com.openai.unity 8.8.1 (#394)

- updated realtime audio transcription settings properties
- bumped deps

---------

Co-authored-by: AG3D <[email protected]>
Co-authored-by: Andries Geens <[email protected]>

Author: 3 people

OpenAI/Packages/com.openai.unity/Runtime/Realtime/ClientSecret.cs.meta

@@ -2,18 +2,43 @@
 
 using Newtonsoft.Json;
 using OpenAI.Models;
+using UnityEngine.Scripting;
 
 namespace OpenAI.Realtime
 {
+    [Preserve]
     public sealed class InputAudioTranscriptionSettings
     {
+        [Preserve]
+        public InputAudioTranscriptionSettings(Model model, string prompt = null, string language = null)
+        {
+            Model = string.IsNullOrWhiteSpace(model?.Id) ? Models.Model.Whisper1 : model;
+            Prompt = prompt;
+            Language = language;
+        }
+
+        [Preserve]
         [JsonConstructor]
-        public InputAudioTranscriptionSettings([JsonProperty("model")] Model model)
+        internal InputAudioTranscriptionSettings(
+            [JsonProperty("model")] string model,
+            [JsonProperty("prompt")] string prompt = null,
+            [JsonProperty("language")] string language = null)
         {
-            Model = string.IsNullOrWhiteSpace(model.Id) ? "whisper-1" : model;
+            Model = model;
+            Prompt = prompt;
+            Language = language;
         }
 
+        [Preserve]
         [JsonProperty("model")]
         public string Model { get; }
+
+        [Preserve]
+        [JsonProperty("prompt")]
+        public string Prompt { get; }
+
+        [Preserve]
+        [JsonProperty("language")]
+        public string Language { get; }
     }
 }

OpenAI/Packages/com.openai.unity/Runtime/Realtime/ExpiresAfter.cs.meta

@@ -104,9 +104,7 @@ public Options(
                 : instructions;
             InputAudioFormat = inputAudioFormat;
             OutputAudioFormat = outputAudioFormat;
-            InputAudioTranscriptionSettings = new(string.IsNullOrWhiteSpace(transcriptionModel)
-                ? "whisper-1"
-                : transcriptionModel);
+            InputAudioTranscriptionSettings = new(transcriptionModel);
             VoiceActivityDetectionSettings = turnDetectionSettings ?? new();
             tools.ProcessTools<Tool>(toolChoice, out var toolList, out var activeTool);
             Tools = toolList?.Where(t => t.IsFunction).Select(tool =>

OpenAI/Packages/com.openai.unity/Runtime/Realtime/InputAudioTranscriptionSettings.cs

@@ -3,22 +3,57 @@
 using Newtonsoft.Json;
 using OpenAI.Extensions;
 using OpenAI.Models;
+using System;
 using System.Collections.Generic;
 using System.Linq;
 using UnityEngine.Scripting;
 
 namespace OpenAI.Realtime
 {
+    [Preserve]
     public sealed class SessionConfiguration
     {
+        [Obsolete("use new ctor overload")]
+        public SessionConfiguration(
+            Model model,
+            Modality modalities,
+            Voice voice,
+            string instructions,
+            RealtimeAudioFormat inputAudioFormat,
+            RealtimeAudioFormat outputAudioFormat,
+            Model transcriptionModel,
+            IVoiceActivityDetectionSettings turnDetectionSettings,
+            IEnumerable<Tool> tools,
+            string toolChoice,
+            float? temperature,
+            int? maxResponseOutputTokens,
+            int? expiresAfter)
+          : this(
+              model: model,
+              modalities: modalities,
+              voice: voice,
+              instructions: instructions,
+              inputAudioFormat: inputAudioFormat,
+              outputAudioFormat: outputAudioFormat,
+              inputAudioTranscriptionSettings: new(transcriptionModel),
+              turnDetectionSettings: turnDetectionSettings,
+              tools: tools,
+              toolChoice: toolChoice,
+              temperature: temperature,
+              maxResponseOutputTokens: maxResponseOutputTokens,
+              expiresAfter: expiresAfter)
+        {
+        }
+
+        [Preserve]
         public SessionConfiguration(
             Model model,
             Modality modalities = Modality.Text | Modality.Audio,
             Voice voice = null,
             string instructions = null,
             RealtimeAudioFormat inputAudioFormat = RealtimeAudioFormat.PCM16,
             RealtimeAudioFormat outputAudioFormat = RealtimeAudioFormat.PCM16,
-            Model transcriptionModel = null,
+            InputAudioTranscriptionSettings inputAudioTranscriptionSettings = null,
             IVoiceActivityDetectionSettings turnDetectionSettings = null,
             IEnumerable<Tool> tools = null,
             string toolChoice = null,
@@ -27,11 +62,9 @@ public SessionConfiguration(
             int? expiresAfter = null)
         {
             ClientSecret = new ClientSecret(expiresAfter);
-            Model = string.IsNullOrWhiteSpace(model.Id)
-                ? Models.Model.GPT4oRealtime
-                : model;
+            Model = string.IsNullOrWhiteSpace(model?.Id) ? Models.Model.GPT4oRealtime : model;
             Modalities = modalities;
-            Voice = voice ?? OpenAI.Voice.Alloy;
+            Voice = string.IsNullOrWhiteSpace(voice?.Id) ? OpenAI.Voice.Alloy : voice;
             Instructions = string.IsNullOrWhiteSpace(instructions)
                 ? "Your knowledge cutoff is 2023-10. You are a helpful, witty, and friendly AI. Act like a human, " +
                   "but remember that you aren't a human and that you can't do human things in the real world. " +
@@ -42,9 +75,7 @@ public SessionConfiguration(
                 : instructions;
             InputAudioFormat = inputAudioFormat;
             OutputAudioFormat = outputAudioFormat;
-            InputAudioTranscriptionSettings = new(string.IsNullOrWhiteSpace(transcriptionModel)
-                ? Models.Model.Whisper1
-                : transcriptionModel);
+            InputAudioTranscriptionSettings = inputAudioTranscriptionSettings;
             VoiceActivityDetectionSettings = turnDetectionSettings ?? new ServerVAD();
             tools.ProcessTools<Tool>(toolChoice, out var toolList, out var activeTool);
             Tools = toolList?.Where(t => t.IsFunction).Select(tool =>
@@ -66,6 +97,7 @@ public SessionConfiguration(
             }
         }
 
+        [Preserve]
         internal SessionConfiguration(
             string model,
             Modality modalities,
@@ -94,6 +126,7 @@ internal SessionConfiguration(
             MaxResponseOutputTokens = maxResponseOutputTokens;
         }
 
+        [Preserve]
         [JsonConstructor]
         internal SessionConfiguration(
             [JsonProperty("client_secret")] ClientSecret clientSecret,

OpenAI/Packages/com.openai.unity/Runtime/Realtime/Options.cs

@@ -1,9 +1,9 @@
 {
   "name": "com.openai.unity",
   "displayName": "OpenAI",
-  "description": "A OpenAI package for the Unity Game Engine to use GPT-4, GPT-3.5, GPT-3 and Dall-E though their RESTful API (currently in beta).\n\nIndependently developed, this is not an official library and I am not affiliated with OpenAI.\n\nAn OpenAI API account is required.",
+  "description": "A OpenAI package for the Unity to use though their RESTful API.\n\nIndependently developed, this is not an official library and I am not affiliated with OpenAI.\n\nAn OpenAI API account is required.",
   "keywords": [],
-  "version": "8.8.0",
+  "version": "8.8.1",
   "unity": "2021.3",
   "documentationUrl": "https://github.com/RageAgainstThePixel/com.openai.unity#documentation",
   "changelogUrl": "https://github.com/RageAgainstThePixel/com.openai.unity/releases",
@@ -17,9 +17,9 @@
     "url": "https://github.com/StephenHodgson"
   },
   "dependencies": {
-    "com.utilities.encoder.wav": "2.2.3",
+    "com.utilities.encoder.wav": "2.2.4",
     "com.utilities.rest": "3.3.8",
-    "com.utilities.websockets": "1.0.4"
+    "com.utilities.websockets": "1.0.5"
   },
   "samples": [
     {