Add OpenAi Chat Options (2)

 - Add OpenAiChatOptions that implements ChatOptions and exposes all OpenAi request options, except messages and stream.
 - Add OpenAiChatOptions field (as defaultOptions) to OpenAiChatClient. Implement star-up/runtime options merging on chat request creation - Add OpenAiChatOptions options field to OpenAiChatProperties. Later is set as OpenAiChatClient#defaultOptions. Use the spring.ai.openai.chat.options.* prefix to set the options.
 - Add tests for properties and options merging.

 Part of #228

Author: tzolov

models/spring-ai-openai/pom.xml

@@ -34,6 +34,12 @@
 			<version>2.0.4</version>
 		</dependency>
 
+		<!-- NOTE: Required only by the @ConstructorBinding. -->
+		<dependency>
+			<groupId>org.springframework.boot</groupId>
+			<artifactId>spring-boot</artifactId>
+		</dependency>
+
 		<dependency>
 			<groupId>io.rest-assured</groupId>
 			<artifactId>json-path</artifactId>

models/spring-ai-openai/src/main/java/org/springframework/ai/openai/OpenAiChatClient.java

@@ -28,14 +28,16 @@
 import org.springframework.ai.chat.ChatResponse;
 import org.springframework.ai.chat.Generation;
 import org.springframework.ai.chat.StreamingChatClient;
-import org.springframework.ai.chat.messages.Message;
 import org.springframework.ai.chat.metadata.ChatGenerationMetadata;
 import org.springframework.ai.chat.metadata.RateLimit;
 import org.springframework.ai.chat.prompt.Prompt;
+import org.springframework.ai.model.ModelOptionsUtils;
 import org.springframework.ai.openai.api.OpenAiApi;
 import org.springframework.ai.openai.api.OpenAiApi.ChatCompletion;
 import org.springframework.ai.openai.api.OpenAiApi.ChatCompletionMessage;
+import org.springframework.ai.openai.api.OpenAiApi.ChatCompletionRequest;
 import org.springframework.ai.openai.api.OpenAiApi.OpenAiApiException;
+import org.springframework.ai.openai.api.OpenAiChatOptions;
 import org.springframework.ai.openai.metadata.OpenAiChatResponseMetadata;
 import org.springframework.ai.openai.metadata.support.OpenAiResponseHeaderExtractor;
 import org.springframework.http.ResponseEntity;
@@ -57,11 +59,15 @@
  */
 public class OpenAiChatClient implements ChatClient, StreamingChatClient {
 
-	private Double temperature = 0.7;
+	private final Logger logger = LoggerFactory.getLogger(getClass());
 
-	private String model = "gpt-3.5-turbo";
+	private static final List<String> REQUEST_JSON_FIELD_NAMES = ModelOptionsUtils
+		.getJsonPropertyValues(ChatCompletionRequest.class);
 
-	private final Logger logger = LoggerFactory.getLogger(getClass());
+	private OpenAiChatOptions defaultOptions = OpenAiChatOptions.builder()
+		.withModel("gpt-3.5-turbo")
+		.withTemperature(0.7f)
+		.build();
 
 	public final RetryTemplate retryTemplate = RetryTemplate.builder()
 		.maxAttempts(10)
@@ -76,40 +82,23 @@ public OpenAiChatClient(OpenAiApi openAiApi) {
 		this.openAiApi = openAiApi;
 	}
 
-	public String getModel() {
-		return this.model;
-	}
-
-	public void setModel(String model) {
-		this.model = model;
-	}
-
-	public Double getTemperature() {
-		return this.temperature;
-	}
-
-	public void setTemperature(Double temperature) {
-		this.temperature = temperature;
+	public OpenAiChatClient withDefaultOptions(OpenAiChatOptions options) {
+		this.defaultOptions = options;
+		return this;
 	}
 
 	@Override
 	public ChatResponse call(Prompt prompt) {
 
 		return this.retryTemplate.execute(ctx -> {
-			List<Message> messages = prompt.getInstructions();
 
-			List<ChatCompletionMessage> chatCompletionMessages = messages.stream()
-				.map(m -> new ChatCompletionMessage(m.getContent(),
-						ChatCompletionMessage.Role.valueOf(m.getMessageType().name())))
-				.toList();
+			ChatCompletionRequest request = createRequest(prompt, false);
 
-			ResponseEntity<ChatCompletion> completionEntity = this.openAiApi
-				.chatCompletionEntity(new OpenAiApi.ChatCompletionRequest(chatCompletionMessages, this.model,
-						this.temperature.floatValue()));
+			ResponseEntity<ChatCompletion> completionEntity = this.openAiApi.chatCompletionEntity(request);
 
 			var chatCompletion = completionEntity.getBody();
 			if (chatCompletion == null) {
-				logger.warn("No chat completion returned for request: {}", chatCompletionMessages);
+				logger.warn("No chat completion returned for request: {}", prompt);
 				return new ChatResponse(List.of());
 			}
 
@@ -128,16 +117,9 @@ public ChatResponse call(Prompt prompt) {
 	@Override
 	public Flux<ChatResponse> stream(Prompt prompt) {
 		return this.retryTemplate.execute(ctx -> {
-			List<Message> messages = prompt.getInstructions();
-
-			List<ChatCompletionMessage> chatCompletionMessages = messages.stream()
-				.map(m -> new ChatCompletionMessage(m.getContent(),
-						ChatCompletionMessage.Role.valueOf(m.getMessageType().name())))
-				.toList();
+			ChatCompletionRequest request = createRequest(prompt, true);
 
-			Flux<OpenAiApi.ChatCompletionChunk> completionChunks = this.openAiApi
-				.chatCompletionStream(new OpenAiApi.ChatCompletionRequest(chatCompletionMessages, this.model,
-						this.temperature.floatValue(), true));
+			Flux<OpenAiApi.ChatCompletionChunk> completionChunks = this.openAiApi.chatCompletionStream(request);
 
 			// For chunked responses, only the first chunk contains the choice role.
 			// The rest of the chunks with same ID share the same role.
@@ -161,4 +143,36 @@ public Flux<ChatResponse> stream(Prompt prompt) {
 		});
 	}
 
+	/**
+	 * Accessible for testing.
+	 */
+	ChatCompletionRequest createRequest(Prompt prompt, boolean stream) {
+
+		List<ChatCompletionMessage> chatCompletionMessages = prompt.getInstructions()
+			.stream()
+			.map(m -> new ChatCompletionMessage(m.getContent(),
+					ChatCompletionMessage.Role.valueOf(m.getMessageType().name())))
+			.toList();
+
+		ChatCompletionRequest request = new ChatCompletionRequest(chatCompletionMessages, stream);
+
+		if (this.defaultOptions != null) {
+			request = ModelOptionsUtils.merge(request, this.defaultOptions, ChatCompletionRequest.class,
+					REQUEST_JSON_FIELD_NAMES);
+		}
+
+		if (prompt.getOptions() != null) {
+			if (prompt.getOptions() instanceof OpenAiChatOptions runtimeOptions) {
+				request = ModelOptionsUtils.merge(runtimeOptions, request, ChatCompletionRequest.class,
+						REQUEST_JSON_FIELD_NAMES);
+			}
+			else {
+				throw new IllegalArgumentException("Prompt options are not of type ChatCompletionRequest:"
+						+ prompt.getOptions().getClass().getSimpleName());
+			}
+		}
+
+		return request;
+	}
+
 }

models/spring-ai-openai/src/main/java/org/springframework/ai/openai/api/OpenAiApi.java

@@ -31,6 +31,7 @@
 import reactor.core.publisher.Flux;
 import reactor.core.publisher.Mono;
 
+import org.springframework.boot.context.properties.bind.ConstructorBinding;
 import org.springframework.core.ParameterizedTypeReference;
 import org.springframework.http.HttpHeaders;
 import org.springframework.http.MediaType;
@@ -183,6 +184,7 @@ public record FunctionTool(
 		 * Create a tool of type 'function' and the given function definition.
 		 * @param function function definition.
 		 */
+		@ConstructorBinding
 		public FunctionTool(Function function) {
 			this(Type.FUNCTION, function);
 		}
@@ -219,13 +221,14 @@ public record Function(
 			 * @param name tool function name.
 			 * @param jsonSchema tool function schema as json.
 			 */
+			@ConstructorBinding
 			public Function(String description, String name, String jsonSchema) {
 				this(description, name, parseJson(jsonSchema));
 			}
 		}
 	}
 
-	/**
+/**
 	 * Creates a model response for the given chat conversation.
 	 *
 	 * @param messages A list of messages comprising the conversation so far.
@@ -269,17 +272,17 @@ public Function(String description, String name, String jsonSchema) {
 	 *
 	 */
 	@JsonInclude(Include.NON_NULL)
-	public record ChatCompletionRequest(
+	public record ChatCompletionRequest (
 			@JsonProperty("messages") List<ChatCompletionMessage> messages,
 			@JsonProperty("model") String model,
 			@JsonProperty("frequency_penalty") Float frequencyPenalty,
-			@JsonProperty("logit_bias") Map<String, Object> logitBias,
+			@JsonProperty("logit_bias") Map<String, Integer> logitBias,
 			@JsonProperty("max_tokens") Integer maxTokens,
 			@JsonProperty("n") Integer n,
 			@JsonProperty("presence_penalty") Float presencePenalty,
 			@JsonProperty("response_format") ResponseFormat responseFormat,
 			@JsonProperty("seed") Integer seed,
-			@JsonProperty("stop") String stop,
+			@JsonProperty("stop") List<String> stop,
 			@JsonProperty("stream") Boolean stream,
 			@JsonProperty("temperature") Float temperature,
 			@JsonProperty("top_p") Float topP,
@@ -331,6 +334,20 @@ public ChatCompletionRequest(List<ChatCompletionMessage> messages, String model,
 					tools, toolChoice, null);
 		}
 
+				/**
+		 * Shortcut constructor for a chat completion request with the given messages, model, tools and tool choice.
+		 * Streaming is set to false, temperature to 0.8 and all other parameters are null.
+		 *
+		 * @param messages A list of messages comprising the conversation so far.
+		 * @param stream If set, partial message deltas will be sent.Tokens will be sent as data-only server-sent events
+		 * as they become available, with the stream terminated by a data: [DONE] message.
+		 */
+		public ChatCompletionRequest(List<ChatCompletionMessage> messages, Boolean stream) {
+			this(messages, null, null, null, null, null, null,
+					null, null, null, stream, null, null,
+					null, null, null);
+		}
+
 		/**
 		 * Specifies a tool the model should use. Use to force the model to call a specific function.
 		 *
@@ -346,6 +363,7 @@ public record ToolChoice(
 			 * Create a tool choice of type 'function' and name 'functionName'.
 			 * @param functionName Function name of the tool.
 			 */
+			@ConstructorBinding
 			public ToolChoice(String functionName) {
 				this("function", Map.of("name", functionName));
 			}