Merge OpenAiStreamingChatModel and OpenAiChatModel classes

docs/docs/integrations/language-models/open-ai.md

@@ -71,16 +71,17 @@ OpenAiChatModel model = OpenAiChatModel.builder()
     .proxy(...)
     .logRequests(...)
     .logResponses(...)
+    .logStreamingResponses(...)
     .tokenizer(...)
     .customHeaders(...)
     .build();
 ```
 See the description of some of the parameters above [here](https://platform.openai.com/docs/api-reference/chat/create).
 
-## OpenAiStreamingChatModel
+## Streaming ChatModel
 
 ```java
-OpenAiStreamingChatModel model = OpenAiStreamingChatModel.withApiKey(System.getenv("OPENAI_API_KEY"));
+OpenAiChatModel model = OpenAiChatModel.withApiKey(System.getenv("OPENAI_API_KEY"), true);
 
 model.generate("Say 'Hello World'", new StreamingResponseHandler<AiMessage>() {
 

langchain4j-ollama/src/test/java/dev/langchain4j/model/ollama/OllamaOpenAiStreamingChatModelIT.java

@@ -4,7 +4,7 @@
 import dev.langchain4j.data.message.UserMessage;
 import dev.langchain4j.model.chat.StreamingChatLanguageModel;
 import dev.langchain4j.model.chat.TestStreamingResponseHandler;
-import dev.langchain4j.model.openai.OpenAiStreamingChatModel;
+import dev.langchain4j.model.openai.OpenAiChatModel;
 import dev.langchain4j.model.output.Response;
 import org.junit.jupiter.api.Test;
 
@@ -18,7 +18,7 @@
  */
 class OllamaOpenAiStreamingChatModelIT extends AbstractOllamaLanguageModelInfrastructure {
 
-    StreamingChatLanguageModel model = OpenAiStreamingChatModel.builder()
+    StreamingChatLanguageModel model = OpenAiChatModel.builder()
             .apiKey("does not matter") // TODO make apiKey optional when using custom baseUrl?
             .baseUrl(ollama.getEndpoint() + "/v1") // TODO add "/v1" by default?
             .modelName(TINY_DOLPHIN_MODEL)

langchain4j-open-ai/src/main/java/dev/langchain4j/model/openai/OpenAiChatModel.java

@@ -1,14 +1,18 @@
 package dev.langchain4j.model.openai;
 
 import dev.ai4j.openai4j.OpenAiClient;
+import dev.ai4j.openai4j.chat.ChatCompletionChoice;
 import dev.ai4j.openai4j.OpenAiHttpException;
 import dev.ai4j.openai4j.chat.ChatCompletionRequest;
 import dev.ai4j.openai4j.chat.ChatCompletionResponse;
+import dev.ai4j.openai4j.chat.Delta;
 import dev.langchain4j.agent.tool.ToolSpecification;
 import dev.langchain4j.data.message.AiMessage;
 import dev.langchain4j.data.message.ChatMessage;
+import dev.langchain4j.model.StreamingResponseHandler;
 import dev.langchain4j.model.Tokenizer;
 import dev.langchain4j.model.chat.ChatLanguageModel;
+import dev.langchain4j.model.chat.StreamingChatLanguageModel;
 import dev.langchain4j.model.chat.TokenCountEstimator;
 import dev.langchain4j.model.chat.listener.ChatLanguageModelRequest;
 import dev.langchain4j.model.chat.listener.ChatLanguageModelResponse;
@@ -26,6 +30,7 @@
 
 import static dev.langchain4j.internal.RetryUtils.withRetry;
 import static dev.langchain4j.internal.Utils.getOrDefault;
+import static dev.langchain4j.internal.Utils.isNullOrEmpty;
 import static dev.langchain4j.model.openai.InternalOpenAiHelper.*;
 import static dev.langchain4j.model.openai.OpenAiModelName.GPT_3_5_TURBO;
 import static dev.langchain4j.spi.ServiceHelper.loadFactories;
@@ -38,7 +43,7 @@
  * You can find description of parameters <a href="https://platform.openai.com/docs/api-reference/chat/create">here</a>.
  */
 @Slf4j
-public class OpenAiChatModel implements ChatLanguageModel, TokenCountEstimator {
+public class OpenAiChatModel implements ChatLanguageModel, StreamingChatLanguageModel, TokenCountEstimator {
 
     private final OpenAiClient client;
     private final String modelName;
@@ -54,6 +59,7 @@ public class OpenAiChatModel implements ChatLanguageModel, TokenCountEstimator {
     private final String user;
     private final Integer maxRetries;
     private final Tokenizer tokenizer;
+    private final boolean isOpenAiModel;
     private final List<ModelListener<ChatLanguageModelRequest, ChatLanguageModelResponse>> listeners;
 
     @Builder
@@ -98,10 +104,14 @@ public OpenAiChatModel(String baseUrl,
                 .proxy(proxy)
                 .logRequests(logRequests)
                 .logResponses(logResponses)
+                .logStreamingResponses(logResponses)
                 .userAgent(DEFAULT_USER_AGENT)
                 .customHeaders(customHeaders)
                 .build();
+
         this.modelName = getOrDefault(modelName, GPT_3_5_TURBO);
+        this.isOpenAiModel = isOpenAiModel(modelName);
+
         this.temperature = getOrDefault(temperature, 0.7);
         this.topP = topP;
         this.stop = stop;
@@ -123,22 +133,22 @@ public String modelName() {
 
     @Override
     public Response<AiMessage> generate(List<ChatMessage> messages) {
-        return generate(messages, null, null);
+        return generateMessage(messages, null, null);
     }
 
     @Override
     public Response<AiMessage> generate(List<ChatMessage> messages, List<ToolSpecification> toolSpecifications) {
-        return generate(messages, toolSpecifications, null);
+        return generateMessage(messages, toolSpecifications, null);
     }
 
     @Override
     public Response<AiMessage> generate(List<ChatMessage> messages, ToolSpecification toolSpecification) {
-        return generate(messages, singletonList(toolSpecification), toolSpecification);
+        return generateMessage(messages, singletonList(toolSpecification), toolSpecification);
     }
 
-    private Response<AiMessage> generate(List<ChatMessage> messages,
-                                         List<ToolSpecification> toolSpecifications,
-                                         ToolSpecification toolThatMustBeExecuted
+    private Response<AiMessage> generateMessage(List<ChatMessage> messages,
+                                                List<ToolSpecification> toolSpecifications,
+                                                ToolSpecification toolThatMustBeExecuted
     ) {
         ChatCompletionRequest.Builder requestBuilder = ChatCompletionRequest.builder()
                 .model(modelName)
@@ -248,4 +258,92 @@ public OpenAiChatModelBuilder modelName(OpenAiChatModelName modelName) {
             return this;
         }
     }
+
+    @Override
+    public void generate(List<ChatMessage> messages, StreamingResponseHandler<AiMessage> handler) {
+        generate(messages, null, null, handler);
+    }
+
+    @Override
+    public void generate(List<ChatMessage> messages, List<ToolSpecification> toolSpecifications, StreamingResponseHandler<AiMessage> handler) {
+        generate(messages, toolSpecifications, null, handler);
+    }
+
+    @Override
+    public void generate(List<ChatMessage> messages, ToolSpecification toolSpecification, StreamingResponseHandler<AiMessage> handler) {
+        generate(messages, null, toolSpecification, handler);
+    }
+
+    private void generate(List<ChatMessage> messages,
+                          List<ToolSpecification> toolSpecifications,
+                          ToolSpecification toolThatMustBeExecuted,
+                          StreamingResponseHandler<AiMessage> handler
+    ) {
+        ChatCompletionRequest.Builder requestBuilder = ChatCompletionRequest.builder()
+                .stream(true)
+                .model(modelName)
+                .messages(toOpenAiMessages(messages))
+                .temperature(temperature)
+                .topP(topP)
+                .stop(stop)
+                .maxTokens(maxTokens)
+                .presencePenalty(presencePenalty)
+                .frequencyPenalty(frequencyPenalty)
+                .logitBias(logitBias)
+                .responseFormat(responseFormat)
+                .seed(seed)
+                .user(user);
+
+        if (toolThatMustBeExecuted != null) {
+            requestBuilder.tools(toTools(singletonList(toolThatMustBeExecuted)));
+            requestBuilder.toolChoice(toolThatMustBeExecuted.name());
+        } else if (!isNullOrEmpty(toolSpecifications)) {
+            requestBuilder.tools(toTools(toolSpecifications));
+        }
+
+        ChatCompletionRequest request = requestBuilder.build();
+
+        int inputTokenCount = countInputTokens(messages, toolSpecifications, toolThatMustBeExecuted);
+        OpenAiStreamingResponseBuilder responseBuilder = new OpenAiStreamingResponseBuilder(inputTokenCount);
+
+        client.chatCompletion(request)
+                .onPartialResponse(partialResponse -> {
+                    responseBuilder.append(partialResponse);
+                    handle(partialResponse, handler);
+                })
+                .onComplete(() -> {
+                    Response<AiMessage> response = responseBuilder.build(tokenizer, toolThatMustBeExecuted != null);
+                    if (!isOpenAiModel) {
+                        response = removeTokenUsage(response);
+                    }
+                    handler.onComplete(response);
+                })
+                .onError(handler::onError)
+                .execute();
+    }
+
+    private int countInputTokens(List<ChatMessage> messages,
+                                 List<ToolSpecification> toolSpecifications,
+                                 ToolSpecification toolThatMustBeExecuted) {
+        int inputTokenCount = tokenizer.estimateTokenCountInMessages(messages);
+        if (toolThatMustBeExecuted != null) {
+            inputTokenCount += tokenizer.estimateTokenCountInForcefulToolSpecification(toolThatMustBeExecuted);
+        } else if (!isNullOrEmpty(toolSpecifications)) {
+            inputTokenCount += tokenizer.estimateTokenCountInToolSpecifications(toolSpecifications);
+        }
+        return inputTokenCount;
+    }
+
+    private static void handle(ChatCompletionResponse partialResponse,
+                               StreamingResponseHandler<AiMessage> handler) {
+        List<ChatCompletionChoice> choices = partialResponse.choices();
+        if (choices == null || choices.isEmpty()) {
+            return;
+        }
+        Delta delta = choices.get(0).delta();
+        String content = delta.content();
+        if (content != null) {
+            handler.onNext(content);
+        }
+    }
 }

langchain4j-open-ai/src/main/java/dev/langchain4j/model/openai/OpenAiStreamingChatModel.java

@@ -39,8 +39,12 @@
  * Represents an OpenAI language model with a chat completion interface, such as gpt-3.5-turbo and gpt-4.
  * The model's response is streamed token by token and should be handled with {@link StreamingResponseHandler}.
  * You can find description of parameters <a href="https://platform.openai.com/docs/api-reference/chat/create">here</a>.
+ *
+ *
+ * @Deprecated use {@link OpenAiChatModel} instead.
  */
 @Slf4j
+@Deprecated()
 public class OpenAiStreamingChatModel implements StreamingChatLanguageModel, TokenCountEstimator {
 
     private final OpenAiClient client;

langchain4j-open-ai/src/main/java/dev/langchain4j/model/openai/spi/OpenAiStreamingChatModelBuilderFactory.java

@@ -7,5 +7,6 @@
 /**
  * A factory for building {@link OpenAiStreamingChatModel.OpenAiStreamingChatModelBuilder} instances.
  */
+@Deprecated
 public interface OpenAiStreamingChatModelBuilderFactory extends Supplier<OpenAiStreamingChatModel.OpenAiStreamingChatModelBuilder> {
 }

langchain4j/src/test/java/dev/langchain4j/service/StreamingAiServicesIT.java

@@ -11,8 +11,8 @@
 import dev.langchain4j.memory.chat.MessageWindowChatMemory;
 import dev.langchain4j.model.azure.AzureOpenAiStreamingChatModel;
 import dev.langchain4j.model.chat.StreamingChatLanguageModel;
+import dev.langchain4j.model.openai.OpenAiChatModel;
 import dev.langchain4j.model.mistralai.MistralAiStreamingChatModel;
-import dev.langchain4j.model.openai.OpenAiStreamingChatModel;
 import dev.langchain4j.model.output.Response;
 import dev.langchain4j.model.output.TokenUsage;
 import org.junit.jupiter.api.Test;
@@ -33,7 +33,7 @@ public class StreamingAiServicesIT {
 
     static Stream<StreamingChatLanguageModel> models() {
         return Stream.of(
-                OpenAiStreamingChatModel.builder()
+                OpenAiChatModel.builder()
                         .baseUrl(System.getenv("OPENAI_BASE_URL"))
                         .apiKey(System.getenv("OPENAI_API_KEY"))
                         .organizationId(System.getenv("OPENAI_ORGANIZATION_ID"))
@@ -238,7 +238,7 @@ void should_execute_a_tool_then_stream_answer(StreamingChatLanguageModel model)
     void should_execute_multiple_tools_sequentially_then_answer() throws Exception {
 
         // TODO test more models
-        StreamingChatLanguageModel streamingChatModel = OpenAiStreamingChatModel.builder()
+        StreamingChatLanguageModel streamingChatModel = OpenAiChatModel.builder()
                 .baseUrl(System.getenv("OPENAI_BASE_URL"))
                 .apiKey(System.getenv("OPENAI_API_KEY"))
                 .organizationId(System.getenv("OPENAI_ORGANIZATION_ID"))
@@ -337,7 +337,7 @@ void should_execute_multiple_tools_in_parallel_then_answer() throws Exception {
         Calculator calculator = spy(new Calculator());
 
         // TODO test more models
-        StreamingChatLanguageModel streamingChatModel = OpenAiStreamingChatModel.builder()
+        StreamingChatLanguageModel streamingChatModel = OpenAiChatModel.builder()
                 .baseUrl(System.getenv("OPENAI_BASE_URL"))
                 .apiKey(System.getenv("OPENAI_API_KEY"))
                 .organizationId(System.getenv("OPENAI_ORGANIZATION_ID"))

langchain4j/src/test/java/dev/langchain4j/service/StreamingAiServicesWithToolsIT.java

@@ -9,7 +9,7 @@
 import dev.langchain4j.memory.chat.MessageWindowChatMemory;
 import dev.langchain4j.model.chat.StreamingChatLanguageModel;
 import dev.langchain4j.model.mistralai.MistralAiStreamingChatModel;
-import dev.langchain4j.model.openai.OpenAiStreamingChatModel;
+import dev.langchain4j.model.openai.OpenAiChatModel;
 import dev.langchain4j.model.output.Response;
 import org.junit.jupiter.params.ParameterizedTest;
 import org.junit.jupiter.params.provider.MethodSource;
@@ -32,7 +32,7 @@ class StreamingAiServicesWithToolsIT {
 
     static Stream<StreamingChatLanguageModel> models() {
         return Stream.of(
-                OpenAiStreamingChatModel.builder()
+                OpenAiChatModel.builder()
                         .baseUrl(System.getenv("OPENAI_BASE_URL"))
                         .apiKey(System.getenv("OPENAI_API_KEY"))
                         .organizationId(System.getenv("OPENAI_ORGANIZATION_ID"))