feat(ai): 调整模型

This commit is contained in:
v-zhangjc9
2025-06-13 19:07:13 +08:00
parent bc32a89fea
commit f11f5e7656
5 changed files with 92 additions and 20 deletions

View File

@@ -9,7 +9,7 @@ spring:
api-key: ENC(K+Hff9QGC+fcyi510VIDd9CaeK/IN5WBJ9rlkUsHEdDgIidW+stHHJlsK0lLPUXXREha+ToQZqqDXJrqSE+GUKCXklFhelD8bRHFXBIeP/ZzT2cxhzgKUXgjw3S0Qw2R)
chat:
options:
model: 'Qwen3-1.7-vllm'
model: 'Qwen3/qwen3-1.7b'
mvc:
async:
request-timeout: 3600000

View File

@@ -1,7 +1,9 @@
package com.lanyuanxiaoyao.service.ai.chat;
import cn.hutool.core.util.StrUtil;
import java.io.IOException;
import java.net.http.HttpClient;
import java.time.Duration;
import java.util.Arrays;
import java.util.List;
import org.noear.solon.ai.rag.Document;
@@ -25,12 +27,12 @@ import org.springframework.web.reactive.function.client.WebClient;
* @author lanyuanxiaoyao
* @version 20250514
*/
public class TestSpringAIToolChat {
public class TestModel {
public static void main(String[] args) throws IOException {
testChatModel();
// testChatModel();
testVisualModel();
// testEmbeddingModel();
// testRerankingModel();
testEmbeddingModel();
testRerankingModel();
}
private static void testChatModel() {
@@ -41,12 +43,14 @@ public class TestSpringAIToolChat {
"Qwen3/qwen3-4b-q4km",
"Qwen3/qwen3-8b-q4km"
)) {
System.out.println(model);
long start = System.currentTimeMillis();
ChatClient client = chatClient(model);
String content = client.prompt()
.user("你好")
.user("你好,详细介绍一下是谁,能帮我做什么?")
.call()
.content();
System.out.println(content);
System.out.println(content.length() * 1000.0 / (System.currentTimeMillis() - start));
}
}
@@ -55,19 +59,17 @@ public class TestSpringAIToolChat {
"Qwen2.5/qwen2.5-vl-7b",
"Qwen2.5/qwen2.5-vl-7b-q4km",
"Qwen2.5/qwen2.5-vl-3b-instruct",
"Qwen2.5/qwen2.5-vl-3b-instruct-awq",
"Qwen2.5/qwen2.5-vl-7b-instruct",
"Qwen2.5/qwen2.5-vl-7b-instruct-awq",
"MiniCPM/minicpm-o-2.6-7.6b",
"MiniCPM/minicpm-o-2.6-7.6b-q4km"
)) {
ChatClient client = chatClient(model);
String content = client.prompt()
.user(spec -> spec.text("图片中有什么").media(MimeTypeUtils.IMAGE_PNG, new FileSystemResource("/Users/lanyuanxiaoyao/Pictures/deepseek.png")))
.user(spec -> spec.text("根据图片中的内容编一个童话小故事").media(MimeTypeUtils.IMAGE_PNG, new FileSystemResource("/Users/lanyuanxiaoyao/Pictures/deepseek.png")))
.call()
.content();
System.out.println(content);
System.out.println(StrUtil.trim(content));
}
}
@@ -78,9 +80,7 @@ public class TestSpringAIToolChat {
"Qwen3/qwen3-embedding-4b-q4km",
"Qwen3/qwen3-embedding-8b-q4km",
"BGE/bge-m3",
"BGE/bge-m3-q4km",
"MiniCPM/minicpm-embedding",
"MiniCPM/minicpm-embedding-light"
"BGE/bge-m3-q4km"
)) {
EmbeddingModel embeddingModel = embeddingModel(model);
float[] worlds = embeddingModel.embed("Hello world");
@@ -90,12 +90,10 @@ public class TestSpringAIToolChat {
private static void testRerankingModel() throws IOException {
for (String model : List.of(
"BGE/beg-reranker-v2",
"MiniCPM/minicpm-reranker",
"MiniCPM/minicpm-reranker-light",
"BGE/beg-reranker-v2",
"BGE/beg-reranker-v2-q4km"
)) {
System.out.println(model);
RerankingModel rerankingModel = rerankingModel(model);
List<Document> list = rerankingModel.rerank(
"你好",
@@ -149,6 +147,7 @@ public class TestSpringAIToolChat {
return RerankingModel.of("http://132.121.206.65:10086/v1/rerank")
.model(model)
.apiKey("*XMySqV%>hR&v>>g*NwCs3tpQ5FVMFEF2VHVTj<MYQd$&@$sY7CgqNyea4giJi4")
.timeout(Duration.ofMinutes(10))
.build();
}

View File

@@ -0,0 +1,73 @@
package com.lanyuanxiaoyao.service.ai.chat;
import cn.hutool.core.util.StrUtil;
import java.net.http.HttpClient;
import org.springframework.ai.chat.client.ChatClient;
import org.springframework.ai.openai.OpenAiChatModel;
import org.springframework.ai.openai.OpenAiChatOptions;
import org.springframework.ai.openai.api.OpenAiApi;
import org.springframework.ai.tool.annotation.Tool;
import org.springframework.ai.tool.annotation.ToolParam;
import org.springframework.http.client.JdkClientHttpRequestFactory;
import org.springframework.http.client.reactive.JdkClientHttpConnector;
import org.springframework.web.client.RestClient;
import org.springframework.web.reactive.function.client.WebClient;
import reactor.core.Disposable;
/**
* @author lanyuanxiaoyao
* @version 20250613
*/
public class TestSpringAiTools {
public static void main(String[] args) {
ChatClient client = ChatClient.builder(
OpenAiChatModel.builder()
.openAiApi(
OpenAiApi.builder()
.baseUrl("http://132.121.206.65:10086")
.apiKey("*XMySqV%>hR&v>>g*NwCs3tpQ5FVMFEF2VHVTj<MYQd$&@$sY7CgqNyea4giJi4")
.restClientBuilder(restClientBuilder())
.webClientBuilder(webClientBuilder())
.build()
)
.defaultOptions(
OpenAiChatOptions.builder()
.model("Qwen3/qwen3-1.7b")
.build()
)
.build()
)
.build();
Disposable disposable = client.prompt()
.tools(new TestTool())
.user("调用submit工具生成一个关于「猪」的笑话")
.stream()
.content()
.subscribe(System.out::println);
while (!disposable.isDisposed()) {
}
}
private static HttpClient httpClient() {
return HttpClient.newBuilder()
.version(HttpClient.Version.HTTP_1_1)
.build();
}
private static RestClient.Builder restClientBuilder() {
return RestClient.builder()
.requestFactory(new JdkClientHttpRequestFactory(httpClient()));
}
private static WebClient.Builder webClientBuilder() {
return WebClient.builder()
.clientConnector(new JdkClientHttpConnector(httpClient()));
}
public static final class TestTool {
@Tool(description = "传入任意动物名称,返回一个关于这个动物的笑话")
public String submit(@ToolParam(description = "动物名称") String animalName) {
return StrUtil.format("{}掉沟里了", animalName);
}
}
}

View File

@@ -145,7 +145,7 @@ public class LlamaSwapTool extends GeneratorTool {
arguments.add("--reasoning-parser deepseek_r1");
}
return new DockerCmd(
"vllm-server-cpu:0.9.1",
"vllm-server-cpu:0.8.5.post1",
name,
model,
StrUtil.format("http://vllm-{}:${PORT}", displayName(model)),

View File

@@ -9,10 +9,10 @@ spring:
api-key: ENC(K+Hff9QGC+fcyi510VIDd9CaeK/IN5WBJ9rlkUsHEdDgIidW+stHHJlsK0lLPUXXREha+ToQZqqDXJrqSE+GUKCXklFhelD8bRHFXBIeP/ZzT2cxhzgKUXgjw3S0Qw2R)
chat:
options:
model: 'Qwen3-1.7-vllm'
model: 'Qwen3/qwen3-1.7b'
embedding:
options:
model: 'Bge-m3-vllm'
model: 'Qwen3/qwen3-embedding-4b'
vectorstore:
qdrant:
host: 132.121.206.65