Finally got tool streaming working!

Signed-off-by: Travis Angevine <[email protected]>

sdk/ai/Azure.AI.Inference/src/Customized/ChatCompletionsOptions.cs

@@ -2,6 +2,7 @@
 // Licensed under the MIT License.
 
 using System;
+using System.Collections.Generic;
 using Azure.Core;
 
 namespace Azure.AI.Inference
@@ -32,6 +33,16 @@ public partial class ChatCompletionsOptions
         /// <summary> A value indicating whether chat completions should be streamed for this request. </summary>
         internal bool? InternalShouldStreamResponse { get; set; }
 
+        /// <summary>
+        /// The collection of context messages associated with this chat completions request.
+        /// Typical usage begins with a chat message for the System role that provides instructions for
+        /// the behavior of the assistant, followed by alternating messages between the User and
+        /// Assistant roles.
+        /// Please note <see cref="ChatRequestMessage"/> is the base class. According to the scenario, a derived class of the base class might need to be assigned here, or this property needs to be casted to one of the possible derived classes.
+        /// The available derived classes include <see cref="ChatRequestAssistantMessage"/>, <see cref="ChatRequestSystemMessage"/>, <see cref="ChatRequestToolMessage"/> and <see cref="ChatRequestUserMessage"/>.
+        /// </summary>
+        public IList<ChatRequestMessage> Messages { get; set; }
+
         /// <summary> Initializes a new instance of <see cref="ChatCompletionsOptions"/>. </summary>
         public ChatCompletionsOptions()
         {

sdk/ai/Azure.AI.Inference/src/Customized/ChatCompletionsToolCall.cs

@@ -1,6 +1,8 @@
 // Copyright (c) Microsoft Corporation. All rights reserved.
 // Licensed under the MIT License.
 
+using System;
+
 namespace Azure.AI.Inference
 {
     public partial class ChatCompletionsToolCall
@@ -22,5 +24,30 @@ public string Arguments
             get => Function.Arguments;
             set => Function.Arguments = value;
         }
+
+        /// <summary> Creates a new <see cref="ChatCompletionsToolCall"/> representing a function call made by the model. </summary>
+        /// <param name="toolCallId"> The ID of the tool call. </param>
+        /// <param name="functionName"> The name of the function that model is calling. </param>
+        /// <param name="functionArguments">
+        ///     The arguments that model is calling the function with, which are generated by the model in JSON format.
+        ///     Note that the model does not always generate valid JSON, and may hallucinate parameters not defined by your
+        ///     function schema. Validate the arguments in your code before calling your function.
+        /// </param>
+        /// <exception cref="ArgumentNullException"> <paramref name="toolCallId"/>, <paramref name="functionName"/> or <paramref name="functionArguments"/> is null. </exception>
+        /// <exception cref="ArgumentException"> <paramref name="toolCallId"/>, <paramref name="functionName"/> or <paramref name="functionArguments"/> is an empty string, and was expected to be non-empty. </exception>
+        public static ChatCompletionsToolCall CreateFunctionToolCall(string toolCallId, string functionName, string functionArguments)
+        {
+            Argument.AssertNotNullOrEmpty(toolCallId, nameof(toolCallId));
+            Argument.AssertNotNullOrEmpty(functionName, nameof(functionName));
+            Argument.AssertNotNullOrEmpty(functionArguments, nameof(functionArguments));
+
+            FunctionCall function = new(functionName, functionArguments);
+
+            return new(
+                id: toolCallId,
+                type: ChatCompletionsToolCallType.Function,
+                function: function,
+                serializedAdditionalRawData: null);
+        }
     }
 }

sdk/ai/Azure.AI.Inference/src/Customized/StreamingChatChoiceUpdate.cs

@@ -0,0 +1,20 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+using System;
+using System.Collections.Generic;
+
+namespace Azure.AI.Inference
+{
+    /// <summary>
+    /// Represents an update to a single prompt completion when the service is streaming updates
+    /// using Server Sent Events (SSE).
+    /// Generally, `n` choices are generated per provided prompt with a default value of 1.
+    /// Token limits and other settings may limit the number of choices generated.
+    /// </summary>
+    public partial class StreamingChatChoiceUpdate
+    {
+        /// <summary> An update to the chat message for a given chat completions prompt. </summary>
+        public StreamingChatResponseMessageUpdate Delta { get; } = StreamingChatResponseMessageUpdate.Empty;
+    }
+}

sdk/ai/Azure.AI.Inference/src/Customized/StreamingChatCompletionsUpdate.Serialization.cs

@@ -25,7 +25,7 @@ internal static List<StreamingChatCompletionsUpdate> DeserializeStreamingChatCom
             string model = default;
             DateTimeOffset created = default;
             string systemFingerprint = null;
-            List<StreamingChoiceData> choiceEntries = new();
+            List<StreamingChatChoiceUpdate> choiceEntries = new();
 
             foreach (JsonProperty property in element.EnumerateObject())
             {
@@ -53,36 +53,33 @@ internal static List<StreamingChatCompletionsUpdate> DeserializeStreamingChatCom
                 {
                     foreach (JsonElement choiceElement in property.Value.EnumerateArray())
                     {
-                        choiceEntries.Add(StreamingChoiceData.DeserializeStreamingChoiceData(choiceElement));
+                        choiceEntries.Add(StreamingChatChoiceUpdate.DeserializeStreamingChatChoiceUpdate(choiceElement));
                     }
                 }
             }
 
             // If a chunk has no choices, we infer an empty one to aid traversal/expansion
             if (choiceEntries.Count == 0)
             {
-                choiceEntries.Add(StreamingChoiceData.Empty);
+                choiceEntries.Add(new StreamingChatChoiceUpdate());
             }
 
             // We inflate the possible combination of information into one StreamingChatCompletionsUpdate per tool
             // call update per choice, inferring an empty item in cases where the dimension isn't populated.
 
             List<StreamingChatCompletionsUpdate> results = new();
 
-            foreach (StreamingChoiceData choiceData in choiceEntries)
+            foreach (StreamingChatChoiceUpdate choiceData in choiceEntries)
             {
-                foreach (StreamingToolCallUpdate toolCallUpdate in choiceData.Delta.ToolCallUpdates)
+                foreach (StreamingChatResponseToolCallUpdate toolCallUpdate in choiceData.Delta.ToolCalls)
                 {
                     results.Add(new(
                         id,
                         model,
                         created,
                         choiceData.Delta.Role,
-                        choiceData.Delta.AuthorName,
-                        choiceData.Delta.ContentUpdate,
+                        choiceData.Delta.Content,
                         choiceData.FinishReason,
-                        choiceData.Delta.FunctionName,
-                        choiceData.Delta.FunctionArgumentsUpdate,
                         toolCallUpdate));
                 }
             }

sdk/ai/Azure.AI.Inference/src/Customized/StreamingChatCompletionsUpdate.cs

@@ -68,65 +68,19 @@ public partial class StreamingChatCompletionsUpdate
         /// </remarks>
         public string ContentUpdate { get; }
 
-        /// <summary>
-        /// Gets the name of a function to be called.
-        /// </summary>
-        /// <remarks>
-        /// Corresponds to e.g. $.choices[0].delta.function_call.name in the underlying REST schema.
-        /// </remarks>
-        public string FunctionName { get; }
-
-        /// <summary>
-        /// Gets a function arguments fragment associated with this update.
-        /// </summary>
-        /// <remarks>
-        /// <para>
-        /// Corresponds to e.g. $.choices[0].delta.function_call.arguments in the underlying REST schema.
-        /// </para>
-        ///
-        /// <para>
-        /// Each update contains only a small number of tokens. When presenting or reconstituting a full, streamed
-        /// arguments body, all <see cref="FunctionArgumentsUpdate"/> values for the same index should be combined.
-        /// </para>
-        ///
-        /// <para>
-        /// As is the case for non-streaming <see cref="FunctionCall.Arguments"/>, the content provided for function
-        /// arguments is not guaranteed to be well-formed JSON or to contain expected data. Callers should validate
-        /// function arguments before using them.
-        /// </para>
-        /// </remarks>
-        public string FunctionArgumentsUpdate { get; }
-
         /// <summary>
         /// An incremental update payload for a tool call that is part of this response.
         /// </summary>
         /// <remarks>
         /// <para>
-        /// Corresponds to e.g. $.choices[0].delta.tool_calls[0].index in the REST API schema.
+        /// Corresponds to e.g. $.choices[0].delta.tool_calls[0] in the REST API schema.
         /// </para>
         /// <para>
         /// To differentiate between parallel streaming tool calls within a single streaming choice, use the value of the
-        /// <see cref="StreamingToolCallUpdate.ToolCallIndex"/> property.
-        /// </para>
-        /// <para>
-        /// Please note <see cref="StreamingToolCallUpdate"/> is the base class. According to the scenario, a derived class
-        /// of the base class might need to be assigned here, or this property needs to be casted to one of the possible
-        /// derived classes.
-        /// The available derived classes include: <see cref="StreamingFunctionToolCallUpdate"/>.
-        /// </para>
-        /// </remarks>
-        public StreamingToolCallUpdate ToolCallUpdate { get; }
-
-        /// <summary>
-        /// Gets an optional name associated with the role of the streamed Chat Completion, typically as previously
-        /// specified in a system message.
-        /// </summary>
-        /// <remarks>
-        /// <para>
-        /// Corresponds to e.g. $.choices[0].delta.name in the underlying REST schema.
+        /// <see cref="StreamingChatResponseToolCallUpdate.Id"/> property.
         /// </para>
         /// </remarks>
-        public string AuthorName { get; }
+        public StreamingChatResponseToolCallUpdate ToolCallUpdate { get; }
 
         /// <summary>
         /// Gets the <see cref="CompletionsFinishReason"/> associated with this update.
@@ -147,22 +101,16 @@ internal StreamingChatCompletionsUpdate(
             string model,
             DateTimeOffset created,
             ChatRole? role = null,
-            string authorName = null,
             string contentUpdate = null,
             CompletionsFinishReason? finishReason = null,
-            string functionName = null,
-            string functionArgumentsUpdate = null,
-            StreamingToolCallUpdate toolCallUpdate = null)
+            StreamingChatResponseToolCallUpdate toolCallUpdate = null)
         {
             Id = id;
             Model = model;
             Created = created;
             Role = role;
-            AuthorName = authorName;
             ContentUpdate = contentUpdate;
             FinishReason = finishReason;
-            FunctionName = functionName;
-            FunctionArgumentsUpdate = functionArgumentsUpdate;
             ToolCallUpdate = toolCallUpdate;
         }
     }

sdk/ai/Azure.AI.Inference/src/Customized/StreamingChatResponseMessageUpdate.Serialization.cs

@@ -0,0 +1,72 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+using System;
+using System.ClientModel.Primitives;
+using System.Collections.Generic;
+using System.Text.Json;
+using Azure.Core;
+
+namespace Azure.AI.Inference
+{
+    public partial class StreamingChatResponseMessageUpdate : IUtf8JsonSerializable, IJsonModel<StreamingChatResponseMessageUpdate>
+    {
+        internal static StreamingChatResponseMessageUpdate DeserializeStreamingChatResponseMessageUpdate(JsonElement element, ModelReaderWriterOptions options = null)
+        {
+            options ??= ModelSerializationExtensions.WireOptions;
+
+            if (element.ValueKind == JsonValueKind.Null)
+            {
+                return null;
+            }
+            ChatRole? role = default;
+            string content = default;
+            List<StreamingChatResponseToolCallUpdate> toolCalls = new List<StreamingChatResponseToolCallUpdate>();
+            IDictionary<string, BinaryData> serializedAdditionalRawData = default;
+            Dictionary<string, BinaryData> rawDataDictionary = new Dictionary<string, BinaryData>();
+            foreach (var property in element.EnumerateObject())
+            {
+                if (property.NameEquals("role"u8))
+                {
+                    if (property.Value.ValueKind == JsonValueKind.Null)
+                    {
+                        continue;
+                    }
+                    role = new ChatRole(property.Value.GetString());
+                    continue;
+                }
+                if (property.NameEquals("content"u8))
+                {
+                    content = property.Value.GetString();
+                    continue;
+                }
+                if (property.NameEquals("tool_calls"u8))
+                {
+                    if (property.Value.ValueKind == JsonValueKind.Null)
+                    {
+                        continue;
+                    }
+                    List<StreamingChatResponseToolCallUpdate> array = new List<StreamingChatResponseToolCallUpdate>();
+                    foreach (var item in property.Value.EnumerateArray())
+                    {
+                        array.Add(StreamingChatResponseToolCallUpdate.DeserializeStreamingChatResponseToolCallUpdate(item, options));
+                    }
+                    toolCalls = array;
+                    continue;
+                }
+                if (options.Format != "W")
+                {
+                    rawDataDictionary.Add(property.Name, BinaryData.FromString(property.Value.GetRawText()));
+                }
+            }
+            serializedAdditionalRawData = rawDataDictionary;
+
+            if (toolCalls.Count == 0)
+            {
+                toolCalls.Add(null);
+            }
+
+            return new StreamingChatResponseMessageUpdate(role, content, toolCalls, serializedAdditionalRawData);
+        }
+    }
+}

sdk/ai/Azure.AI.Inference/src/Customized/StreamingChatResponseMessageUpdate.cs

@@ -0,0 +1,24 @@
+// Copyright (c) Microsoft Corporation. All rights reserved.
+// Licensed under the MIT License.
+
+using System;
+using System.Collections.Generic;
+using static Azure.AI.Inference.StreamingChatCompletionsUpdate;
+
+namespace Azure.AI.Inference
+{
+    /// <summary> A representation of a chat message update as received in a streaming response. </summary>
+    public partial class StreamingChatResponseMessageUpdate
+    {
+        /// <summary>
+        /// The tool calls that must be resolved and have their outputs appended to subsequent input messages for the chat
+        /// completions request to resolve as configured.
+        /// </summary>
+        public IReadOnlyList<StreamingChatResponseToolCallUpdate> ToolCalls { get; set; } = new List<StreamingChatResponseToolCallUpdate>();
+
+        internal static StreamingChatResponseMessageUpdate Empty { get; } = new()
+        {
+            ToolCalls = new List<StreamingChatResponseToolCallUpdate>() { null },
+        };
+    }
+}