[OpenAI] Whisper support

sdk/openai/azure-ai-openai/CHANGELOG.md

@@ -4,6 +4,10 @@
 
 ### Features Added
 
+- Support for `Whisper` endpoints was added.
+- Translation and Transcription of audio files is available
+- The above features are available both in Azure and non-Azure OpenAI
+
 ### Breaking Changes
 
 ### Bugs Fixed

sdk/openai/azure-ai-openai/assets.json

@@ -2,5 +2,5 @@
   "AssetsRepo": "Azure/azure-sdk-assets",
   "AssetsRepoPrefixPath": "java",
   "TagPrefix": "java/openai/azure-ai-openai",
-  "Tag": "java/openai/azure-ai-openai_57107e7a09"
+  "Tag": "java/openai/azure-ai-openai_298c639caa"
 }

sdk/openai/azure-ai-openai/src/main/java/com/azure/ai/openai/OpenAIServiceVersion.java

@@ -21,7 +21,10 @@ public enum OpenAIServiceVersion implements ServiceVersion {
     V2023_07_01_PREVIEW("2023-07-01-preview"),
 
     /** Enum value 2023-08-01-preview. */
-    V2023_08_01_PREVIEW("2023-08-01-preview");
+    V2023_08_01_PREVIEW("2023-08-01-preview"),
+
+    /** Enum value 2023-09-01-preview. */
+    V2023_09_01_PREVIEW("2023-09-01-preview");
 
     private final String version;
 
@@ -41,6 +44,6 @@ public String getVersion() {
      * @return The latest {@link OpenAIServiceVersion}.
      */
     public static OpenAIServiceVersion getLatest() {
-        return V2023_08_01_PREVIEW;
+        return V2023_09_01_PREVIEW;
     }
 }

sdk/openai/azure-ai-openai/src/main/java/com/azure/ai/openai/implementation/MultipartBoundaryGenerator.java

@@ -0,0 +1,17 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+package com.azure.ai.openai.implementation;
+
+/**
+ * Interface implemented by classes that would generate a boundary string to use in multipart type of requests.
+ * The main purpose of this class is to allow to mock behaviour for tests
+ */
+public interface MultipartBoundaryGenerator {
+
+    /**
+     * Generates a new multipart boundary value each time the method is called
+     * @return a {@link String} value containing a boundary to be used in HTTP multipart requests
+     */
+    String generateBoundary();
+}

sdk/openai/azure-ai-openai/src/main/java/com/azure/ai/openai/implementation/MultipartDataHelper.java

@@ -0,0 +1,203 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+package com.azure.ai.openai.implementation;
+
+import com.azure.ai.openai.models.AudioTranscriptionOptions;
+import com.azure.ai.openai.models.AudioTranslationOptions;
+import com.azure.core.util.BinaryData;
+
+import java.io.ByteArrayOutputStream;
+import java.io.IOException;
+import java.nio.charset.Charset;
+import java.nio.charset.StandardCharsets;
+import java.util.ArrayList;
+import java.util.List;
+import java.util.UUID;
+
+/**
+ * Helper class for marshaling {@link AudioTranscriptionOptions} and {@link AudioTranslationOptions} objects to be used
+ * in multipart HTTP requests according to RFC7578.
+ */
+public class MultipartDataHelper {
+
+    /**
+     * Value to be used as part of the divider for the multipart requests.
+     */
+    private final String boundary;
+
+    /**
+     * The actual part separator in the request. This is obtained by prepending "--" to the "boundary".
+     */
+    private final String partSeparator;
+
+    /**
+     * The marker for the ending of a multipart request. This is obtained by post-pending "--" to the "partSeparator".
+     */
+    private final String endMarker;
+
+    /**
+     * Charset used for encoding the multipart HTTP request.
+     */
+    private final Charset encoderCharset = StandardCharsets.UTF_8;
+
+    /**
+     * Line separator for the multipart HTTP request.
+     */
+    private static final String CRLF = "\r\n";
+
+    /**
+     * Default constructor used in the code. The boundary is a random value.
+     */
+    public MultipartDataHelper() {
+        this(() -> UUID.randomUUID().toString().substring(0, 16));
+    }
+
+    /**
+     * Constructor accepting a boundary generator. Used for testing.
+     * @param boundaryGenerator Generates the value for "boundary".
+     */
+    public MultipartDataHelper(MultipartBoundaryGenerator boundaryGenerator) {
+        this.boundary = boundaryGenerator.generateBoundary();
+        partSeparator = "--" + boundary;
+        endMarker = partSeparator + "--";
+    }
+
+    /**
+     *
+     * @return the "boundary" value.
+     */
+    public String getBoundary() {
+        return boundary;
+    }
+
+    /**
+     * This methods marshals the passed request into ready to be sent
+     * @param requestOptions object to be marshalled for the multipart HTTP request
+     * @param fileName the name of the file that is being sent as a part of this request
+     * @return the marshalled data and its length
+     * @param <T> {@link AudioTranscriptionOptions} and {@link AudioTranslationOptions} are the only types supported.
+     *           This represents the type information of the request object.
+     */
+    public <T> MultipartDataSerializationResult serializeRequest(T requestOptions, String fileName) {
+        if (requestOptions instanceof AudioTranslationOptions) {
+            AudioTranslationOptions audioTranslationOptions = (AudioTranslationOptions) requestOptions;
+            byte[] file = audioTranslationOptions.getFile();
+            List<MultipartField> fields = formatAudioTranslationOptions(audioTranslationOptions);
+            return serializeRequestFields(file, fields, fileName);
+        } else if (requestOptions instanceof AudioTranscriptionOptions) {
+            AudioTranscriptionOptions audioTranscriptionOptions = (AudioTranscriptionOptions) requestOptions;
+            byte[] file = audioTranscriptionOptions.getFile();
+            List<MultipartField> fields = formatAudioTranscriptionOptions(audioTranscriptionOptions);
+            return serializeRequestFields(file, fields, fileName);
+        } else {
+            throw new IllegalArgumentException("Only AudioTranslationOptions and AudioTranscriptionOptions currently supported");
+        }
+    }
+
+    /**
+     *
+     * @param file is the byte[] representation of the file in the request object.
+     * @param fields a list of the members other than the file in the request object.
+     * @param fileName the name of the file passed in the "file" field of the request object.
+     * @return a structure containing the marshalled data and its length.
+     */
+    private MultipartDataSerializationResult serializeRequestFields(byte[] file, List<MultipartField> fields, String fileName) {
+        ByteArrayOutputStream byteArrayOutputStream = new ByteArrayOutputStream();
+
+        // Multipart preamble
+        String fileFieldPreamble = partSeparator
+            + CRLF + "Content-Disposition: form-data; name=\"file\"; filename=\""
+            + fileName + "\""
+            + CRLF + "Content-Type: application/octet-stream" + CRLF + CRLF;
+        try {
+            // Writing the file into the request as a byte stream
+            byteArrayOutputStream.write(fileFieldPreamble.getBytes(encoderCharset));
+            byteArrayOutputStream.write(file);
+
+            // Adding other fields to the request
+            for (MultipartField field : fields) {
+                byteArrayOutputStream.write(serializeField(field));
+            }
+            byteArrayOutputStream.write((CRLF + endMarker).getBytes(encoderCharset));
+        } catch (IOException e) {
+            throw new RuntimeException(e);
+        }
+
+        byte[] totalData = byteArrayOutputStream.toByteArray();
+        return new MultipartDataSerializationResult(BinaryData.fromBytes(totalData), totalData.length);
+    }
+
+    /**
+     * Adds member fields apart from the file to the multipart HTTP request
+     * @param audioTranslationOptions request object
+     * @return a list of the fields in the request (except for "file")
+     */
+    private List<MultipartField> formatAudioTranslationOptions(AudioTranslationOptions audioTranslationOptions) {
+        List<MultipartField> fields = new ArrayList<>();
+        if (audioTranslationOptions.getResponseFormat() != null) {
+            fields.add(new MultipartField(
+                "response_format",
+                audioTranslationOptions.getResponseFormat().toString()));
+        }
+        if (audioTranslationOptions.getModel() != null) {
+            fields.add(new MultipartField("model",
+                    audioTranslationOptions.getModel()
+            ));
+        }
+        if (audioTranslationOptions.getPrompt() != null) {
+            fields.add(new MultipartField("prompt",
+                audioTranslationOptions.getPrompt()));
+        }
+        if (audioTranslationOptions.getTemperature() != null) {
+            fields.add(new MultipartField("temperature",
+                String.valueOf(audioTranslationOptions.getTemperature())));
+        }
+        return fields;
+    }
+
+    /**
+     * Adds member fields apart from the file to the multipart HTTP request
+     * @param audioTranscriptionOptions request object
+     * @return a list of the fields in the request (except for "file")
+     */
+    private List<MultipartField> formatAudioTranscriptionOptions(AudioTranscriptionOptions audioTranscriptionOptions) {
+        List<MultipartField> fields = new ArrayList<>();
+        if (audioTranscriptionOptions.getResponseFormat() != null) {
+            fields.add(new MultipartField("response_format",
+                audioTranscriptionOptions.getResponseFormat().toString()));
+        }
+        if (audioTranscriptionOptions.getModel() != null) {
+            fields.add(new MultipartField("model",
+                    audioTranscriptionOptions.getModel()
+            ));
+        }
+        if (audioTranscriptionOptions.getPrompt() != null) {
+            fields.add(new MultipartField("prompt",
+                audioTranscriptionOptions.getPrompt()));
+        }
+        if (audioTranscriptionOptions.getTemperature() != null) {
+            fields.add(new MultipartField("temperature",
+                String.valueOf(audioTranscriptionOptions.getTemperature())));
+        }
+        if (audioTranscriptionOptions.getLanguage() != null) {
+            fields.add(new MultipartField("language",
+                audioTranscriptionOptions.getLanguage()));
+        }
+        return fields;
+    }
+
+    /**
+     * This method formats a field for a multipart HTTP request and returns its byte[] representation
+     * @param field the field of the request to be marshalled
+     * @return byte[] representation of a field for a multipart HTTP request
+     */
+    private byte[] serializeField(MultipartField field) {
+        String serialized = CRLF + partSeparator
+            + CRLF + "Content-Disposition: form-data; name=\""
+            + field.getWireName() + "\"" + CRLF + CRLF
+            + field.getValue();
+
+        return serialized.getBytes(encoderCharset);
+    }
+}

sdk/openai/azure-ai-openai/src/main/java/com/azure/ai/openai/implementation/MultipartDataSerializationResult.java

@@ -0,0 +1,50 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+package com.azure.ai.openai.implementation;
+
+import com.azure.core.util.BinaryData;
+
+/**
+ * This class is used as a stand-in representation of marshalled data to be used in an HTTP multipart request.
+ */
+public class MultipartDataSerializationResult {
+
+    /**
+     * Represents the length of the content of this request. The value is to be used for the "Content-Length" header
+     * of the HTTP request
+     */
+    private final long dataLength;
+
+    /**
+     * The multipart form data of the request.
+     */
+    private final BinaryData data;
+
+    /**
+     * Constructor bundling both data and its length
+     * @param data the multipart form data of the request
+     * @param contentLength the length of the multipart form data of the request
+     */
+    public MultipartDataSerializationResult(BinaryData data, long contentLength) {
+        this.dataLength = contentLength;
+        this.data = data;
+    }
+
+    /**
+     *
+     * @return the result of marshaling a multipart HTTP request
+     */
+    public BinaryData getData() {
+        return data;
+    }
+
+    /**
+     *
+     * @return the length of a multipart HTTP request data
+     */
+    public long getDataLength() {
+        return dataLength;
+    }
+
+}

sdk/openai/azure-ai-openai/src/main/java/com/azure/ai/openai/implementation/MultipartField.java

@@ -0,0 +1,46 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+package com.azure.ai.openai.implementation;
+
+/**
+ * A field of a request for a multipart HTTP request.
+ */
+public class MultipartField {
+
+    /**
+     * The JSON key name of this field.
+     */
+    private final String wireName;
+
+    /**
+     * The JSON value of this field.
+     */
+    private final String value;
+
+    /**
+     *
+     * @param wireName The JSON key name of this field.
+     * @param value The JSON value of this field.
+     */
+    public MultipartField(String wireName, String value) {
+        this.wireName = wireName;
+        this.value = value;
+    }
+
+    /**
+     *
+     * @return The JSON key name of this field.
+     */
+    public String getWireName() {
+        return wireName;
+    }
+
+    /**
+     *
+     * @return The JSON value of this field.
+     */
+    public String getValue() {
+        return value;
+    }
+}