mono2: transcribe

commands/interaction/context/transcribe.js

@@ -1,7 +1,7 @@
 const { Constants } = require('detritus-client');
 const { InteractionCallbackTypes, ApplicationCommandTypes, MessageFlags } = Constants;
 
-const { transcribeWithSpeakerLabelsObelisk } = require('../../../labscore/api/obelisk');
+const { AudioTranscribe} = require('../../../labscore/api/obelisk');
 
 const { createEmbed } = require('../../../labscore/utils/embed');
 const { codeblock } = require('../../../labscore/utils/markdown');
@@ -34,26 +34,20 @@ module.exports = {
           flags: MessageFlags.EPHEMERAL
         })
         
-        const recog = await transcribeWithSpeakerLabelsObelisk(context, message.attachments.first().url)
+        const recog = await AudioTranscribe(context, message.attachments.first().url)
   
         return editOrReply(context, {
           embeds: [createEmbed("default", context, {
-            description: codeblock("md", [ recog.response.body.transcription.substr(0,3900) ]),
+            description: codeblock("md", [ recog.response.body.transcript.substr(0,3900) ]),
             footer: {
               iconUrl: STATICS.google,
-              text: `Google Cloud • Confidence: ${(recog.response.body.confidence* 100).toFixed(1)}% • ${context.application.name}`
+              text: `Google Speech to Text • ${context.application.name}`
             }
           })],
           flags: MessageFlags.EPHEMERAL
         })
         
       } catch (e) {
-        console.log(e)
-        if(e.response?.body?.status && e.response.body.status == 2) return editOrReply(context, {
-          embeds: [createEmbed("warning", context, e.response.body.message)],
-          flags: MessageFlags.EPHEMERAL
-        })
-
         return editOrReply(context, {
           embeds: [createEmbed("error", context, "Unable to transcribe message.")],
           flags: MessageFlags.EPHEMERAL

commands/message/google/transcribe.js

@@ -1,7 +1,7 @@
 const { createEmbed } = require('../../../labscore/utils/embed')
 const { editOrReply } = require('../../../labscore/utils/message')
 
-const { transcribeWithSpeakerLabelsObelisk } = require('../../../labscore/api/obelisk');
+const { AudioTranscribe} = require('../../../labscore/api/obelisk');
 
 const { STATICS } = require('../../../labscore/utils/statics');
 const { codeblock, icon } = require('../../../labscore/utils/markdown');
@@ -32,19 +32,19 @@ module.exports = {
       if(!msg.attachments.first()) return editOrReply(context, createEmbed("warning", context, "No voice message found."))
       if(!msg.attachments.first().url.split('?')[0].endsWith('voice-message.ogg')) return editOrReply(context, createEmbed("warning", context, "No voice message found."))
       
-      const recog = await transcribeWithSpeakerLabelsObelisk(context, msg.attachments.first().url)
+      const recog = await AudioTranscribe(context, msg.attachments.first().url)
 
       return editOrReply(context, createEmbed("default", context, {
-        description: codeblock("md", [ recog.response.body.transcription ]),
+        description: codeblock("md", [ recog.response.body.transcript ]),
         footer: {
           iconUrl: STATICS.google,
-          text: `Google Cloud • Confidence: ${(recog.response.body.confidence* 100).toFixed(1)}% • ${context.application.name}`
+          text: `Google Speech to Text • ${context.application.name}`
         }
       }))
       
     } catch (e) {
       console.log(e)
-      if(e.response?.body?.status && e.response.body.status == 2) return editOrReply(context, createEmbed("warning", context, e.response.body.message))
+      if(e.response?.body?.error) return editOrReply(context, createEmbed("warning", context, e.response.body.error.message))
       return editOrReply(context, createEmbed("error", context, `Unable to transcribe audio (too long?).`))
     }
   },

labscore/api/obelisk/endpoints.js

@@ -7,6 +7,8 @@ const ObeliskApi = Object.freeze({
   HOST: ObeliskHosts.prod,
 
   // monolith2
+  AUDIO_TRANSCRIBE: "/audio/v1/transcribe",
+
   LLM_PRIVATE_BARD: "/llm/v1/_private:bard",
   LLM_MODELS_GENERATE: "/llm/v1/generate",
 

labscore/api/obelisk/index.js

@@ -40,6 +40,13 @@ async function request(path, type, headers, args, host) {
 }
 
 // monolith2
+module.exports.AudioTranscribe = async function(context, url){
+  return await request(ObeliskApi.AUDIO_TRANSCRIBE, "POST", {}, {
+    url,
+    type: "VOICE_MESSAGE"
+  })
+}
+
 module.exports.LlmPrivateBard = async function(context, prompt){
   return await request(ObeliskApi.LLM_PRIVATE_BARD, "POST", {}, {
     prompt