我在Java里使用langchain4j，搭建本地的RAG。

配置文件

langchain4j:
  community:
    dashscope:
      chat-model:
        api-key: sk-41be519a5cf24bf9a280a51b8348c3b7
        model-name: deepseek-v3
        #model_name: qwen-max
        #model-name: qwen-max-latest
      embedding-model:
        api-key: sk-41be519a5cf24bf9a280a51b8348c3b7
        model-name: text-embedding-v3

EmbeddingStore初始化的方法

@Bean
    public EmbeddingStore<TextSegment> initEmbeddingStore() {

        return PgVectorEmbeddingStore.builder()
            .table(pgConfig.getTable())
            .dropTableFirst(true)
            .createTable(true)
            .host(pgConfig.getHost())
            .port(pgConfig.getPort())
            .user(pgConfig.getUser())
            .password(pgConfig.getPassword())
            .dimension(1024)
            .database(pgConfig.getDatabase())
            .build();

    }

文件加载的方法

    public String load(@RequestParam(value = "maxSegmentSizeInChars",required = false,defaultValue = "200") int maxSegmentSizeInChars
            , @RequestParam(value = "maxOverlapSizeInChars",required = false ,defaultValue = "0") int maxOverlapSizeInChars) {

        List<Document> documents = FileSystemDocumentLoader.loadDocuments("D:\\work\\lecture-langchain-20250525\\documents");
        // EmbeddingStoreIngestor.ingest(documents,embeddingStore);
        EmbeddingStoreIngestor.builder().embeddingStore(embeddingStore)
                .embeddingModel(embeddingModel)
                .documentSplitter(new DocumentByLineSplitter(maxSegmentSizeInChars, maxOverlapSizeInChars))
                .build().ingest(documents);
        return "数据加载成功";
    }

查看pgvector中的vector字段的属性也是1024，但是提问时，提示ERROR: different vector dimensions 1024 and 384
这个是模型的问题吗。我使用本地ollama中的all-minilm时没有遇见此类问题