mirror of
https://github.com/tencentmusic/supersonic.git
synced 2025-12-14 13:47:09 +00:00
(improvement)(launcher)Move langchain4j related classes from launcher-common to common.
This commit is contained in:
jerryjzhang
30
common/src/main/java/dev/langchain4j/model/ChatModel.java
Normal file
30
common/src/main/java/dev/langchain4j/model/ChatModel.java
Normal file
@@ -0,0 +1,30 @@
|
||||
package dev.langchain4j.model;
|
||||
|
||||
public enum ChatModel {
|
||||
ZHIPU("glm"),
|
||||
ALI("qwen");
|
||||
|
||||
private final String modelName;
|
||||
|
||||
private ChatModel(String modelName) {
|
||||
this.modelName = modelName;
|
||||
}
|
||||
|
||||
public String toString() {
|
||||
return this.modelName;
|
||||
}
|
||||
|
||||
public static ChatModel from(String stringValue) {
|
||||
ChatModel[] var1 = values();
|
||||
int var2 = var1.length;
|
||||
|
||||
for (int var3 = 0; var3 < var2; ++var3) {
|
||||
ChatModel model = var1[var3];
|
||||
if (model.modelName.equals(stringValue)) {
|
||||
return model;
|
||||
}
|
||||
}
|
||||
|
||||
throw new IllegalArgumentException("Unknown role: '" + stringValue + "'");
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,61 @@
|
||||
package dev.langchain4j.model.embedding;
|
||||
|
||||
import java.io.IOException;
|
||||
import java.net.MalformedURLException;
|
||||
import java.net.URL;
|
||||
import java.nio.file.Files;
|
||||
import java.nio.file.Path;
|
||||
import java.nio.file.Paths;
|
||||
import org.apache.commons.lang3.StringUtils;
|
||||
|
||||
/**
|
||||
* An embedding model that runs within your Java application's process.
|
||||
* Any BERT-based model (e.g., from HuggingFace) can be used, as long as it is in ONNX format.
|
||||
* Information on how to convert models into ONNX format can be found <a
|
||||
* href="https://huggingface.co/docs/optimum/exporters/onnx/usage_guides/export_a_model">here</a>.
|
||||
* Many models already converted to ONNX format are available <a href="https://huggingface.co/Xenova">here</a>.
|
||||
* Copy from dev.langchain4j.model.embedding.OnnxEmbeddingModel.
|
||||
*/
|
||||
public class S2OnnxEmbeddingModel extends AbstractInProcessEmbeddingModel {
|
||||
|
||||
private final OnnxBertBiEncoder model;
|
||||
|
||||
/**
|
||||
* @param pathToModel The path to the .onnx model file (e.g., "/home/me/model.onnx").
|
||||
*/
|
||||
public S2OnnxEmbeddingModel(String pathToModel, String vocabularyPath) {
|
||||
URL resource = AbstractInProcessEmbeddingModel.class.getResource("/bert-vocabulary-en.txt");
|
||||
if (StringUtils.isNotBlank(vocabularyPath)) {
|
||||
try {
|
||||
resource = Paths.get(vocabularyPath).toUri().toURL();
|
||||
} catch (MalformedURLException e) {
|
||||
throw new RuntimeException(e);
|
||||
}
|
||||
}
|
||||
model = loadFromFileSystem(Paths.get(pathToModel), resource);
|
||||
}
|
||||
|
||||
/**
|
||||
* @param pathToModel The path to the .onnx model file (e.g., "/home/me/model.onnx").
|
||||
*/
|
||||
public S2OnnxEmbeddingModel(String pathToModel) {
|
||||
this(pathToModel, null);
|
||||
}
|
||||
|
||||
@Override
|
||||
protected OnnxBertBiEncoder model() {
|
||||
return model;
|
||||
}
|
||||
|
||||
static OnnxBertBiEncoder loadFromFileSystem(Path pathToModel, URL vocabularyFile) {
|
||||
try {
|
||||
return new OnnxBertBiEncoder(
|
||||
Files.newInputStream(pathToModel),
|
||||
vocabularyFile,
|
||||
PoolingMode.MEAN
|
||||
);
|
||||
} catch (IOException e) {
|
||||
throw new RuntimeException(e);
|
||||
}
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,227 @@
|
||||
package dev.langchain4j.model.openai;
|
||||
|
||||
import dev.ai4j.openai4j.OpenAiClient;
|
||||
import dev.ai4j.openai4j.chat.ChatCompletionChoice;
|
||||
import dev.ai4j.openai4j.chat.ChatCompletionRequest;
|
||||
import dev.ai4j.openai4j.chat.ChatCompletionResponse;
|
||||
import dev.ai4j.openai4j.chat.ChatCompletionRequest.Builder;
|
||||
import dev.langchain4j.agent.tool.ToolSpecification;
|
||||
import dev.langchain4j.data.message.AiMessage;
|
||||
import dev.langchain4j.data.message.ChatMessage;
|
||||
import dev.langchain4j.internal.RetryUtils;
|
||||
import dev.langchain4j.internal.Utils;
|
||||
import dev.langchain4j.model.ChatModel;
|
||||
import dev.langchain4j.model.Tokenizer;
|
||||
import dev.langchain4j.model.chat.ChatLanguageModel;
|
||||
import dev.langchain4j.model.chat.TokenCountEstimator;
|
||||
import dev.langchain4j.model.output.Response;
|
||||
|
||||
import java.net.Proxy;
|
||||
import java.time.Duration;
|
||||
import java.util.Collections;
|
||||
import java.util.List;
|
||||
|
||||
public class FullOpenAiChatModel implements ChatLanguageModel, TokenCountEstimator {
|
||||
|
||||
private final OpenAiClient client;
|
||||
private final String modelName;
|
||||
private final Double temperature;
|
||||
private final Double topP;
|
||||
private final List<String> stop;
|
||||
private final Integer maxTokens;
|
||||
private final Double presencePenalty;
|
||||
private final Double frequencyPenalty;
|
||||
private final Integer maxRetries;
|
||||
private final Tokenizer tokenizer;
|
||||
|
||||
public FullOpenAiChatModel(String baseUrl, String apiKey, String modelName, Double temperature,
|
||||
Double topP, List<String> stop, Integer maxTokens, Double presencePenalty,
|
||||
Double frequencyPenalty, Duration timeout, Integer maxRetries, Proxy proxy,
|
||||
Boolean logRequests, Boolean logResponses, Tokenizer tokenizer) {
|
||||
baseUrl = Utils.getOrDefault(baseUrl, "https://api.openai.com/v1");
|
||||
if ("demo".equals(apiKey)) {
|
||||
baseUrl = "http://langchain4j.dev/demo/openai/v1";
|
||||
}
|
||||
|
||||
timeout = Utils.getOrDefault(timeout, Duration.ofSeconds(60L));
|
||||
this.client = OpenAiClient.builder().openAiApiKey(apiKey)
|
||||
.baseUrl(baseUrl).callTimeout(timeout).connectTimeout(timeout)
|
||||
.readTimeout(timeout).writeTimeout(timeout).proxy(proxy)
|
||||
.logRequests(logRequests).logResponses(logResponses).build();
|
||||
this.modelName = Utils.getOrDefault(modelName, "gpt-3.5-turbo");
|
||||
this.temperature = Utils.getOrDefault(temperature, 0.7D);
|
||||
this.topP = topP;
|
||||
this.stop = stop;
|
||||
this.maxTokens = maxTokens;
|
||||
this.presencePenalty = presencePenalty;
|
||||
this.frequencyPenalty = frequencyPenalty;
|
||||
this.maxRetries = Utils.getOrDefault(maxRetries, 3);
|
||||
this.tokenizer = Utils.getOrDefault(tokenizer, new OpenAiTokenizer(this.modelName));
|
||||
}
|
||||
|
||||
public Response<AiMessage> generate(List<ChatMessage> messages) {
|
||||
return this.generate(messages, null, null);
|
||||
}
|
||||
|
||||
public Response<AiMessage> generate(List<ChatMessage> messages, List<ToolSpecification> toolSpecifications) {
|
||||
return this.generate(messages, toolSpecifications, null);
|
||||
}
|
||||
|
||||
public Response<AiMessage> generate(List<ChatMessage> messages, ToolSpecification toolSpecification) {
|
||||
return this.generate(messages, Collections.singletonList(toolSpecification), toolSpecification);
|
||||
}
|
||||
|
||||
private Response<AiMessage> generate(List<ChatMessage> messages,
|
||||
List<ToolSpecification> toolSpecifications,
|
||||
ToolSpecification toolThatMustBeExecuted) {
|
||||
Builder requestBuilder = null;
|
||||
if (modelName.contains(ChatModel.ZHIPU.toString()) || modelName.contains(ChatModel.ALI.toString())) {
|
||||
requestBuilder = ChatCompletionRequest.builder()
|
||||
.model(this.modelName)
|
||||
.messages(ImproveInternalOpenAiHelper.toOpenAiMessages(messages, this.modelName));
|
||||
} else {
|
||||
requestBuilder = ChatCompletionRequest.builder()
|
||||
.model(this.modelName)
|
||||
.messages(ImproveInternalOpenAiHelper.toOpenAiMessages(messages, this.modelName))
|
||||
.temperature(this.temperature).topP(this.topP).stop(this.stop).maxTokens(this.maxTokens)
|
||||
.presencePenalty(this.presencePenalty).frequencyPenalty(this.frequencyPenalty);
|
||||
}
|
||||
if (toolSpecifications != null && !toolSpecifications.isEmpty()) {
|
||||
requestBuilder.functions(InternalOpenAiHelper.toFunctions(toolSpecifications));
|
||||
}
|
||||
|
||||
if (toolThatMustBeExecuted != null) {
|
||||
requestBuilder.functionCall(toolThatMustBeExecuted.name());
|
||||
}
|
||||
|
||||
ChatCompletionRequest request = requestBuilder.build();
|
||||
ChatCompletionResponse response = (ChatCompletionResponse) RetryUtils.withRetry(() -> {
|
||||
return (ChatCompletionResponse) this.client.chatCompletion(request).execute();
|
||||
}, this.maxRetries);
|
||||
return Response.from(InternalOpenAiHelper.aiMessageFrom(response),
|
||||
InternalOpenAiHelper.tokenUsageFrom(response.usage()),
|
||||
InternalOpenAiHelper.finishReasonFrom(
|
||||
((ChatCompletionChoice) response.choices().get(0)).finishReason()));
|
||||
}
|
||||
|
||||
public int estimateTokenCount(List<ChatMessage> messages) {
|
||||
return this.tokenizer.estimateTokenCountInMessages(messages);
|
||||
}
|
||||
|
||||
public static FullOpenAiChatModel.FullOpenAiChatModelBuilder builder() {
|
||||
return new FullOpenAiChatModel.FullOpenAiChatModelBuilder();
|
||||
}
|
||||
|
||||
public static class FullOpenAiChatModelBuilder {
|
||||
|
||||
private String baseUrl;
|
||||
private String apiKey;
|
||||
private String modelName;
|
||||
private Double temperature;
|
||||
private Double topP;
|
||||
private List<String> stop;
|
||||
private Integer maxTokens;
|
||||
private Double presencePenalty;
|
||||
private Double frequencyPenalty;
|
||||
private Duration timeout;
|
||||
private Integer maxRetries;
|
||||
private Proxy proxy;
|
||||
private Boolean logRequests;
|
||||
private Boolean logResponses;
|
||||
private Tokenizer tokenizer;
|
||||
|
||||
FullOpenAiChatModelBuilder() {
|
||||
}
|
||||
|
||||
public FullOpenAiChatModel.FullOpenAiChatModelBuilder baseUrl(String baseUrl) {
|
||||
this.baseUrl = baseUrl;
|
||||
return this;
|
||||
}
|
||||
|
||||
public FullOpenAiChatModel.FullOpenAiChatModelBuilder apiKey(String apiKey) {
|
||||
this.apiKey = apiKey;
|
||||
return this;
|
||||
}
|
||||
|
||||
public FullOpenAiChatModel.FullOpenAiChatModelBuilder modelName(String modelName) {
|
||||
this.modelName = modelName;
|
||||
return this;
|
||||
}
|
||||
|
||||
public FullOpenAiChatModel.FullOpenAiChatModelBuilder temperature(Double temperature) {
|
||||
this.temperature = temperature;
|
||||
return this;
|
||||
}
|
||||
|
||||
public FullOpenAiChatModel.FullOpenAiChatModelBuilder topP(Double topP) {
|
||||
this.topP = topP;
|
||||
return this;
|
||||
}
|
||||
|
||||
public FullOpenAiChatModel.FullOpenAiChatModelBuilder stop(List<String> stop) {
|
||||
this.stop = stop;
|
||||
return this;
|
||||
}
|
||||
|
||||
public FullOpenAiChatModel.FullOpenAiChatModelBuilder maxTokens(Integer maxTokens) {
|
||||
this.maxTokens = maxTokens;
|
||||
return this;
|
||||
}
|
||||
|
||||
public FullOpenAiChatModel.FullOpenAiChatModelBuilder presencePenalty(Double presencePenalty) {
|
||||
this.presencePenalty = presencePenalty;
|
||||
return this;
|
||||
}
|
||||
|
||||
public FullOpenAiChatModel.FullOpenAiChatModelBuilder frequencyPenalty(Double frequencyPenalty) {
|
||||
this.frequencyPenalty = frequencyPenalty;
|
||||
return this;
|
||||
}
|
||||
|
||||
public FullOpenAiChatModel.FullOpenAiChatModelBuilder timeout(Duration timeout) {
|
||||
this.timeout = timeout;
|
||||
return this;
|
||||
}
|
||||
|
||||
public FullOpenAiChatModel.FullOpenAiChatModelBuilder maxRetries(Integer maxRetries) {
|
||||
this.maxRetries = maxRetries;
|
||||
return this;
|
||||
}
|
||||
|
||||
public FullOpenAiChatModel.FullOpenAiChatModelBuilder proxy(Proxy proxy) {
|
||||
this.proxy = proxy;
|
||||
return this;
|
||||
}
|
||||
|
||||
public FullOpenAiChatModel.FullOpenAiChatModelBuilder logRequests(Boolean logRequests) {
|
||||
this.logRequests = logRequests;
|
||||
return this;
|
||||
}
|
||||
|
||||
public FullOpenAiChatModel.FullOpenAiChatModelBuilder logResponses(Boolean logResponses) {
|
||||
this.logResponses = logResponses;
|
||||
return this;
|
||||
}
|
||||
|
||||
public FullOpenAiChatModel.FullOpenAiChatModelBuilder tokenizer(Tokenizer tokenizer) {
|
||||
this.tokenizer = tokenizer;
|
||||
return this;
|
||||
}
|
||||
|
||||
public FullOpenAiChatModel build() {
|
||||
return new FullOpenAiChatModel(this.baseUrl, this.apiKey, this.modelName, this.temperature,
|
||||
this.topP, this.stop, this.maxTokens, this.presencePenalty, this.frequencyPenalty,
|
||||
this.timeout, this.maxRetries, this.proxy, this.logRequests, this.logResponses, this.tokenizer);
|
||||
}
|
||||
|
||||
public String toString() {
|
||||
return "FullOpenAiChatModel.FullOpenAiChatModelBuilder(baseUrl=" + this.baseUrl
|
||||
+ ", apiKey=" + this.apiKey + ", modelName=" + this.modelName + ", temperature="
|
||||
+ this.temperature + ", topP=" + this.topP + ", stop=" + this.stop + ", maxTokens="
|
||||
+ this.maxTokens + ", presencePenalty=" + this.presencePenalty + ", frequencyPenalty="
|
||||
+ this.frequencyPenalty + ", timeout=" + this.timeout + ", maxRetries=" + this.maxRetries
|
||||
+ ", proxy=" + this.proxy + ", logRequests=" + this.logRequests + ", logResponses="
|
||||
+ this.logResponses + ", tokenizer=" + this.tokenizer + ")";
|
||||
}
|
||||
}
|
||||
}
|
||||
@@ -0,0 +1,66 @@
|
||||
package dev.langchain4j.model.openai;
|
||||
|
||||
import dev.ai4j.openai4j.chat.FunctionCall;
|
||||
import dev.ai4j.openai4j.chat.Message;
|
||||
import dev.ai4j.openai4j.chat.Role;
|
||||
import dev.langchain4j.data.message.AiMessage;
|
||||
import dev.langchain4j.data.message.ChatMessage;
|
||||
import dev.langchain4j.data.message.SystemMessage;
|
||||
import dev.langchain4j.data.message.ToolExecutionResultMessage;
|
||||
import dev.langchain4j.data.message.UserMessage;
|
||||
import dev.langchain4j.model.ChatModel;
|
||||
import java.util.List;
|
||||
import java.util.stream.Collectors;
|
||||
|
||||
public class ImproveInternalOpenAiHelper {
|
||||
|
||||
public ImproveInternalOpenAiHelper() {
|
||||
}
|
||||
|
||||
public static List<Message> toOpenAiMessages(List<ChatMessage> messages, String modelName) {
|
||||
List<Message> messageList = messages.stream()
|
||||
.map(message -> toOpenAiMessage(message, modelName)).collect(Collectors.toList());
|
||||
return messageList;
|
||||
}
|
||||
|
||||
public static Message toOpenAiMessage(ChatMessage message, String modelName) {
|
||||
return Message.builder().role(roleFrom(message, modelName))
|
||||
.name(nameFrom(message)).content(message.text())
|
||||
.functionCall(functionCallFrom(message)).build();
|
||||
}
|
||||
|
||||
private static String nameFrom(ChatMessage message) {
|
||||
if (message instanceof UserMessage) {
|
||||
return ((UserMessage) message).name();
|
||||
} else {
|
||||
return message instanceof ToolExecutionResultMessage
|
||||
? ((ToolExecutionResultMessage) message).toolName() : null;
|
||||
}
|
||||
}
|
||||
|
||||
private static FunctionCall functionCallFrom(ChatMessage message) {
|
||||
if (message instanceof AiMessage) {
|
||||
AiMessage aiMessage = (AiMessage) message;
|
||||
if (aiMessage.toolExecutionRequest() != null) {
|
||||
return FunctionCall.builder().name(aiMessage.toolExecutionRequest().name())
|
||||
.arguments(aiMessage.toolExecutionRequest().arguments()).build();
|
||||
}
|
||||
}
|
||||
|
||||
return null;
|
||||
}
|
||||
|
||||
public static Role roleFrom(ChatMessage message, String modelName) {
|
||||
if (modelName.contains(ChatModel.ZHIPU.toString()) || modelName.contains(ChatModel.ALI.toString())) {
|
||||
return Role.USER;
|
||||
}
|
||||
if (message instanceof AiMessage) {
|
||||
return Role.ASSISTANT;
|
||||
} else if (message instanceof ToolExecutionResultMessage) {
|
||||
return Role.FUNCTION;
|
||||
} else {
|
||||
return message instanceof SystemMessage ? Role.SYSTEM : Role.USER;
|
||||
}
|
||||
}
|
||||
|
||||
}
|
||||
Reference in New Issue
Block a user