Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Enable semantic cache with @CacheResult #659

Draft
wants to merge 6 commits into
base: main
Choose a base branch
from
Draft

Enable semantic cache with @CacheResult #659

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
ed1d11c
Enable semantic cache with @CacheResult
andreadimaio Jul 2, 2024
1fd0fb9
Allow selection of preferred embedding model
andreadimaio Jun 6, 2024
a78a4fc
Avoid using cache and tools in the same @RegisterAiService
andreadimaio Jun 6, 2024
9530cc0
Compose the cache message with SystemMessage + UserMessage
andreadimaio Jun 6, 2024
02ea764
Generate the id for the Cache with a unique context if possible
andreadimaio Jun 7, 2024
d961991
Add cache documentation
andreadimaio Jun 11, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
16 changes: 16 additions & 0 deletions core/deployment/src/main/java/io/quarkiverse/langchain4j/deployment/AiCacheBuildConfig.java
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,16 @@
package io.quarkiverse.langchain4j.deployment;

import static io.quarkus.runtime.annotations.ConfigPhase.BUILD_TIME;

import io.quarkus.runtime.annotations.ConfigRoot;
import io.smallrye.config.ConfigMapping;

@ConfigRoot(phase = BUILD_TIME)
@ConfigMapping(prefix = "quarkus.langchain4j.cache")
public interface AiCacheBuildConfig {

/**
* Ai Cache embedding model related settings
*/
CacheEmbeddingModelConfig embedding();
}
22 changes: 22 additions & 0 deletions core/deployment/src/main/java/io/quarkiverse/langchain4j/deployment/AiCacheBuildItem.java
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,22 @@
package io.quarkiverse.langchain4j.deployment;

import io.quarkus.builder.item.SimpleBuildItem;

public final class AiCacheBuildItem extends SimpleBuildItem {

private boolean enable;
private String embeddingModelName;

public AiCacheBuildItem(boolean enable, String embeddingModelName) {
this.enable = enable;
this.embeddingModelName = embeddingModelName;
}

public boolean isEnable() {
return enable;
}

public String getEmbeddingModelName() {
return embeddingModelName;
}
}
81 changes: 81 additions & 0 deletions core/deployment/src/main/java/io/quarkiverse/langchain4j/deployment/AiCacheProcessor.java
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,81 @@
package io.quarkiverse.langchain4j.deployment;

import jakarta.enterprise.context.ApplicationScoped;

import org.jboss.jandex.AnnotationInstance;
import org.jboss.jandex.AnnotationTarget;
import org.jboss.jandex.ClassInfo;
import org.jboss.jandex.ClassType;
import org.jboss.jandex.IndexView;

import dev.langchain4j.model.embedding.EmbeddingModel;
import io.quarkiverse.langchain4j.ModelName;
import io.quarkiverse.langchain4j.runtime.AiCacheRecorder;
import io.quarkiverse.langchain4j.runtime.NamedConfigUtil;
import io.quarkiverse.langchain4j.runtime.cache.AiCacheProvider;
import io.quarkiverse.langchain4j.runtime.cache.AiCacheStore;
import io.quarkiverse.langchain4j.runtime.cache.config.AiCacheConfig;
import io.quarkus.arc.deployment.SyntheticBeanBuildItem;
import io.quarkus.arc.deployment.UnremovableBeanBuildItem;
import io.quarkus.deployment.annotations.BuildProducer;
import io.quarkus.deployment.annotations.BuildStep;
import io.quarkus.deployment.annotations.ExecutionTime;
import io.quarkus.deployment.annotations.Record;
import io.quarkus.deployment.builditem.CombinedIndexBuildItem;

public class AiCacheProcessor {

@BuildStep
@Record(ExecutionTime.RUNTIME_INIT)
void setupBeans(AiCacheBuildConfig cacheBuildConfig,
AiCacheConfig cacheConfig,
AiCacheRecorder recorder,
CombinedIndexBuildItem indexBuildItem,
BuildProducer<AiCacheBuildItem> aiCacheBuildItemProducer,
BuildProducer<UnremovableBeanBuildItem> unremovableProducer,
BuildProducer<SyntheticBeanBuildItem> syntheticBeanProducer) {

IndexView index = indexBuildItem.getIndex();
boolean enableCache = false;

for (AnnotationInstance instance : index.getAnnotations(LangChain4jDotNames.REGISTER_AI_SERVICES)) {
if (instance.target().kind() != AnnotationTarget.Kind.CLASS) {
continue;
}

ClassInfo declarativeAiServiceClassInfo = instance.target().asClass();

if (declarativeAiServiceClassInfo.hasAnnotation(LangChain4jDotNames.CACHE_RESULT)) {
enableCache = true;
break;
}
}

String embeddingModelName = NamedConfigUtil.DEFAULT_NAME;
if (cacheBuildConfig.embedding() != null)
embeddingModelName = cacheBuildConfig.embedding().name().orElse(NamedConfigUtil.DEFAULT_NAME);

aiCacheBuildItemProducer.produce(new AiCacheBuildItem(enableCache, embeddingModelName));

if (enableCache) {
SyntheticBeanBuildItem.ExtendedBeanConfigurator configurator = SyntheticBeanBuildItem
.configure(AiCacheProvider.class)
.setRuntimeInit()
.addInjectionPoint(ClassType.create(AiCacheStore.class))
.scope(ApplicationScoped.class)
.createWith(recorder.messageWindow(cacheConfig, embeddingModelName))
.defaultBean();

if (NamedConfigUtil.isDefault(embeddingModelName)) {
configurator.addInjectionPoint(ClassType.create(LangChain4jDotNames.EMBEDDING_MODEL));
} else {
configurator.addInjectionPoint(ClassType.create(LangChain4jDotNames.EMBEDDING_MODEL),
AnnotationInstance.builder(ModelName.class).add("value", embeddingModelName).build());
}

syntheticBeanProducer.produce(configurator.done());
unremovableProducer.produce(UnremovableBeanBuildItem.beanTypes(AiCacheStore.class));
unremovableProducer.produce(UnremovableBeanBuildItem.beanTypes(EmbeddingModel.class));
}
}
}
103 changes: 76 additions & 27 deletions core/deployment/src/main/java/io/quarkiverse/langchain4j/deployment/AiServicesProcessor.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -52,7 +52,6 @@
import org.objectweb.asm.tree.analysis.AnalyzerException;

import dev.langchain4j.exception.IllegalConfigurationException;
import dev.langchain4j.service.Moderate;
import io.quarkiverse.langchain4j.ModelName;
import io.quarkiverse.langchain4j.ToolBox;
import io.quarkiverse.langchain4j.deployment.config.LangChain4jBuildConfig;
Expand Down Expand Up @@ -189,6 +188,7 @@ public void findDeclarativeServices(CombinedIndexBuildItem indexBuildItem,

Set<String> chatModelNames = new HashSet<>();
Set<String> moderationModelNames = new HashSet<>();

for (AnnotationInstance instance : index.getAnnotations(LangChain4jDotNames.REGISTER_AI_SERVICES)) {
if (instance.target().kind() != AnnotationTarget.Kind.CLASS) {
continue; // should never happen
Expand All @@ -210,14 +210,11 @@ public void findDeclarativeServices(CombinedIndexBuildItem indexBuildItem,
}

String chatModelName = NamedConfigUtil.DEFAULT_NAME;
String moderationModelName = NamedConfigUtil.DEFAULT_NAME;

if (chatLanguageModelSupplierClassDotName == null) {
AnnotationValue modelNameValue = instance.value("modelName");
if (modelNameValue != null) {
String modelNameValueStr = modelNameValue.asString();
if ((modelNameValueStr != null) && !modelNameValueStr.isEmpty()) {
chatModelName = modelNameValueStr;
}
}
chatModelName = getModelName(modelNameValue);
chatModelNames.add(chatModelName);
}

Expand All @@ -243,6 +240,18 @@ public void findDeclarativeServices(CombinedIndexBuildItem indexBuildItem,
}
}

// the default value depends on whether tools exists or not - if they do, then we require a AiCacheProvider bean
DotName aiCacheProviderSupplierClassDotName = LangChain4jDotNames.BEAN_AI_CACHE_PROVIDER_SUPPLIER;
AnnotationValue aiCacheProviderSupplierValue = instance.value("cacheProviderSupplier");
if (aiCacheProviderSupplierValue != null) {
aiCacheProviderSupplierClassDotName = aiCacheProviderSupplierValue.asClass().name();
if (!aiCacheProviderSupplierClassDotName
.equals(LangChain4jDotNames.BEAN_AI_CACHE_PROVIDER_SUPPLIER)) {
validateSupplierAndRegisterForReflection(aiCacheProviderSupplierClassDotName, index,
reflectiveClassProducer);
}
}

DotName retrieverClassDotName = null;
AnnotationValue retrieverValue = instance.value("retriever");
if (retrieverValue != null) {
Expand Down Expand Up @@ -296,17 +305,11 @@ public void findDeclarativeServices(CombinedIndexBuildItem indexBuildItem,
}

// determine whether the method is annotated with @Moderate
String moderationModelName = NamedConfigUtil.DEFAULT_NAME;
for (MethodInfo method : declarativeAiServiceClassInfo.methods()) {
if (method.hasAnnotation(LangChain4jDotNames.MODERATE)) {
if (moderationModelSupplierClassName.equals(LangChain4jDotNames.BEAN_IF_EXISTS_MODERATION_MODEL_SUPPLIER)) {
AnnotationValue modelNameValue = instance.value("modelName");
if (modelNameValue != null) {
String modelNameValueStr = modelNameValue.asString();
if ((modelNameValueStr != null) && !modelNameValueStr.isEmpty()) {
moderationModelName = modelNameValueStr;
}
}
moderationModelName = getModelName(modelNameValue);
moderationModelNames.add(moderationModelName);
}
break;
Expand All @@ -325,13 +328,15 @@ public void findDeclarativeServices(CombinedIndexBuildItem indexBuildItem,
chatLanguageModelSupplierClassDotName,
toolDotNames,
chatMemoryProviderSupplierClassDotName,
aiCacheProviderSupplierClassDotName,
retrieverClassDotName,
retrievalAugmentorSupplierClassName,
customRetrievalAugmentorSupplierClassIsABean,
auditServiceSupplierClassName,
moderationModelSupplierClassName,
cdiScope,
chatModelName, moderationModelName));
chatModelName,
moderationModelName));
}

for (String chatModelName : chatModelNames) {
Expand Down Expand Up @@ -365,7 +370,8 @@ public void handleDeclarativeServices(AiServicesRecorder recorder,
List<DeclarativeAiServiceBuildItem> declarativeAiServiceItems,
List<SelectedChatModelProviderBuildItem> selectedChatModelProvider,
BuildProducer<SyntheticBeanBuildItem> syntheticBeanProducer,
BuildProducer<UnremovableBeanBuildItem> unremoveableProducer) {
BuildProducer<UnremovableBeanBuildItem> unremoveableProducer,
AiCacheBuildItem aiCacheBuildItem) {

boolean needsChatModelBean = false;
boolean needsStreamingChatModelBean = false;
Expand All @@ -374,6 +380,8 @@ public void handleDeclarativeServices(AiServicesRecorder recorder,
boolean needsRetrievalAugmentorBean = false;
boolean needsAuditServiceBean = false;
boolean needsModerationModelBean = false;
boolean needsAiCacheProvider = false;

Set<DotName> allToolNames = new HashSet<>();

for (DeclarativeAiServiceBuildItem bi : declarativeAiServiceItems) {
Expand All @@ -390,6 +398,10 @@ public void handleDeclarativeServices(AiServicesRecorder recorder,
? bi.getChatMemoryProviderSupplierClassDotName().toString()
: null;

String aiCacheProviderSupplierClassName = bi.getAiCacheProviderSupplierClassDotName() != null
? bi.getAiCacheProviderSupplierClassDotName().toString()
: null;

String retrieverClassName = bi.getRetrieverClassDotName() != null
? bi.getRetrieverClassDotName().toString()
: null;
Expand All @@ -407,7 +419,7 @@ public void handleDeclarativeServices(AiServicesRecorder recorder,
: null);

// determine whether the method returns Multi<String>
boolean injectStreamingChatModelBean = false;
boolean needsStreamingChatModel = false;
for (MethodInfo method : declarativeAiServiceClassInfo.methods()) {
if (!LangChain4jDotNames.MULTI.equals(method.returnType().name())) {
continue;
Expand All @@ -423,29 +435,41 @@ public void handleDeclarativeServices(AiServicesRecorder recorder,
throw illegalConfiguration("Only Multi<String> is supported as a Multi return type. Offending method is '"
+ method.declaringClass().name().toString() + "#" + method.name() + "'");
}
injectStreamingChatModelBean = true;
needsStreamingChatModel = true;
}

boolean injectModerationModelBean = false;
boolean needsModerationModel = false;
for (MethodInfo method : declarativeAiServiceClassInfo.methods()) {
if (method.hasAnnotation(Moderate.class)) {
injectModerationModelBean = true;
if (method.hasAnnotation(LangChain4jDotNames.MODERATE)) {
needsModerationModel = true;
break;
}
}

String chatModelName = bi.getChatModelName();
String moderationModelName = bi.getModerationModelName();
boolean enableCache = aiCacheBuildItem.isEnable();

// It is not possible to use the cache in combination with the tools.
if (!toolClassNames.isEmpty() && enableCache
&& declarativeAiServiceClassInfo.hasAnnotation(LangChain4jDotNames.CACHE_RESULT)) {
throw new RuntimeException("The cache cannot be used in combination with the tools. Affected class: %s"
.formatted(serviceClassName));
}

SyntheticBeanBuildItem.ExtendedBeanConfigurator configurator = SyntheticBeanBuildItem
.configure(QuarkusAiServiceContext.class)
.forceApplicationClass()
.createWith(recorder.createDeclarativeAiService(
new DeclarativeAiServiceCreateInfo(serviceClassName, chatLanguageModelSupplierClassName,
toolClassNames, chatMemoryProviderSupplierClassName, retrieverClassName,
toolClassNames, chatMemoryProviderSupplierClassName, aiCacheProviderSupplierClassName,
retrieverClassName,
retrievalAugmentorSupplierClassName,
auditServiceClassSupplierName, moderationModelSupplierClassName, chatModelName,
moderationModelName,
injectStreamingChatModelBean, injectModerationModelBean)))
needsStreamingChatModel,
needsModerationModel,
enableCache)))
.setRuntimeInit()
.addQualifier()
.annotation(LangChain4jDotNames.QUARKUS_AI_SERVICE_CONTEXT_QUALIFIER).addValue("value", serviceClassName)
Expand All @@ -455,15 +479,15 @@ public void handleDeclarativeServices(AiServicesRecorder recorder,
if ((chatLanguageModelSupplierClassName == null) && !selectedChatModelProvider.isEmpty()) {
if (NamedConfigUtil.isDefault(chatModelName)) {
configurator.addInjectionPoint(ClassType.create(LangChain4jDotNames.CHAT_MODEL));
if (injectStreamingChatModelBean) {
if (needsStreamingChatModel) {
configurator.addInjectionPoint(ClassType.create(LangChain4jDotNames.STREAMING_CHAT_MODEL));
needsStreamingChatModelBean = true;
}
} else {
configurator.addInjectionPoint(ClassType.create(LangChain4jDotNames.CHAT_MODEL),
AnnotationInstance.builder(ModelName.class).add("value", chatModelName).build());

if (injectStreamingChatModelBean) {
if (needsStreamingChatModel) {
configurator.addInjectionPoint(ClassType.create(LangChain4jDotNames.STREAMING_CHAT_MODEL),
AnnotationInstance.builder(ModelName.class).add("value", chatModelName).build());
needsStreamingChatModelBean = true;
Expand Down Expand Up @@ -519,7 +543,7 @@ public void handleDeclarativeServices(AiServicesRecorder recorder,
}

if (LangChain4jDotNames.BEAN_IF_EXISTS_MODERATION_MODEL_SUPPLIER.toString()
.equals(moderationModelSupplierClassName) && injectModerationModelBean) {
.equals(moderationModelSupplierClassName) && needsModerationModel) {

if (NamedConfigUtil.isDefault(moderationModelName)) {
configurator.addInjectionPoint(ClassType.create(LangChain4jDotNames.MODERATION_MODEL));
Expand All @@ -531,6 +555,16 @@ public void handleDeclarativeServices(AiServicesRecorder recorder,
needsModerationModelBean = true;
}

if (enableCache) {

if (LangChain4jDotNames.BEAN_AI_CACHE_PROVIDER_SUPPLIER.toString().equals(aiCacheProviderSupplierClassName)) {
configurator.addInjectionPoint(ClassType.create(LangChain4jDotNames.AI_CACHE_PROVIDER));
} else {
configurator.addInjectionPoint(ClassType.create(LangChain4jDotNames.AI_CACHE_PROVIDER));
}
needsAiCacheProvider = true;
}

syntheticBeanProducer.produce(configurator.done());
}

Expand All @@ -555,6 +589,9 @@ public void handleDeclarativeServices(AiServicesRecorder recorder,
if (needsModerationModelBean) {
unremoveableProducer.produce(UnremovableBeanBuildItem.beanTypes(LangChain4jDotNames.MODERATION_MODEL));
}
if (needsAiCacheProvider) {
unremoveableProducer.produce(UnremovableBeanBuildItem.beanTypes(LangChain4jDotNames.AI_CACHE_PROVIDER));
}
if (!allToolNames.isEmpty()) {
unremoveableProducer.produce(UnremovableBeanBuildItem.beanTypes(allToolNames));
}
Expand Down Expand Up @@ -877,6 +914,8 @@ private AiServiceMethodCreateInfo gatherMethodMetadata(MethodInfo method, boolea

boolean requiresModeration = method.hasAnnotation(LangChain4jDotNames.MODERATE);
Class<?> returnType = JandexUtil.load(method.returnType(), Thread.currentThread().getContextClassLoader());
boolean requiresCache = method.declaringClass().hasDeclaredAnnotation(LangChain4jDotNames.CACHE_RESULT)
|| method.hasDeclaredAnnotation(LangChain4jDotNames.CACHE_RESULT);

List<MethodParameterInfo> params = method.parameters();

Expand Down Expand Up @@ -914,7 +953,7 @@ private AiServiceMethodCreateInfo gatherMethodMetadata(MethodInfo method, boolea
List<String> methodToolClassNames = gatherMethodToolClassNames(method);

return new AiServiceMethodCreateInfo(method.declaringClass().name().toString(), method.name(), systemMessageInfo,
userMessageInfo, memoryIdParamPosition, requiresModeration,
userMessageInfo, memoryIdParamPosition, requiresModeration, requiresCache,
returnType, metricsTimedInfo, metricsCountedInfo, spanInfo, responseSchemaInfo, methodToolClassNames);
}

Expand Down Expand Up @@ -1249,6 +1288,16 @@ static Map<String, Integer> toNameToArgsPositionMap(List<TemplateParameterInfo>
}
}

private String getModelName(AnnotationValue value) {
if (value != null) {
String modelNameValueStr = value.asString();
if ((modelNameValueStr != null) && !modelNameValueStr.isEmpty()) {
return modelNameValueStr;
}
}
return NamedConfigUtil.DEFAULT_NAME;
}

public static final class AiServicesMethodBuildItem extends MultiBuildItem {

private final MethodInfo methodInfo;
Expand Down
Loading
Loading