Update to LangChain4j 0.32.0

core/deployment/src/main/java/io/quarkiverse/langchain4j/deployment/AiServicesProcessor.java

@@ -31,7 +31,6 @@
 
 import jakarta.annotation.PreDestroy;
 import jakarta.enterprise.context.Dependent;
-import jakarta.enterprise.inject.Instance;
 import jakarta.inject.Inject;
 
 import org.jboss.jandex.AnnotationInstance;
@@ -132,7 +131,6 @@ public class AiServicesProcessor {
 
     private static final MethodDescriptor QUARKUS_AI_SERVICES_CONTEXT_REMOVE_CHAT_MEMORY_IDS = MethodDescriptor.ofMethod(
             QuarkusAiServiceContext.class, "removeChatMemoryIds", void.class, Object[].class);
-    public static final DotName CDI_INSTANCE = DotName.createSimple(Instance.class);
     private static final String[] EMPTY_STRING_ARRAY = new String[0];
     private static final String METRICS_DEFAULT_NAME = "langchain4j.aiservices";
 
@@ -494,7 +492,7 @@ public void handleDeclarativeServices(AiServicesRecorder recorder,
                     .equals(retrievalAugmentorSupplierClassName)) {
                 // Use a CDI bean of type `RetrievalAugmentor` if one exists, otherwise
                 // don't use an augmentor.
-                configurator.addInjectionPoint(ParameterizedType.create(CDI_INSTANCE,
+                configurator.addInjectionPoint(ParameterizedType.create(DotNames.CDI_INSTANCE,
                         new Type[] { ClassType.create(LangChain4jDotNames.RETRIEVAL_AUGMENTOR) }, null));
                 needsRetrievalAugmentorBean = true;
             } else {
@@ -513,7 +511,7 @@ public void handleDeclarativeServices(AiServicesRecorder recorder,
             }
 
             if (LangChain4jDotNames.BEAN_IF_EXISTS_AUDIT_SERVICE_SUPPLIER.toString().equals(auditServiceClassSupplierName)) {
-                configurator.addInjectionPoint(ParameterizedType.create(CDI_INSTANCE,
+                configurator.addInjectionPoint(ParameterizedType.create(DotNames.CDI_INSTANCE,
                         new Type[] { ClassType.create(LangChain4jDotNames.AUDIT_SERVICE) }, null));
                 needsAuditServiceBean = true;
             }

core/deployment/src/main/java/io/quarkiverse/langchain4j/deployment/DotNames.java

@@ -5,9 +5,13 @@
 import java.util.List;
 import java.util.Set;
 
+import jakarta.enterprise.inject.Instance;
+
 import org.jboss.jandex.DotName;
 
-class DotNames {
+import dev.langchain4j.model.chat.listener.ChatModelListener;
+
+public class DotNames {
 
     public static final DotName BOOLEAN = DotName.createSimple(Boolean.class);
     public static final DotName PRIMITIVE_BOOLEAN = DotName.createSimple(boolean.class);
@@ -32,4 +36,7 @@ class DotNames {
     public static final DotName SET = DotName.createSimple(Set.class);
 
     public static final DotName OBJECT = DotName.createSimple(Object.class.getName());
+    public static final DotName CDI_INSTANCE = DotName.createSimple(Instance.class);
+
+    public static final DotName CHAT_MODEL_LISTENER = DotName.createSimple(ChatModelListener.class);
 }

core/deployment/src/main/java/io/quarkiverse/langchain4j/deployment/ListenersProcessor.java

@@ -0,0 +1,34 @@
+package io.quarkiverse.langchain4j.deployment;
+
+import java.util.Optional;
+
+import io.quarkiverse.langchain4j.runtime.listeners.MetricsChatModelListener;
+import io.quarkiverse.langchain4j.runtime.listeners.SpanChatModelListener;
+import io.quarkus.arc.deployment.AdditionalBeanBuildItem;
+import io.quarkus.deployment.Capabilities;
+import io.quarkus.deployment.Capability;
+import io.quarkus.deployment.annotations.BuildProducer;
+import io.quarkus.deployment.annotations.BuildStep;
+import io.quarkus.deployment.metrics.MetricsCapabilityBuildItem;
+import io.quarkus.runtime.metrics.MetricsFactory;
+
+public class ListenersProcessor {
+
+    @BuildStep
+    public void spanListeners(Capabilities capabilities,
+            Optional<MetricsCapabilityBuildItem> metricsCapability,
+            BuildProducer<AdditionalBeanBuildItem> additionalBeanProducer) {
+        var addOpenTelemetrySpan = capabilities.isPresent(Capability.OPENTELEMETRY_TRACER);
+        if (addOpenTelemetrySpan) {
+            additionalBeanProducer.produce(
+                    AdditionalBeanBuildItem.builder().addBeanClass(SpanChatModelListener.class).setUnremovable().build());
+        }
+
+        var addMicrometerMetrics = metricsCapability.isPresent()
+                && metricsCapability.get().metricsSupported(MetricsFactory.MICROMETER);
+        if (addMicrometerMetrics) {
+            additionalBeanProducer.produce(
+                    AdditionalBeanBuildItem.builder().addBeanClass(MetricsChatModelListener.class).setUnremovable().build());
+        }
+    }
+}

core/runtime/src/main/java/io/quarkiverse/langchain4j/runtime/InProcessEmbeddingRecorder.java

@@ -1,11 +1,13 @@
 package io.quarkiverse.langchain4j.runtime;
 
+import java.util.concurrent.Executor;
 import java.util.function.Supplier;
 
 import org.jboss.logging.Logger;
 
 import dev.langchain4j.model.embedding.AbstractInProcessEmbeddingModel;
 import io.quarkus.runtime.annotations.Recorder;
+import io.smallrye.mutiny.infrastructure.Infrastructure;
 
 @Recorder
 public class InProcessEmbeddingRecorder {
@@ -15,9 +17,18 @@ public Supplier<?> instantiate(String className) {
             @Override
             public Object get() {
                 try {
+                    Executor executor = Infrastructure.getDefaultWorkerPool();
                     Class<? extends AbstractInProcessEmbeddingModel> loaded = (Class<? extends AbstractInProcessEmbeddingModel>) InProcessEmbeddingRecorder.class
                             .getClassLoader().loadClass(className);
-                    return loaded.getConstructor().newInstance();
+                    try {
+                        return loaded.getDeclaredConstructor(Executor.class).newInstance(executor);
+                    } catch (NoSuchMethodException e) {
+                        Logger.getLogger(InProcessEmbeddingRecorder.class)
+                                .warn("Cannot locate in-process embedding model's %s " +
+                                        "constructor that takes a java.util.concurrent.Executor, " +
+                                        "using the default constructor as fallback", className, e);
+                        return loaded.getDeclaredConstructor().newInstance();
+                    }
                 } catch (Exception e) {
                     Logger.getLogger(InProcessEmbeddingRecorder.class)
                             .errorf("Failed to instantiate in-process embedding model %s", className, e);

core/runtime/src/main/java/io/quarkiverse/langchain4j/runtime/listeners/MetricsChatModelListener.java

@@ -0,0 +1,126 @@
+package io.quarkiverse.langchain4j.runtime.listeners;
+
+import java.util.concurrent.TimeUnit;
+
+import org.jboss.logging.Logger;
+
+import dev.langchain4j.data.message.AiMessage;
+import dev.langchain4j.model.chat.listener.ChatModelErrorContext;
+import dev.langchain4j.model.chat.listener.ChatModelListener;
+import dev.langchain4j.model.chat.listener.ChatModelRequest;
+import dev.langchain4j.model.chat.listener.ChatModelRequestContext;
+import dev.langchain4j.model.chat.listener.ChatModelResponse;
+import dev.langchain4j.model.chat.listener.ChatModelResponseContext;
+import dev.langchain4j.model.output.TokenUsage;
+import io.micrometer.core.instrument.Clock;
+import io.micrometer.core.instrument.Counter;
+import io.micrometer.core.instrument.Meter;
+import io.micrometer.core.instrument.Metrics;
+import io.micrometer.core.instrument.Tag;
+import io.micrometer.core.instrument.Tags;
+import io.micrometer.core.instrument.Timer;
+
+/**
+ * Creates metrics that follow the
+ * <a href="https://github.com/open-telemetry/semantic-conventions/blob/main/docs/gen-ai/gen-ai-spans.md">Semantic Conventions
+ * for GenAI Metrics</a>
+ */
+public class MetricsChatModelListener implements ChatModelListener {
+
+    private static final Logger log = Logger.getLogger(MetricsChatModelListener.class);
+
+    public static final String START_TIME_KEY_NAME = "startTime";
+    private final Meter.MeterProvider<Counter> inputTokenUsage;
+    private final Meter.MeterProvider<Counter> outputTokenUsage;
+    private final Meter.MeterProvider<Timer> duration;
+
+    public MetricsChatModelListener() {
+        inputTokenUsage = Counter.builder("gen_ai.client.token.usage")
+                .description("Measures number of input tokens used")
+                .tag("gen_ai.operation.name", "completion")
+                .tag("gen_ai.token.type", "input")
+                .withRegistry(Metrics.globalRegistry);
+        outputTokenUsage = Counter.builder("gen_ai.client.token.usage")
+                .description("Measures number of output tokens used")
+                .tag("gen_ai.operation.name", "completion")
+                .tag("gen_ai.token.type", "output")
+                .withRegistry(Metrics.globalRegistry);
+        duration = Timer.builder("gen_ai.client.operation.duration")
+                .description("GenAI operation duration")
+                .tag("gen_ai.operation.name", "completion")
+                .withRegistry(Metrics.globalRegistry);
+    }
+
+    @Override
+    public void onRequest(ChatModelRequestContext requestContext) {
+        final long startTime = Clock.SYSTEM.monotonicTime();
+        requestContext.attributes().put(START_TIME_KEY_NAME, startTime);
+    }
+
+    @Override
+    public void onResponse(ChatModelResponseContext responseContext) {
+        final long endTime = Clock.SYSTEM.monotonicTime();
+
+        ChatModelRequest request = responseContext.request();
+        ChatModelResponse response = responseContext.response();
+        Tags tags = Tags.of(
+                Tag.of("gen_ai.request.model", request.model()),
+                Tag.of("gen_ai.response.model", response.model()));
+
+        recordTokenUsage(responseContext, tags);
+        recordDuration(responseContext, endTime, tags);
+    }
+
+    @Override
+    public void onError(ChatModelErrorContext errorContext) {
+        final long endTime = Clock.SYSTEM.monotonicTime();
+
+        Long startTime = (Long) errorContext.attributes().get(START_TIME_KEY_NAME);
+        if (startTime == null) {
+            // should never happen
+            log.warn("No start time found in response");
+            return;
+        }
+
+        AiMessage aiMessage = errorContext.partialResponse().aiMessage();
+        if (aiMessage == null) {
+            return;
+        }
+
+        Tags tags = Tags.of(
+                Tag.of("gen_ai.request.model", errorContext.request().model()),
+                Tag.of("gen_ai.response.model", errorContext.partialResponse().model()),
+                Tag.of("error.type", aiMessage.text()));
+        duration.withTags(tags).record(endTime - startTime, TimeUnit.NANOSECONDS);
+    }
+
+    private void recordTokenUsage(ChatModelResponseContext responseContext, Tags tags) {
+        TokenUsage tokenUsage = responseContext.response().tokenUsage();
+        if (tokenUsage == null) {
+            return;
+        }
+
+        Integer inputTokenCount = tokenUsage.inputTokenCount();
+        if (inputTokenCount != null) {
+            inputTokenUsage
+                    .withTags(tags)
+                    .increment(inputTokenCount);
+        }
+        Integer outputTokenCount = tokenUsage.outputTokenCount();
+        if (outputTokenCount != null) {
+            outputTokenUsage
+                    .withTags(tags)
+                    .increment(outputTokenCount);
+        }
+    }
+
+    private void recordDuration(ChatModelResponseContext responseContext, long endTime, Tags tags) {
+        Long startTime = (Long) responseContext.attributes().get(START_TIME_KEY_NAME);
+        if (startTime == null) {
+            // should never happen
+            log.warn("No start time found in response");
+            return;
+        }
+        duration.withTags(tags).record(endTime - startTime, TimeUnit.NANOSECONDS);
+    }
+}

core/runtime/src/main/java/io/quarkiverse/langchain4j/runtime/listeners/SpanChatModelListener.java

@@ -0,0 +1,103 @@
+package io.quarkiverse.langchain4j.runtime.listeners;
+
+import java.util.Map;
+
+import jakarta.inject.Inject;
+
+import org.jboss.logging.Logger;
+
+import dev.langchain4j.model.chat.listener.ChatModelErrorContext;
+import dev.langchain4j.model.chat.listener.ChatModelListener;
+import dev.langchain4j.model.chat.listener.ChatModelRequest;
+import dev.langchain4j.model.chat.listener.ChatModelRequestContext;
+import dev.langchain4j.model.chat.listener.ChatModelResponse;
+import dev.langchain4j.model.chat.listener.ChatModelResponseContext;
+import dev.langchain4j.model.output.TokenUsage;
+import io.opentelemetry.api.trace.Span;
+import io.opentelemetry.api.trace.Tracer;
+import io.opentelemetry.context.Scope;
+
+/**
+ * Creates a span that follows the
+ * <a href="https://github.com/open-telemetry/semantic-conventions/blob/main/docs/gen-ai/gen-ai-spans.md">Semantic Conventions
+ * for GenAI operations</a>
+ */
+public class SpanChatModelListener implements ChatModelListener {
+
+    private static final Logger log = Logger.getLogger(SpanChatModelListener.class);
+
+    private static final String OTEL_SCOPE_KEY_NAME = "OTelScope";
+    private static final String OTEL_SPAN_KEY_NAME = "OTelSpan";
+
+    private final Tracer tracer;
+
+    @Inject
+    public SpanChatModelListener(Tracer tracer) {
+        this.tracer = tracer;
+    }
+
+    @Override
+    public void onRequest(ChatModelRequestContext requestContext) {
+        ChatModelRequest request = requestContext.request();
+        Span span = tracer.spanBuilder("completion " + request.model())
+                .setAttribute("gen_ai.request.model", request.model())
+                .setAttribute("gen_ai.request.temperature", request.temperature())
+                .setAttribute("gen_ai.request.top_p", request.topP())
+                .startSpan();
+        Scope scope = span.makeCurrent();
+
+        var attributes = requestContext.attributes();
+        attributes.put(OTEL_SCOPE_KEY_NAME, scope);
+        attributes.put(OTEL_SPAN_KEY_NAME, span);
+    }
+
+    @Override
+    public void onResponse(ChatModelResponseContext responseContext) {
+        var attributes = responseContext.attributes();
+        Span span = (Span) attributes.get(OTEL_SPAN_KEY_NAME);
+        if (span != null) {
+            ChatModelResponse response = responseContext.response();
+            span.setAttribute("gen_ai.response.id", response.id()).setAttribute("gen_ai.response.model", response.model());
+            if (response.finishReason() != null) {
+                span.setAttribute("gen_ai.response.finish_reasons", response.finishReason().toString());
+            }
+            TokenUsage tokenUsage = response.tokenUsage();
+            if (tokenUsage != null) {
+                span.setAttribute("gen_ai.usage.completion_tokens", tokenUsage.outputTokenCount())
+                        .setAttribute("gen_ai.usage.prompt_tokens", tokenUsage.inputTokenCount());
+            }
+            span.end();
+        } else {
+            // should never happen
+            log.warn("No Span found in response");
+        }
+        safeCloseScope(attributes);
+    }
+
+    @Override
+    public void onError(ChatModelErrorContext errorContext) {
+        var attributes = errorContext.attributes();
+        Span span = (Span) attributes.get(OTEL_SPAN_KEY_NAME);
+        if (span != null) {
+            span.recordException(errorContext.error());
+        } else {
+            // should never happen
+            log.warn("No Span found in response");
+        }
+        safeCloseScope(errorContext.attributes());
+    }
+
+    private void safeCloseScope(Map<Object, Object> attributes) {
+        Scope scope = (Scope) attributes.get(OTEL_SCOPE_KEY_NAME);
+        if (scope == null) {
+            // should never happen
+            log.warn("No Scope found in response");
+        } else {
+            try {
+                scope.close();
+            } catch (Exception e) {
+                log.warn("Error closing scope", e);
+            }
+        }
+    }
+}

docs/modules/ROOT/pages/includes/attributes.adoc

@@ -1,3 +1,3 @@
 :project-version: 0.15.1
-:langchain4j-version: 0.31.0
+:langchain4j-version: 0.32.0-SNAPSHOT
 :examples-dir: ./../examples/

embedding-stores/milvus/deployment/src/main/java/io/quarkiverse/langchain4j/milvus/MilvusBuildConfig.java

@@ -30,7 +30,7 @@ interface MilvusDevServicesBuildTimeConfig {
         /**
          * Container image for Milvus.
          */
-        @WithDefault("docker.io/milvusdb/milvus:v2.3.4")
+        @WithDefault("docker.io/milvusdb/milvus:v2.3.16")
         String milvusImageName();
 
         /**