upgrade to gpt-4o-mini

Author: nobodyiam

pom.xml

@@ -6,7 +6,7 @@
   <parent>
     <groupId>org.springframework.boot</groupId>
     <artifactId>spring-boot-starter-parent</artifactId>
-    <version>3.1.3</version>
+    <version>3.3.2</version>
     <relativePath/> <!-- lookup parent from repository -->
   </parent>
   <groupId>com.apolloconfig.apollo.ai</groupId>
@@ -16,10 +16,10 @@
   <description>a smart qa bot</description>
   <properties>
     <java.version>17</java.version>
-    <openai-gpt3-java.version>0.16.0</openai-gpt3-java.version>
-    <guava.version>32.1.2-jre</guava.version>
+    <openai-gpt3-java.version>0.18.2</openai-gpt3-java.version>
+    <guava.version>33.2.1-jre</guava.version>
     <flexmark.version>0.64.8</flexmark.version>
-    <milvus.version>2.3.0</milvus.version>
+    <milvus.version>2.4.2</milvus.version>
   </properties>
 
   <dependencyManagement>

src/main/java/com/apolloconfig/apollo/ai/qabot/openai/OpenAiService.java

@@ -17,8 +17,9 @@
 @Component
 class OpenAiService implements AiService {
 
-  private static final String DEFAULT_MODEL = "gpt-3.5-turbo";
+  private static final String DEFAULT_MODEL = "gpt-4o-mini";
   private static final String DEFAULT_EMBEDDING_MODEL = "text-embedding-ada-002";
+  private static final int DEFAULT_MAX_TOKENS = 5000;
 
   private final com.theokanning.openai.service.OpenAiService service;
 
@@ -37,7 +38,7 @@ public Flowable<ChatCompletionChunk> getCompletionFromMessages(List<ChatMessage>
 
   public Flowable<ChatCompletionChunk> getCompletionFromMessages(List<ChatMessage> messages,
       double temperature) {
-    return getCompletionFromMessages(messages, DEFAULT_MODEL, temperature, 500);
+    return getCompletionFromMessages(messages, DEFAULT_MODEL, temperature, DEFAULT_MAX_TOKENS);
   }
 
   public Flowable<ChatCompletionChunk> getCompletionFromMessages(List<ChatMessage> messages,

src/main/resources/application.yaml

@@ -59,7 +59,7 @@ milvus:
 
 qa:
   # the topK number of chunks retrieved from milvus database
-  topK: 3
+  topK: 5
   # the prompt used to generate the answer
   prompt: |
     You are an AI assistant for answering questions for apollo, which is a very popular configuration management system suitable for microservice configuration management scenarios.