Merge pull request #1813 from quarkiverse/#1812

Disable TopK in Anthropic module when thinking is enabled

Author: geoand

model-providers/anthropic/runtime/src/main/java/io/quarkiverse/langchain4j/anthropic/runtime/AnthropicRecorder.java

@@ -5,6 +5,8 @@
 import java.time.Duration;
 import java.util.function.Supplier;
 
+import org.jboss.logging.Logger;
+
 import dev.langchain4j.model.anthropic.AnthropicChatModel;
 import dev.langchain4j.model.anthropic.AnthropicStreamingChatModel;
 import dev.langchain4j.model.chat.ChatModel;
@@ -20,6 +22,8 @@
 
 @Recorder
 public class AnthropicRecorder {
+    private static final Logger LOG = Logger.getLogger(AnthropicRecorder.class);
+
     private static final String DUMMY_KEY = "dummy";
 
     private final RuntimeValue<LangChain4jAnthropicConfig> runtimeConfig;
@@ -47,14 +51,15 @@ public Supplier<ChatModel> chatModel(String configName) {
                     .logRequests(firstOrDefault(false, chatModelConfig.logRequests(), anthropicConfig.logRequests()))
                     .logResponses(firstOrDefault(false, chatModelConfig.logResponses(), anthropicConfig.logResponses()))
                     .timeout(anthropicConfig.timeout().orElse(Duration.ofSeconds(10)))
-                    .topK(chatModelConfig.topK())
                     .maxTokens(chatModelConfig.maxTokens())
                     .maxRetries(chatModelConfig.maxRetries());
 
             if (chatModelConfig.temperature().isPresent()) {
                 builder.temperature(chatModelConfig.temperature().getAsDouble());
             }
 
+            builder.topK(chatModelConfig.topK().orElse(40));
+
             if (chatModelConfig.topP().isPresent()) {
                 builder.topP(chatModelConfig.topP().getAsDouble());
             }
@@ -65,6 +70,10 @@ public Supplier<ChatModel> chatModel(String configName) {
 
             ChatModelConfig.ThinkingConfig thinkingConfig = chatModelConfig.thinking();
             if (thinkingConfig.type().isPresent()) {
+                if (chatModelConfig.topK().isPresent()) {
+                    LOG.warn("TopK was not configured because thinking was enabled");
+                }
+                builder.topK(null);
                 builder.thinkingType(thinkingConfig.type().get());
             }
 
@@ -115,7 +124,7 @@ public Supplier<StreamingChatModel> streamingChatModel(String configName) {
                     .logRequests(firstOrDefault(false, chatModelConfig.logRequests(), anthropicConfig.logRequests()))
                     .logResponses(firstOrDefault(false, chatModelConfig.logResponses(), anthropicConfig.logResponses()))
                     .timeout(anthropicConfig.timeout().orElse(Duration.ofSeconds(10)))
-                    .topK(chatModelConfig.topK())
+                    .topK(chatModelConfig.topK().orElse(40))
                     .maxTokens(chatModelConfig.maxTokens());
 
             if (chatModelConfig.temperature().isPresent()) {

model-providers/anthropic/runtime/src/main/java/io/quarkiverse/langchain4j/anthropic/runtime/config/ChatModelConfig.java

@@ -3,6 +3,7 @@
 import java.util.List;
 import java.util.Optional;
 import java.util.OptionalDouble;
+import java.util.OptionalInt;
 
 import io.quarkus.runtime.annotations.ConfigDocDefault;
 import io.quarkus.runtime.annotations.ConfigGroup;
@@ -47,8 +48,8 @@ public interface ChatModelConfig {
      * Reduces the probability of generating nonsense. A higher value (e.g. 100) will give more diverse answers, while a lower
      * value (e.g. 10) will be more conservative
      */
-    @WithDefault("40")
-    Integer topK();
+    @ConfigDocDefault("40")
+    OptionalInt topK();
 
     /**
      * The maximum number of times to retry. 1 means exactly one attempt, with retrying disabled.