Refactor to static

mochi-neko · mochi-neko · commit 71b66fb70992 · 2023-06-30T14:43:13.000+09:00
diff --git a/Assets/Mochineko/Whisper_API.Samples/Mochineko.Whisper_API.Samples.asmdef b/Assets/Mochineko/Whisper_API.Samples/Mochineko.Whisper_API.Samples.asmdef
@@ -8,7 +8,7 @@
     "includePlatforms": [],
     "excludePlatforms": [],
     "allowUnsafeCode": false,
-    "overrideReferences": false,
+    "overrideReferences": true,
     "precompiledReferences": [],
     "autoReferenced": false,
     "defineConstraints": [],
diff --git a/Assets/Mochineko/Whisper_API.Samples/TranscriptionSample.cs b/Assets/Mochineko/Whisper_API.Samples/TranscriptionSample.cs
@@ -1,5 +1,6 @@
 #nullable enable
 using System;
+using System.Net.Http;
 using Cysharp.Threading.Tasks;
 using UnityEngine;
 using UnityEngine.Assertions;
@@ -21,16 +22,13 @@ public sealed class TranscriptionSample : MonoBehaviour
         /// </summary>
         [SerializeField] private string filePath = string.Empty;
 
-        private Transcription? connection;
-
+        private static readonly HttpClient httpClient = new();
+        
         private void Start()
         {
             // API Key must be set.
             Assert.IsNotNull(apiKey);
 
-            // Create instance of WhisperTranscriptionConnection.
-            connection = new Transcription(apiKey);
-            
             // If you want to specify response format, language, etc..., please use other initialization:
             // connection = new WhisperTranscriptionConnection(apiKey, new APIRequestBody(
             //     file: "",
@@ -45,12 +43,6 @@ private void Start()
         public async void Transcribe()
         {
             // Validations
-            if (connection == null)
-            {
-                Debug.LogError($"[Whisper_API.Transcription.Samples] Connection is null.");
-                return;
-            }
-
             if (string.IsNullOrEmpty(filePath))
             {
                 Debug.LogError($"[Whisper_API.Transcription.Samples] File path is null or empty.");
@@ -61,8 +53,13 @@ public async void Transcribe()
             try
             {
                 // Transcribe speech into text by Whisper transcription API.
-                result = await connection
-                    .TranscribeFromFileAsync(filePath, this.GetCancellationTokenOnDestroy());
+                result = await Transcription
+                    .TranscribeFromFileAsync(
+                        apiKey,
+                        httpClient,
+                        filePath, 
+                        Model.Whisper1,
+                        this.GetCancellationTokenOnDestroy());
             }
             catch (Exception e)
             {
diff --git a/Assets/Mochineko/Whisper_API.Tests/Mochineko.Whisper_API.Tests.asmdef b/Assets/Mochineko/Whisper_API.Tests/Mochineko.Whisper_API.Tests.asmdef
@@ -5,7 +5,8 @@
         "GUID:27619889b8ba8c24980f49ee34dbb44a",
         "GUID:0acc523941302664db1f4e527237feb3",
         "GUID:e372c541aba5148868e12aa078ca7c20",
-        "GUID:12e7ae99596e14eef8970a67eb588899"
+        "GUID:12e7ae99596e14eef8970a67eb588899",
+        "GUID:f51ebe6a0ceec4240a699833d6309b23"
     ],
     "includePlatforms": [
         "Editor"
diff --git a/Assets/Mochineko/Whisper_API.Tests/TranscriptionTest.cs b/Assets/Mochineko/Whisper_API.Tests/TranscriptionTest.cs
@@ -27,9 +27,15 @@ public async Task Transcribe()
                 Application.dataPath,
                 "Mochineko/Whisper_API.Tests/test.wav");
 
-            var connection = new Transcription(apiKey, Model.Whisper1);
-
-            var result = await connection.TranscribeFromFileAsync(filePath, CancellationToken.None);
+            using var httpClient = new System.Net.Http.HttpClient();
+
+            var result = await Transcription
+                .TranscribeFromFileAsync(
+                    apiKey,
+                    httpClient,
+                    filePath,
+                    Model.Whisper1,
+                    CancellationToken.None);
 
             string.IsNullOrEmpty(result).Should().BeFalse();
 
diff --git a/Assets/Mochineko/Whisper_API/Mochineko.Whisper_API.asmdef b/Assets/Mochineko/Whisper_API/Mochineko.Whisper_API.asmdef
@@ -1,13 +1,17 @@
 {
     "name": "Mochineko.Whisper_API",
     "rootNamespace": "",
-    "references": [],
+    "references": [
+        "GUID:f51ebe6a0ceec4240a699833d6309b23"
+    ],
     "includePlatforms": [],
     "excludePlatforms": [],
     "allowUnsafeCode": false,
-    "overrideReferences": false,
-    "precompiledReferences": [],
-    "autoReferenced": true,
+    "overrideReferences": true,
+    "precompiledReferences": [
+        "Newtonsoft.Json.dll"
+    ],
+    "autoReferenced": false,
     "defineConstraints": [],
     "versionDefines": [],
     "noEngineReferences": true
diff --git a/Assets/Mochineko/Whisper_API/Transcription.cs b/Assets/Mochineko/Whisper_API/Transcription.cs
@@ -1,84 +1,69 @@
 #nullable enable
 using System;
-using System.Collections.Generic;
 using System.IO;
 using System.Net.Http;
 using System.Threading;
-using System.Threading.Tasks;
-using Newtonsoft.Json;
+using Cysharp.Threading.Tasks;
 
 namespace Mochineko.Whisper_API
 {
     /// <summary>
     /// OpenAI Whisper transcription API.
     /// https://platform.openai.com/docs/api-reference/audio/create
     /// </summary>
-    public sealed class Transcription
+    public static class Transcription
     {
         private const string EndPoint = "https://api.openai.com/v1/audio/transcriptions";
 
-        private readonly IReadOnlyDictionary<string, string> headers;
-        private readonly TranscriptionRequestBody requestBody;
-
-        private static HttpClient HttpClient
-            => HttpClientPool.PooledClient;
-
         /// <summary>
-        /// Create an instance of Whisper transcription API connection.
+        /// Transcribes speech audio into text by Whisper transcription API.
         /// </summary>
-        /// <param name="apiKey">API key generated by OpenAI</param>
-        /// <param name="model">Speech recognition model</param>
-        /// <exception cref="ArgumentNullException">API Key must be set</exception>
-        public Transcription(string apiKey, Model model = Model.Whisper1)
+        public static async UniTask<string> TranscribeAsync(
+            string apiKey,
+            HttpClient httpClient,
+            Stream fileStream,
+            string fileName,
+            Model model,
+            CancellationToken cancellationToken,
+            string? prompt = null,
+            string? responseFormat = null,
+            float? temperature = null,
+            string? language = null
+        )
         {
-            if (string.IsNullOrEmpty(apiKey))
-            {
-                throw new ArgumentNullException(nameof(apiKey));
-            }
+            cancellationToken.ThrowIfCancellationRequested();
 
-            this.headers = CreateHeader(apiKey);
-            this.requestBody = new TranscriptionRequestBody(string.Empty, model);
-        }
-        
-        /// <summary>
-        /// Create an instance of Whisper transcription API connection.
-        /// </summary>
-        /// <param name="apiKey">API key generated by OpenAI</param>
-        /// <param name="requestBody">Request parameters</param>
-        /// <exception cref="ArgumentNullException">API Key must be set</exception>
-        public Transcription(string apiKey, TranscriptionRequestBody requestBody)
-        {
             if (string.IsNullOrEmpty(apiKey))
             {
                 throw new ArgumentNullException(nameof(apiKey));
             }
 
-            this.headers = CreateHeader(apiKey);
-            this.requestBody = requestBody;
-        }
-        
-        private static IReadOnlyDictionary<string, string> CreateHeader(string apiKey)
-            => new Dictionary<string, string>
+            if (string.IsNullOrEmpty(fileName))
             {
-                ["Authorization"] = $"Bearer {apiKey}",
-            };
+                throw new ArgumentNullException(nameof(fileName));
+            }
 
-        private static HttpRequestMessage CreateRequestMessage(
-            IReadOnlyDictionary<string, string> headers,
-            TranscriptionRequestBody requestBody,
-            Stream fileStream)
-        {
-            var requestMessage = new HttpRequestMessage(HttpMethod.Post, EndPoint);
-            foreach (var header in headers)
+            if (!TranscriptionRequestBody.IsAvailableFormat(fileName))
             {
-                requestMessage.Headers.Add(header.Key, header.Value);
+                throw new InvalidDataException(fileName);
             }
+            
+            var requestBody = new TranscriptionRequestBody(
+                fileName,
+                model,
+                prompt,
+                responseFormat,
+                temperature,
+                language);
+
+            using var requestMessage = new HttpRequestMessage(HttpMethod.Post, EndPoint);
+            requestMessage.Headers.Add("Authorization", $"Bearer {apiKey}");
 
             var requestContent = new MultipartFormDataContent();
 
             requestContent.Add(
                 content: new StringContent(
-                    content: $"{requestBody.Model}",
+                    content: requestBody.Model,
                     encoding: System.Text.Encoding.UTF8),
                 name: "model");
 
@@ -89,92 +74,9 @@ private static HttpRequestMessage CreateRequestMessage(
 
             requestMessage.Content = requestContent;
 
-            return requestMessage;
-        }
-
-        /// <summary>
-        /// Transcribes speech audio into text by Whisper transcription API.
-        /// </summary>
-        /// <param name="filePath"></param>
-        /// <param name="cancellationToken"></param>
-        /// <returns>Response text</returns>
-        /// <exception cref="Exception">System exceptions</exception>
-        /// <exception cref="APIErrorException">API error response</exception>
-        /// <exception cref="HttpRequestException">Network error</exception>
-        /// <exception cref="TaskCanceledException">Cancellation or timeout</exception>
-        /// <exception cref="JsonSerializationException">JSON error</exception>
-        /// <exception cref="ArgumentNullException">File path is null or empty</exception>
-        /// <exception cref="FileNotFoundException">Audio file not found</exception>
-        /// <exception cref="InvalidDataException">Not available audio format</exception>
-        /// <exception cref="OperationCanceledException">Cancelled on called</exception>
-        public async Task<string> TranscribeFromFileAsync(string filePath, CancellationToken cancellationToken)
-        {
-            cancellationToken.ThrowIfCancellationRequested();
-
-            if (string.IsNullOrEmpty(filePath))
-            {
-                throw new ArgumentNullException(nameof(filePath));
-            }
-
-            if (!File.Exists(filePath))
-            {
-                throw new FileNotFoundException(filePath);
-            }
-
-            if (!TranscriptionRequestBody.IsAvailableFormat(filePath))
-            {
-                throw new InvalidDataException(filePath);
-            }
-
-            await using var stream = File.OpenRead(filePath);
-
-            return await TranscribeAsync(
-                fileStream: stream,
-                fileName: Path.GetFileName(filePath),
-                cancellationToken);
-        }
-
-        /// <summary>
-        /// Transcribes speech audio into text by Whisper transcription API.
-        /// </summary>
-        /// <param name="fileStream">File data stream of speech audio</param>
-        /// <param name="fileName">File name of speech audio</param>
-        /// <param name="cancellationToken">Cancellation token</param>
-        /// <returns>Response text.</returns>
-        /// <exception cref="Exception">System exceptions</exception>
-        /// <exception cref="APIErrorException">API error response</exception>
-        /// <exception cref="HttpRequestException">Network error</exception>
-        /// <exception cref="TaskCanceledException">Cancellation or timeout</exception>
-        /// <exception cref="JsonSerializationException">JSON error</exception>
-        /// <exception cref="ArgumentNullException">Argument is null or empty</exception>
-        /// <exception cref="InvalidDataException">Not available audio format</exception>
-        /// <exception cref="OperationCanceledException">Cancelled on called</exception>
-        public async Task<string> TranscribeAsync(
-            Stream fileStream,
-            string fileName,
-            CancellationToken cancellationToken)
-        {
-            cancellationToken.ThrowIfCancellationRequested();
-
-            if (string.IsNullOrEmpty(fileName))
-            {
-                throw new ArgumentNullException(nameof(fileName));
-            }
-
-            if (!TranscriptionRequestBody.IsAvailableFormat(fileName))
-            {
-                throw new InvalidDataException(fileName);
-            }
-
-            requestBody.File = fileName;
-
-            using var requestMessage = CreateRequestMessage(
-                headers,
-                requestBody,
-                fileStream);
-
             // Post request and receive response
-            using var responseMessage = await HttpClient.SendAsync(requestMessage, cancellationToken);
+            using var responseMessage = await httpClient
+                .SendAsync(requestMessage, cancellationToken);
             if (responseMessage == null)
             {
                 throw new Exception($"[Whisper_API.Transcription] HttpResponseMessage is null.");
@@ -209,5 +111,46 @@ public async Task<string> TranscribeAsync(
                 }
             }
         }
+        
+        /// <summary>
+        /// Transcribes speech audio into text by Whisper transcription API.
+        /// </summary>
+        public static async UniTask<string> TranscribeFromFileAsync(
+            string apiKey,
+            HttpClient httpClient,
+            string filePath,
+            Model model,
+            CancellationToken cancellationToken,
+            string? prompt = null,
+            string? responseFormat = null,
+            float? temperature = null,
+            string? language = null)
+        {
+            cancellationToken.ThrowIfCancellationRequested();
+
+            if (string.IsNullOrEmpty(filePath))
+            {
+                throw new ArgumentNullException(nameof(filePath));
+            }
+
+            if (!File.Exists(filePath))
+            {
+                throw new FileNotFoundException(filePath);
+            }
+
+            await using var stream = File.OpenRead(filePath);
+
+            return await TranscribeAsync(
+                apiKey,
+                httpClient,
+                fileStream: stream,
+                fileName: Path.GetFileName(filePath),
+                model,
+                cancellationToken,
+                prompt,
+                responseFormat,
+                temperature,
+                language);
+        }
     }
 }
diff --git a/Assets/Mochineko/Whisper_API/TranscriptionRequestBody.cs b/Assets/Mochineko/Whisper_API/TranscriptionRequestBody.cs
