feat(knowledge): 实现外部直接插入知识库

service-ai/service-ai-knowledge/src/main/java/com/lanyuanxiaoyao/service/ai/knowledge/controller/KnowledgeBaseController.java

@@ -94,7 +94,7 @@ public class KnowledgeBaseController {
 
   @PostMapping("submit_text")
   public void submitText(
-      @RequestParam(value = "id") Long id,
+      @RequestParam("id") Long id,
       @RequestParam(value = "mode", defaultValue = "NORMAL") String mode,
       @RequestParam(value = "type", defaultValue = "text") String type,
       @RequestParam(value = "content", required = false) String content,
@@ -109,6 +109,16 @@ public class KnowledgeBaseController {
     }
   }
 
+  @PostMapping("submit_text_directly")
+  public void submitDirectly(
+      @RequestParam("id") Long id,
+      @RequestParam(value = "name", required = false) String name,
+      @RequestParam(value = "split_key", defaultValue = "\n\n") String splitKey,
+      @RequestBody String content
+  ) {
+    embeddingService.submitDirectly(id, name, Lists.immutable.of(content.split(splitKey)));
+  }
+
   @PostMapping("query")
   public ImmutableList<String> query(
       @RequestParam("id") Long id,

service-ai/service-ai-knowledge/src/main/java/com/lanyuanxiaoyao/service/ai/knowledge/service/EmbeddingService.java

@@ -9,6 +9,7 @@ import com.lanyuanxiaoyao.service.ai.knowledge.entity.Knowledge;
 import com.lanyuanxiaoyao.service.ai.knowledge.entity.vo.DataFileVO;
 import com.yomahub.liteflow.core.FlowExecutor;
 import java.nio.charset.StandardCharsets;
+import java.util.HashMap;
 import java.util.List;
 import java.util.concurrent.ExecutorService;
 import java.util.concurrent.Executors;
@@ -104,4 +105,32 @@ public class EmbeddingService {
       }
     });
   }
+
+  public void submitDirectly(Long id, String name, ImmutableList<String> contents) {
+    executors.submit(() -> {
+      Knowledge knowledge = knowledgeBaseService.get(id);
+      String groupName = name;
+      if (StrUtil.isBlank(groupName)) {
+        groupName = StrUtil.format("外部-{}", IdUtil.nanoId(10));
+      }
+      Long groupId = groupService.add(knowledge.getId(), groupName);
+      EmbeddingContext context = EmbeddingContext.builder()
+          .vectorSourceId(knowledge.getVectorSourceId())
+          .groupId(groupId)
+          .build();
+      context.setDocuments(
+          contents
+              .collect(StrUtil::trim)
+              .collect(content ->
+                  Document.builder()
+                      .text(content)
+                      .metadata(new HashMap<>())
+                      .build()
+              )
+              .toList()
+      );
+      executor.execute2Resp("embedding_submit_directly", null, context);
+      groupService.finish(groupId);
+    });
+  }
 }

service-ai/service-ai-knowledge/src/main/java/com/lanyuanxiaoyao/service/ai/knowledge/service/node/EmbeddingNodes.java

@@ -195,6 +195,7 @@ public class EmbeddingNodes {
         .call()
         .content();
     Assert.notBlank(response, "LLM response is empty");
+    logger.info("LLM response: \n{}", response);
     // noinspection DataFlowIssue
     return Arrays.stream(StrUtil.trim(response).split("---"))
         .map(text -> text.replaceAll("(?!^.+) +$", ""))
@@ -209,6 +210,8 @@ public class EmbeddingNodes {
   @LiteflowMethod(value = LiteFlowMethodEnum.PROCESS, nodeId = "import_vector_source", nodeName = "导入向量库", nodeType = NodeTypeEnum.COMMON)
   public void importVectorSource(NodeComponent node) {
     EmbeddingContext context = node.getContextBean(EmbeddingContext.class);
+    Assert.notNull(context.getVectorSourceId(), "VectorSourceId is null");
+    Assert.notNull(context.getGroupId(), "GroupId is null");
     if (ObjectUtil.isNotEmpty(context.getDocuments())) {
       VectorStore vs = QdrantVectorStore.builder(qdrantClient, embeddingModel)
           .collectionName(String.valueOf(context.getVectorSourceId()))

service-ai/service-ai-knowledge/src/main/resources/config/flow.xml

@@ -20,4 +20,7 @@
     <chain id="embedding_submit">
         SER(embedding_preview, import_vector_source)
     </chain>
+    <chain id="embedding_submit_directly">
+        SER(import_vector_source)
+    </chain>
 </flow>