feat: Introduce OpenAiProviderDialect for flexible request shaping and remove reasoning properties from ModelProfile

- Added OpenAiProviderDialect to manage request features for OpenAI-compatible providers.
- Refactored ChatCompletionClient and TextCompletionClient to utilize the new dialect for handling reasoning and usage settings.
- Removed reasoningEffort and reasoningSummaryDetails from ModelProfile and related services.
- Updated README to reflect changes in request shaping and dialect usage.

README.md

@@ -250,8 +250,6 @@ Each profile file is a `.properties` resource on the classpath.
 | `cost1MCachedInTokens` | Double | No | Cost per 1 million cached input tokens. Defaults to `cost1MInTokens`. | `0.35`, `0` |
 | `cost1MOutTokens` | Double | Yes | Cost per 1 million output tokens. | `0.71`, `15.00`, `0` |
 | `cost1MReasoningOutTokens` | Double | No | Cost per 1 million reasoning output tokens. Defaults to `0`. | `0`, `60.00` |
-| `reasoningEffort` | String | No | Reasoning effort for reasoning-capable models. | `low`, `medium`, `high` |
-| `reasoningSummaryDetails` | String | No | How reasoning content should be summarized. | `auto`, `concise`, `detailed` |
 
 Notes:
 
@@ -260,7 +258,6 @@ Notes:
 - Cost properties are only used for statistics and accounting.
 - `cost1MCachedInTokens` defaults to the normal input token cost when not provided.
 - `cost1MReasoningOutTokens` defaults to `0` when not provided.
-- Reasoning properties only apply to models that expose separate reasoning output.
 - `tokenEncoding` is primarily relevant for tokenizer implementations that support multiple encodings, such as `jTokkit`.
 
 ### Example Profile Files
@@ -282,8 +279,6 @@ url=http://localhost:5001
 apiKey=
 cost1MInTokens=0
 cost1MOutTokens=0
-reasoningEffort=medium
-reasoningSummaryDetails=auto
 ```
 
 Local text completion:
@@ -322,6 +317,52 @@ Tool calling is currently supported only for chat completion models. `TextComple
 
 These built-in providers are OpenAI-compatible today, but that is an implementation detail of the current client set, not a hard architectural limit of the library. Additional provider integrations can be added in the library itself, and applications can plug in their own client implementations where needed.
 
+If you need to adapt request shaping for a specific OpenAI-compatible provider, you can optionally register an `OpenAiProviderDialect` bean. This allows applications to suppress request fields that a provider does not support while keeping `ModelProfile` provider-agnostic and runtime-driven.
+
+The default dialect behavior is:
+
+- `sendUsage(...) = true`
+- `sendReasoning(...) = true`
+- `sendMaxCompletionTokens(...) = true`
+- `reasoningEffort(...) = "medium"`
+- `reasoningSummaryDetails(...) = "concise"`
+
+Example:
+
+```java
+@Bean
+OpenAiProviderDialect azureOpenAiProviderDialect() {
+    return new OpenAiProviderDialect() {
+        @Override
+        public boolean sendUsage(final ModelProfile modelProfile) {
+            return false;
+        }
+
+        @Override
+        public boolean sendReasoning(final ModelProfile modelProfile) {
+            return false;
+        }
+
+        @Override
+        public boolean sendMaxCompletionTokens(final ModelProfile modelProfile) {
+            return false;
+        }
+
+        @Override
+        public String reasoningEffort(final ModelProfile modelProfile) {
+            return "medium";
+        }
+
+        @Override
+        public String reasoningSummaryDetails(final ModelProfile modelProfile) {
+            return "concise";
+        }
+    };
+}
+```
+
+If no `OpenAiProviderDialect` bean is present, all OpenAI-compatible request features remain enabled by default.
+
 For text completions, select a prompt strategy with the `prompts` property. lm-lib ships built-in strategies for common model families including Llama, GPT-OSS, Gemma, Mistral, Phi, Qwen, Vicuna, ChatML, Alpaca, and a no-formatting strategy.
 
 Example:

src/main/java/de/extio/lmlib/client/oai/OpenAiProviderDialect.java

@@ -0,0 +1,26 @@
+package de.extio.lmlib.client.oai;
+
+import de.extio.lmlib.profile.ModelProfile;
+
+public interface OpenAiProviderDialect {
+
+	default boolean sendUsage(final ModelProfile modelProfile) {
+		return true;
+	}
+
+	default boolean sendReasoning(final ModelProfile modelProfile) {
+		return true;
+	}
+
+	default boolean sendMaxCompletionTokens(final ModelProfile modelProfile) {
+		return true;
+	}
+
+	default String reasoningEffort(final ModelProfile modelProfile) {
+		return "medium";
+	}
+
+	default String reasoningSummaryDetails(final ModelProfile modelProfile) {
+		return "concise";
+	}
+}

src/main/java/de/extio/lmlib/client/oai/completion/AbstractCompletionClient.java

@@ -35,6 +35,7 @@ import de.extio.lmlib.client.Conversation;
 import de.extio.lmlib.client.ToolCallData;
 import de.extio.lmlib.client.oai.Model;
 import de.extio.lmlib.client.oai.ModelsResponse;
+import de.extio.lmlib.client.oai.OpenAiProviderDialect;
 import de.extio.lmlib.profile.ModelCategory;
 import de.extio.lmlib.profile.ModelProfile;
 import de.extio.lmlib.profile.ModelProfile.ModelProvider;
@@ -45,6 +46,9 @@ public abstract class AbstractCompletionClient implements Client, DisposableBean
 	
 	private static final Logger LOGGER = LoggerFactory.getLogger(AbstractCompletionClient.class);
 
+	private static final OpenAiProviderDialect DEFAULT_OPENAI_PROVIDER_DIALECT = new OpenAiProviderDialect() {
+	};
+	
 	@Autowired
 	@Qualifier("lmLibRestClientBuilder")
 	protected RestClient.Builder restClientBuilder;
@@ -55,6 +59,9 @@ public abstract class AbstractCompletionClient implements Client, DisposableBean
 	@Autowired
 	protected ModelProfileService modelProfileService;
 
+	@Autowired(required = false)
+	protected OpenAiProviderDialect openAiProviderDialect;
+	
 	@Value("${client.collectStatistics:false}")
 	protected boolean collectStatistics;
 	
@@ -137,6 +144,30 @@ public abstract class AbstractCompletionClient implements Client, DisposableBean
 	
 	protected abstract Completion requestCompletion(final Conversation conversation, final ModelProfile modelProfile, final Consumer<Chunk> chunkConsumer, final ToolCallData toolCallData);
 
+	protected OpenAiProviderDialect getOpenAiProviderDialect() {
+		return this.openAiProviderDialect == null ? DEFAULT_OPENAI_PROVIDER_DIALECT : this.openAiProviderDialect;
+	}
+
+	protected boolean sendUsage(final ModelProfile modelProfile) {
+		return this.getOpenAiProviderDialect().sendUsage(modelProfile);
+	}
+
+	protected boolean sendReasoning(final ModelProfile modelProfile) {
+		return this.getOpenAiProviderDialect().sendReasoning(modelProfile);
+	}
+
+	protected boolean sendMaxCompletionTokens(final ModelProfile modelProfile) {
+		return this.getOpenAiProviderDialect().sendMaxCompletionTokens(modelProfile);
+	}
+
+	protected String reasoningEffort(final ModelProfile modelProfile) {
+		return this.getOpenAiProviderDialect().reasoningEffort(modelProfile);
+	}
+
+	protected String reasoningSummaryDetails(final ModelProfile modelProfile) {
+		return this.getOpenAiProviderDialect().reasoningSummaryDetails(modelProfile);
+	}
+
 	protected String getPrimaryModelName(final ModelProfile modelProfile) {
 		return this.getModelNames(modelProfile, false).stream().findFirst().orElse("");
 	}

src/main/java/de/extio/lmlib/client/oai/completion/chat/ChatCompletionClient.java

@@ -48,45 +48,7 @@ public class ChatCompletionClient extends AbstractCompletionClient {
 	@Override
 	protected Completion requestCompletion(final Conversation conversation, final ModelProfile modelProfile, final Consumer<Chunk> chunkConsumer, final ToolCallData toolCallData) {
 		final var chat = createChats(conversation, modelProfile);
-		
+		final var request = this.createRequest(chat, modelProfile, chunkConsumer != null, toolCallData);
-		final var request = new ChatCompletionRequest();
-		request.setModel(this.getPrimaryModelName(modelProfile));
-		request.setMessages(chat);
-		request.setMaxTokens(modelProfile.maxTokens());
-		request.setMaxCompletionTokens(modelProfile.maxTokens());
-		request.setTemperature(modelProfile.temperature());
-		request.setTopP(modelProfile.topP());
-		request.setStream(chunkConsumer != null);
-		if (chunkConsumer != null) {
-			this.configureStreamOptions(true, request::setStreamOptions);
-		} else {
-			request.setUsage(true);
-		}
-		if (modelProfile.reasoningEffort() != null && !modelProfile.reasoningEffort().isBlank()) {
-			request.setReasoning(new ChatCompletionRequest.ChatCompletionsRequestReasoning());
-			try {
-				request.getReasoning().setEffort(ChatCompletionRequest.ChatCompletionsRequestReasoningEffort.valueOf(modelProfile.reasoningEffort()));
-			}
-			catch (final IllegalArgumentException e) {
-				throw new IllegalArgumentException("Invalid reasoning effort in the model profile: " + modelProfile.reasoningEffort(), e);
-			}
-		}
-		if (modelProfile.reasoningSummaryDetails() != null && !modelProfile.reasoningSummaryDetails().isBlank()) {
-			if (request.getReasoning() == null) {
-				request.setReasoning(new ChatCompletionRequest.ChatCompletionsRequestReasoning());
-			}
-			try {
-				request.getReasoning().setSummary(ChatCompletionRequest.ChatCompletionsRequestReasoningSummaryDetails.valueOf(modelProfile.reasoningSummaryDetails()).name());
-			}
-			catch (final IllegalArgumentException e) {
-				throw new IllegalArgumentException("Invalid reasoning summary details in the model profile: " + modelProfile.reasoningSummaryDetails(), e);
-			}
-		}
-		if (toolCallData != null && toolCallData.hasTools()) {
-			request.setTools(this.createTools(toolCallData.tools()));
-			request.setToolChoice(this.createToolChoice(toolCallData));
-			request.setParallelToolCalls(toolCallData.parallelToolCalls());
-		}
 		
 		String requestBody;
 		try {
@@ -203,6 +165,53 @@ public class ChatCompletionClient extends AbstractCompletionClient {
 		return new Completion(content, reasoning, finishReason, statistics, toolCalls);
 	}
 
+	ChatCompletionRequest createRequest(final List<ChatMessage> chat, final ModelProfile modelProfile, final boolean streaming, final ToolCallData toolCallData) {
+		final var request = new ChatCompletionRequest();
+		request.setModel(this.getPrimaryModelName(modelProfile));
+		request.setMessages(chat);
+		request.setMaxTokens(modelProfile.maxTokens());
+		if (this.sendMaxCompletionTokens(modelProfile)) {
+			request.setMaxCompletionTokens(modelProfile.maxTokens());
+		}
+		request.setTemperature(modelProfile.temperature());
+		request.setTopP(modelProfile.topP());
+		request.setStream(streaming);
+		if (streaming) {
+			this.configureStreamOptions(true, request::setStreamOptions);
+		}
+		else if (this.sendUsage(modelProfile)) {
+			request.setUsage(true);
+		}
+		final var reasoningEffort = this.reasoningEffort(modelProfile);
+		if (this.sendReasoning(modelProfile) && reasoningEffort != null && !reasoningEffort.isBlank()) {
+			request.setReasoning(new ChatCompletionRequest.ChatCompletionsRequestReasoning());
+			try {
+				request.getReasoning().setEffort(ChatCompletionRequest.ChatCompletionsRequestReasoningEffort.valueOf(reasoningEffort));
+			}
+			catch (final IllegalArgumentException e) {
+				throw new IllegalArgumentException("Invalid reasoning effort for the current OAI provider dialect: " + reasoningEffort, e);
+			}
+		}
+		final var reasoningSummaryDetails = this.reasoningSummaryDetails(modelProfile);
+		if (this.sendReasoning(modelProfile) && reasoningSummaryDetails != null && !reasoningSummaryDetails.isBlank()) {
+			if (request.getReasoning() == null) {
+				request.setReasoning(new ChatCompletionRequest.ChatCompletionsRequestReasoning());
+			}
+			try {
+				request.getReasoning().setSummary(ChatCompletionRequest.ChatCompletionsRequestReasoningSummaryDetails.valueOf(reasoningSummaryDetails).name());
+			}
+			catch (final IllegalArgumentException e) {
+				throw new IllegalArgumentException("Invalid reasoning summary details for the current OAI provider dialect: " + reasoningSummaryDetails, e);
+			}
+		}
+		if (toolCallData != null && toolCallData.hasTools()) {
+			request.setTools(this.createTools(toolCallData.tools()));
+			request.setToolChoice(this.createToolChoice(toolCallData));
+			request.setParallelToolCalls(toolCallData.parallelToolCalls());
+		}
+		return request;
+	}
+
 	private List<ChatCompletionRequest.ChatCompletionTool> createTools(final List<ToolDefinition> toolDefinitions) {
 		return toolDefinitions.stream().map(toolDefinition -> {
 			final var function = new ChatCompletionRequest.ChatCompletionFunction();

src/main/java/de/extio/lmlib/client/oai/completion/text/TextCompletionClient.java

@@ -53,19 +53,7 @@ public class TextCompletionClient extends AbstractCompletionClient {
 		}
 		
 		final var prompt = this.createPrompt(conversation, modelProfile, promptStrategy);
-		
+		final var request = this.createRequest(prompt, modelProfile, chunkConsumer != null);
-		final var request = new CompletionRequest();
-		request.setModel(this.getPrimaryModelName(modelProfile));
-		request.setPrompt(prompt);
-		request.setMaxTokens(modelProfile.maxTokens());
-		request.setTemperature(modelProfile.temperature());
-		request.setTopP(modelProfile.topP());
-		request.setStream(chunkConsumer != null);
-		if (chunkConsumer != null) {
-			this.configureStreamOptions(true, request::setStreamOptions);
-		} else {
-			request.setUsage(true);
-		}
 		
 		String requestBody;
 		try {
@@ -197,6 +185,23 @@ public class TextCompletionClient extends AbstractCompletionClient {
 		return new Completion(content, reasoning, finishReason, statistics);
 	}
 
+	CompletionRequest createRequest(final String prompt, final ModelProfile modelProfile, final boolean streaming) {
+		final var request = new CompletionRequest();
+		request.setModel(this.getPrimaryModelName(modelProfile));
+		request.setPrompt(prompt);
+		request.setMaxTokens(modelProfile.maxTokens());
+		request.setTemperature(modelProfile.temperature());
+		request.setTopP(modelProfile.topP());
+		request.setStream(streaming);
+		if (streaming) {
+			this.configureStreamOptions(true, request::setStreamOptions);
+		}
+		else if (this.sendUsage(modelProfile)) {
+			request.setUsage(true);
+		}
+		return request;
+	}
+	
 	private String createPrompt(final Conversation conversation, final ModelProfile modelProfile, final PromptStrategy promptStrategy) {
 		StringBuilder prompt = this.formatConversation(conversation, promptStrategy);
 		

src/main/java/de/extio/lmlib/profile/ModelProfile.java

@@ -18,8 +18,6 @@ public record ModelProfile(
 		BigDecimal costPerCachedInToken,
 		BigDecimal costPerOutToken,
 		BigDecimal costPerReasoningOutToken,
-		String reasoningEffort,
-		String reasoningSummaryDetails,
 		String category) {
 	
 	public ModelProfile {

src/main/java/de/extio/lmlib/profile/ModelProfileService.java

@@ -90,12 +90,6 @@ public class ModelProfileService {
 					.map(Object::toString)
 					.map(value -> new BigDecimal(value).divide(new BigDecimal(1000000.0)))
 					.orElse(BigDecimal.ZERO);
-			final var reasoningEffort = Optional.ofNullable(resource.getProperty("reasoningEffort"))
-					.map(Object::toString)
-					.orElse(null);
-			final var reasoningSummaryDetails = Optional.ofNullable(resource.getProperty("reasoningSummaryDetails"))
-					.map(Object::toString)
-					.orElse(null);
 			final var category = Optional
 					.ofNullable(modelCategory)
 					.map(ModelCategory::getShortName)
@@ -112,7 +106,7 @@ public class ModelProfileService {
 						.orElse(null);
 			}
 			
-			return new ModelProfile(promptTemplate, tokenizer, tokenEncoding, maxTokens, maxContextLength, temperature, topP, modelProvider, modelNameCfg, url, apiKey, costPerInToken, costPerCachedInToken, costPerOutToken, costPerReasoningOutToken, reasoningEffort, reasoningSummaryDetails, category);
+			return new ModelProfile(promptTemplate, tokenizer, tokenEncoding, maxTokens, maxContextLength, temperature, topP, modelProvider, modelNameCfg, url, apiKey, costPerInToken, costPerCachedInToken, costPerOutToken, costPerReasoningOutToken, category);
 		}
 		catch (final IOException e) {
 			LOGGER.error("Error while reading model profile", e);

src/test/java/de/extio/lmlib/client/oai/completion/chat/ChatCompletionClientRequestTest.java

@@ -0,0 +1,121 @@
+package de.extio.lmlib.client.oai.completion.chat;
+
+import static org.junit.jupiter.api.Assertions.assertEquals;
+import static org.junit.jupiter.api.Assertions.assertFalse;
+import static org.junit.jupiter.api.Assertions.assertNotNull;
+import static org.junit.jupiter.api.Assertions.assertNull;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+
+import java.math.BigDecimal;
+import java.util.List;
+
+import org.junit.jupiter.api.Test;
+
+import de.extio.lmlib.client.oai.OpenAiProviderDialect;
+import de.extio.lmlib.profile.ModelProfile;
+import de.extio.lmlib.profile.ModelProfile.ModelProvider;
+
+class ChatCompletionClientRequestTest {
+
+	@Test
+	void defaultsToAllRequestFieldsWhenNoDialectIsConfigured() {
+		final var client = new TestChatCompletionClient();
+		final var request = client.createRequest(
+				List.of(new ChatMessage("user", "hello", null)),
+				this.createProfile(),
+				false,
+				null);
+
+		assertTrue(request.getUsage());
+		assertNotNull(request.getReasoning());
+		assertEquals(ChatCompletionRequest.ChatCompletionsRequestReasoningEffort.medium, request.getReasoning().getEffort());
+		assertEquals("concise", request.getReasoning().getSummary());
+		assertEquals(2048, request.getMaxCompletionTokens());
+	}
+
+	@Test
+	void omitsDialectDisabledRequestFields() {
+		final var client = new TestChatCompletionClient();
+		client.setDialect(new OpenAiProviderDialect() {
+			@Override
+			public boolean sendUsage(final ModelProfile modelProfile) {
+				return false;
+			}
+
+			@Override
+			public boolean sendReasoning(final ModelProfile modelProfile) {
+				return false;
+			}
+
+			@Override
+			public boolean sendMaxCompletionTokens(final ModelProfile modelProfile) {
+				return false;
+			}
+		});
+
+		final var request = client.createRequest(
+				List.of(new ChatMessage("user", "hello", null)),
+				this.createProfile(),
+				false,
+				null);
+
+		assertNull(request.getUsage());
+		assertNull(request.getReasoning());
+		assertNull(request.getMaxCompletionTokens());
+		assertEquals(2048, request.getMaxTokens());
+		assertFalse(request.isStream());
+	}
+
+	@Test
+	void usesDialectProvidedReasoningDefaults() {
+		final var client = new TestChatCompletionClient();
+		client.setDialect(new OpenAiProviderDialect() {
+			@Override
+			public String reasoningEffort(final ModelProfile modelProfile) {
+				return "high";
+			}
+
+			@Override
+			public String reasoningSummaryDetails(final ModelProfile modelProfile) {
+				return "detailed";
+			}
+		});
+
+		final var request = client.createRequest(
+				List.of(new ChatMessage("user", "hello", null)),
+				this.createProfile(),
+				false,
+				null);
+
+		assertNotNull(request.getReasoning());
+		assertEquals(ChatCompletionRequest.ChatCompletionsRequestReasoningEffort.high, request.getReasoning().getEffort());
+		assertEquals("detailed", request.getReasoning().getSummary());
+	}
+
+	private ModelProfile createProfile() {
+		return new ModelProfile(
+				"",
+				"fallback",
+				"cl100k_base",
+				2048,
+				8192,
+				0.4,
+				1.0,
+				ModelProvider.OAI_CHAT_COMPLETION,
+				"test-model",
+				"http://localhost:5001",
+				"",
+				BigDecimal.ZERO,
+				BigDecimal.ZERO,
+				BigDecimal.ZERO,
+				BigDecimal.ZERO,
+				"TEST");
+	}
+
+	private static final class TestChatCompletionClient extends ChatCompletionClient {
+
+		private void setDialect(final OpenAiProviderDialect openAiProviderDialect) {
+			this.openAiProviderDialect = openAiProviderDialect;
+		}
+	}
+}

src/test/java/de/extio/lmlib/client/oai/completion/text/TextCompletionClientRequestTest.java

@@ -0,0 +1,65 @@
+package de.extio.lmlib.client.oai.completion.text;
+
+import static org.junit.jupiter.api.Assertions.assertNull;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+
+import java.math.BigDecimal;
+
+import org.junit.jupiter.api.Test;
+
+import de.extio.lmlib.client.oai.OpenAiProviderDialect;
+import de.extio.lmlib.profile.ModelProfile;
+import de.extio.lmlib.profile.ModelProfile.ModelProvider;
+
+class TextCompletionClientRequestTest {
+
+	@Test
+	void defaultsToUsageWhenNoDialectIsConfigured() {
+		final var client = new TestTextCompletionClient();
+		final var request = client.createRequest("prompt", this.createProfile(), false);
+
+		assertTrue(request.getUsage());
+	}
+
+	@Test
+	void omitsUsageWhenDialectDisablesIt() {
+		final var client = new TestTextCompletionClient();
+		client.setDialect(new OpenAiProviderDialect() {
+			@Override
+			public boolean sendUsage(final ModelProfile modelProfile) {
+				return false;
+			}
+		});
+
+		final var request = client.createRequest("prompt", this.createProfile(), false);
+
+		assertNull(request.getUsage());
+	}
+
+	private ModelProfile createProfile() {
+		return new ModelProfile(
+				"gpt-oss",
+				"fallback",
+				"cl100k_base",
+				2048,
+				8192,
+				0.4,
+				1.0,
+				ModelProvider.OAI_TEXT_COMPLETION,
+				"test-model",
+				"http://localhost:5001",
+				"",
+				BigDecimal.ZERO,
+				BigDecimal.ZERO,
+				BigDecimal.ZERO,
+				BigDecimal.ZERO,
+				"TEST");
+	}
+
+	private static final class TestTextCompletionClient extends TextCompletionClient {
+
+		private void setDialect(final OpenAiProviderDialect openAiProviderDialect) {
+			this.openAiProviderDialect = openAiProviderDialect;
+		}
+	}
+}

src/test/java/de/extio/lmlib/token/TokenizerResolverTest.java

@@ -55,8 +55,6 @@ public class TokenizerResolverTest {
 				BigDecimal.ZERO,
 				BigDecimal.ZERO,
 				BigDecimal.ZERO,
-				null,
-				null,
 				"TEST");
 	}
 	

src/test/resources/local-chatcompletion.properties

@@ -14,5 +14,3 @@ cost1MInTokens=0
 cost1MCachedInTokens=0
 cost1MOutTokens=0
 cost1MReasoningOutTokens=0
-reasoningEffort=medium
-reasoningSummaryDetails=auto