feat(mes): 支持工艺流程工序产出产品多选功能

云

2026-08-04 cfbddbd52f8c4d35b06c8b7e7e68b6cf22ce400a

 yudao-module-ai/src/main/java/cn/iocoder/yudao/module/ai/framework/ai/core/model/AiModelFactoryImpl.java

@@ -7,22 +7,20 @@
import cn.hutool.extra.spring.SpringUtil;
import cn.iocoder.yudao.module.ai.enums.model.AiPlatformEnum;
import cn.iocoder.yudao.module.ai.framework.ai.config.AiAutoConfiguration;
import cn.iocoder.yudao.module.ai.framework.ai.config.YudaoAiProperties;
import lombok.extern.slf4j.Slf4j;
import org.springframework.ai.chat.model.ChatModel;
import org.springframework.ai.embedding.BatchingStrategy;
import org.springframework.ai.embedding.EmbeddingModel;
import org.springframework.ai.embedding.TokenCountBatchingStrategy;
import org.springframework.ai.openai.OpenAiChatModel;
import org.springframework.ai.vectorstore.VectorStore;
import org.springframework.ai.vectorstore.milvus.MilvusVectorStore;
import org.springframework.ai.vectorstore.milvus.autoconfigure.MilvusServiceClientProperties;
import org.springframework.ai.vectorstore.milvus.autoconfigure.MilvusVectorStoreProperties;
import java.util.Map;

/**
 * AI Model 模型工厂实现类
 *
 * 使用 OpenAI 兼容接口对接通义千问 + Milvus 向量存储
 * 使用 OpenAI 兼容接口对接通义千问 + 可配置向量存储（Milvus）
 */
@Slf4j
public class AiModelFactoryImpl implements AiModelFactory {
@@ -61,7 +59,6 @@
    public VectorStore getOrCreateVectorStore(Class<? extends VectorStore> type,
                                              EmbeddingModel embeddingModel,
                                              Map<String, Class<?>> metadataFields) {
        // metadataFields 参与缓存 key，确保不同知识库使用不同配置时不会复用
        String cacheKey = buildCacheKey(VectorStore.class, embeddingModel, type, metadataFields.hashCode());
        return Singleton.get(cacheKey, (Func0<VectorStore>) () -> {
            if (type == MilvusVectorStore.class) {
@@ -72,20 +69,20 @@
    }

    private MilvusVectorStore buildMilvusVectorStore(EmbeddingModel embeddingModel) {
        MilvusVectorStoreProperties serverProperties = SpringUtil.getBean(MilvusVectorStoreProperties.class);
        MilvusServiceClientProperties clientProperties = SpringUtil.getBean(MilvusServiceClientProperties.class);
        YudaoAiProperties aiProperties = SpringUtil.getBean(YudaoAiProperties.class);
        YudaoAiProperties.VectorStore.Milvus milvusConfig = aiProperties.getVectorStore().getMilvus();

        var connectParam = io.milvus.param.ConnectParam.newBuilder()
                .withHost(clientProperties.getHost())
                .withPort(clientProperties.getPort())
                .withDatabaseName(serverProperties.getDatabaseName())
                .withHost(milvusConfig.getHost())
                .withPort(milvusConfig.getPort())
                .withDatabaseName(milvusConfig.getDatabaseName())
                .build();
        var milvusClient = new io.milvus.client.MilvusServiceClient(connectParam);

        MilvusVectorStore vectorStore = MilvusVectorStore.builder(milvusClient, embeddingModel)
                .databaseName(serverProperties.getDatabaseName())
                .collectionName(serverProperties.getCollectionName())
                .initializeSchema(serverProperties.isInitializeSchema())
                .databaseName(milvusConfig.getDatabaseName())
                .collectionName(milvusConfig.getCollectionName())
                .initializeSchema(milvusConfig.isInitializeSchema())
                .batchingStrategy(new TokenCountBatchingStrategy())
                .build();
        try {