEST DIRECTLY TO WHISPER ASR WITHOUT JARVIS SERVER API
+    val okHttpClient = OkHttpClient()
+        .newBuilder()
+        .connectTimeout(60, TimeUnit.SECONDS)
+        .build()
+
+    val fileBody: RequestBody = RequestBody.create("audio/x-wav".toMediaTypeOrNull(), file.readBytes())
+
+    val multipartBody: MultipartBody = MultipartBody.Builder()
+        .setType(MultipartBody.FORM) // Header to show we are sending a Multipart Form Data
+        .addFormDataPart("audio_file", file.name, fileBody) // file param
+        .build()
+
+    val request = Request.Builder()
+        .addHeader("Accept", "application/json")
+        .addHeader("Content-Type", "multipart/form-data")
+        .url("http://192.168.1.208:9000/asr?task=transcribe&language=fr&output=json")
+        .post(multipartBody)
+        .build()
+
+    okHttpClient.newCall(request).execute().use { response ->
+        if (!response.isSuccessful) throw IOException("Unexpected code $response")
+        return response.body!!.string()
+    }*/
 }