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

Added suport to anthropic prompt cache #1413

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
+128 −7
Open

Added suport to anthropic prompt cache #1413

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
a77c7a8
Added suport to anthropic prompt cache
Claudio-code Sep 26, 2024
0631de8
Adjusted source code format
Claudio-code Sep 26, 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
13 changes: 12 additions & 1 deletion ...pring-ai-anthropic/src/main/java/org/springframework/ai/anthropic/AnthropicChatModel.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,7 +35,9 @@
import org.springframework.ai.anthropic.api.AnthropicApi.ContentBlock;
import org.springframework.ai.anthropic.api.AnthropicApi.ContentBlock.Type;
import org.springframework.ai.anthropic.api.AnthropicApi.Role;
import org.springframework.ai.anthropic.api.AnthropicCacheType;
import org.springframework.ai.anthropic.metadata.AnthropicUsage;
import org.springframework.ai.chat.messages.AbstractMessage;
import org.springframework.ai.chat.messages.AssistantMessage;
import org.springframework.ai.chat.messages.MessageType;
import org.springframework.ai.chat.messages.ToolResponseMessage;
Expand Down Expand Up @@ -362,7 +364,16 @@ ChatCompletionRequest createRequest(Prompt prompt, boolean stream) {
.filter(message -> message.getMessageType() != MessageType.SYSTEM)
.map(message -> {
if (message.getMessageType() == MessageType.USER) {
List<ContentBlock> contents = new ArrayList<>(List.of(new ContentBlock(message.getContent())));
AbstractMessage abstractMessage = (AbstractMessage) message;
List<ContentBlock> contents;
if (abstractMessage.getCache() != null) {
AnthropicCacheType cacheType = AnthropicCacheType.valueOf(abstractMessage.getCache());
contents = new ArrayList<>(
List.of(new ContentBlock(message.getContent(), cacheType.cacheControl())));
}
else {
contents = new ArrayList<>(List.of(new ContentBlock(message.getContent())));
}
if (message instanceof UserMessage userMessage) {
if (!CollectionUtils.isEmpty(userMessage.getMedia())) {
List<ContentBlock> mediaContent = userMessage.getMedia()
Expand Down
30 changes: 24 additions & 6 deletions .../spring-ai-anthropic/src/main/java/org/springframework/ai/anthropic/api/AnthropicApi.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,6 +23,7 @@
import java.util.function.Consumer;
import java.util.function.Predicate;

import org.springframework.ai.anthropic.api.AnthropicApi.ChatCompletionRequest.CacheControl;
import org.springframework.ai.anthropic.api.StreamHelper.ChatCompletionResponseBuilder;
import org.springframework.ai.model.ChatModelDescription;
import org.springframework.ai.model.ModelOptionsUtils;
Expand Down Expand Up @@ -71,6 +72,8 @@ public class AnthropicApi {

public static final String BETA_MAX_TOKENS = "max-tokens-3-5-sonnet-2024-07-15";

public static final String BETA_PROMPT_CACHING = "prompt-caching-2024-07-31";

private static final Predicate<String> SSE_DONE_PREDICATE = "[DONE]"::equals;

private final RestClient restClient;
Expand Down Expand Up @@ -267,6 +270,14 @@ public ChatCompletionRequest(String model, List<AnthropicMessage> messages, Stri
public record Metadata(@JsonProperty("user_id") String userId) {
}

/**
* @param type is the cache type supported by anthropic. <a href=
* "https://docs.anthropic.com/en/docs/build-with-claude/prompt-caching#cache-limitations">Doc</a>
*/
@JsonInclude(Include.NON_NULL)
public record CacheControl(String type) {
}

public static ChatCompletionRequestBuilder builder() {
return new ChatCompletionRequestBuilder();
}
Expand Down Expand Up @@ -433,7 +444,10 @@ public record ContentBlock( // @formatter:off

// tool_result response only
@JsonProperty("tool_use_id") String toolUseId,
@JsonProperty("content") String content
@JsonProperty("content") String content,

// cache object
@JsonProperty("cache_control") CacheControl cacheControl
) {
// @formatter:on

Expand All @@ -442,25 +456,29 @@ public ContentBlock(String mediaType, String data) {
}

public ContentBlock(Source source) {
this(Type.IMAGE, source, null, null, null, null, null, null, null);
this(Type.IMAGE, source, null, null, null, null, null, null, null, null);
}

public ContentBlock(String text) {
this(Type.TEXT, null, text, null, null, null, null, null, null);
this(Type.TEXT, null, text, null, null, null, null, null, null, null);
}

public ContentBlock(String text, CacheControl cache) {
this(Type.TEXT, null, text, null, null, null, null, null, null, cache);
}

// Tool result
public ContentBlock(Type type, String toolUseId, String content) {
this(type, null, null, null, null, null, null, toolUseId, content);
this(type, null, null, null, null, null, null, toolUseId, content, null);
}

public ContentBlock(Type type, Source source, String text, Integer index) {
this(type, source, text, index, null, null, null, null, null);
this(type, source, text, index, null, null, null, null, null, null);
}

// Tool use input JSON delta streaming
public ContentBlock(Type type, String id, String name, Map<String, Object> input) {
this(type, null, null, null, id, name, input, null, null);
this(type, null, null, null, id, name, input, null, null, null);
}

/**
Expand Down
21 changes: 21 additions & 0 deletions ...g-ai-anthropic/src/main/java/org/springframework/ai/anthropic/api/AnthropicCacheType.java
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,21 @@
package org.springframework.ai.anthropic.api;

import org.springframework.ai.anthropic.api.AnthropicApi.ChatCompletionRequest.CacheControl;

import java.util.function.Supplier;

public enum AnthropicCacheType {

EPHEMERAL(() -> new CacheControl("ephemeral"));

private Supplier<CacheControl> value;

AnthropicCacheType(Supplier<CacheControl> value) {
this.value = value;
}

public CacheControl cacheControl() {
return this.value.get();
}

}
18 changes: 18 additions & 0 deletions ...pring-ai-anthropic/src/test/java/org/springframework/ai/anthropic/api/AnthropicApiIT.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,8 +26,11 @@
import org.springframework.ai.anthropic.api.AnthropicApi.ChatCompletionResponse;
import org.springframework.ai.anthropic.api.AnthropicApi.ContentBlock;
import org.springframework.ai.anthropic.api.AnthropicApi.Role;
import org.springframework.ai.retry.RetryUtils;
import org.springframework.http.ResponseEntity;

import org.springframework.web.client.RestClient;
import org.springframework.web.reactive.function.client.WebClient;
import reactor.core.publisher.Flux;

/**
Expand All @@ -38,6 +41,21 @@ public class AnthropicApiIT {

AnthropicApi anthropicApi = new AnthropicApi(System.getenv("ANTHROPIC_API_KEY"));

@Test
void chatWithPromptCache() {
AnthropicApi anthropicApiBeta = new AnthropicApi(AnthropicApi.DEFAULT_BASE_URL,
System.getenv("ANTHROPIC_API_KEY"), AnthropicApi.DEFAULT_ANTHROPIC_VERSION, RestClient.builder(),
WebClient.builder(), RetryUtils.DEFAULT_RESPONSE_ERROR_HANDLER, AnthropicApi.BETA_PROMPT_CACHING);
AnthropicMessage chatCompletionMessage = new AnthropicMessage(
List.of(new ContentBlock("Tell me a Joke?", AnthropicCacheType.EPHEMERAL.cacheControl())), Role.USER);

ResponseEntity<ChatCompletionResponse> response = anthropicApiBeta
.chatCompletionEntity(new ChatCompletionRequest(AnthropicApi.ChatModel.CLAUDE_3_HAIKU.getValue(),
List.of(chatCompletionMessage), null, 100, 0.8, false));

assertThat(response).isNotNull();
}

@Test
void chatCompletionEntity() {

Expand Down
32 changes: 32 additions & 0 deletions spring-ai-core/src/main/java/org/springframework/ai/chat/messages/AbstractMessage.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -41,11 +41,25 @@ public abstract class AbstractMessage implements Message {

protected final String textContent;

protected String cache;

/**
* Additional options for the message to influence the response, not a generative map.
*/
protected final Map<String, Object> metadata;

protected AbstractMessage(MessageType messageType, String textContent, Map<String, Object> metadata, String cache) {
Assert.notNull(messageType, "Message type must not be null");
if (messageType == MessageType.SYSTEM || messageType == MessageType.USER) {
Assert.notNull(textContent, "Content must not be null for SYSTEM or USER messages");
}
this.messageType = messageType;
this.textContent = textContent;
this.metadata = new HashMap<>(metadata);
this.metadata.put(MESSAGE_TYPE, messageType);
this.cache = cache;
}

protected AbstractMessage(MessageType messageType, String textContent, Map<String, Object> metadata) {
Assert.notNull(messageType, "Message type must not be null");
if (messageType == MessageType.SYSTEM || messageType == MessageType.USER) {
Expand All @@ -70,6 +84,20 @@ protected AbstractMessage(MessageType messageType, Resource resource, Map<String
this.metadata.put(MESSAGE_TYPE, messageType);
}

protected AbstractMessage(MessageType messageType, Resource resource, Map<String, Object> metadata, String cache) {
Assert.notNull(resource, "Resource must not be null");
try (InputStream inputStream = resource.getInputStream()) {
this.textContent = StreamUtils.copyToString(inputStream, Charset.defaultCharset());
}
catch (IOException ex) {
throw new RuntimeException("Failed to read resource", ex);
}
this.messageType = messageType;
this.metadata = new HashMap<>(metadata);
this.metadata.put(MESSAGE_TYPE, messageType);
this.cache = cache;
}

@Override
public String getContent() {
return this.textContent;
Expand All @@ -85,6 +113,10 @@ public MessageType getMessageType() {
return this.messageType;
}

public String getCache() {
return cache;
}

@Override
public boolean equals(Object o) {
if (this == o)
Expand Down
21 changes: 21 additions & 0 deletions spring-ai-core/src/main/java/org/springframework/ai/chat/messages/UserMessage.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,6 +35,10 @@ public class UserMessage extends AbstractMessage implements MediaContent {

protected final List<Media> media;

public UserMessage(String textContent, String cache) {
this(MessageType.USER, textContent, new ArrayList<>(), Map.of(), cache);
}

public UserMessage(String textContent) {
this(MessageType.USER, textContent, new ArrayList<>(), Map.of());
}
Expand All @@ -44,6 +48,11 @@ public UserMessage(Resource resource) {
this.media = new ArrayList<>();
}

public UserMessage(Resource resource, String cache) {
super(MessageType.USER, resource, Map.of(), cache);
this.media = new ArrayList<>();
}

public UserMessage(String textContent, List<Media> media) {
this(MessageType.USER, textContent, media, Map.of());
}
Expand All @@ -63,6 +72,13 @@ public UserMessage(MessageType messageType, String textContent, Collection<Media
this.media = new ArrayList<>(media);
}

public UserMessage(MessageType messageType, String textContent, Collection<Media> media,
Map<String, Object> metadata, String cache) {
super(messageType, textContent, metadata, cache);
Assert.notNull(media, "media data must not be null");
this.media = new ArrayList<>(media);
}

public List<Media> getMedia(String... dummy) {
return this.media;
}
Expand All @@ -83,4 +99,9 @@ public String getContent() {
return this.textContent;
}

@Override
public String getCache() {
return super.getCache();
}

}