Class MilvusV2EmbeddingStore.Builder

java.lang.Object
dev.langchain4j.store.embedding.milvus.v2.MilvusV2EmbeddingStore.Builder
Enclosing class:
MilvusV2EmbeddingStore
public static class MilvusV2EmbeddingStore.Builder extends Object

  • Constructor Details

    • Builder

      public Builder()

  • Method Details

    • milvusClient

      public MilvusV2EmbeddingStore.Builder milvusClient(io.milvus.v2.client.MilvusClientV2 milvusClientV2)

    • host

      public MilvusV2EmbeddingStore.Builder host(String host)
      Parameters:
      host - The host of the self-managed Milvus instance. Default value: "localhost".
      Returns:
      builder

    • port

      public MilvusV2EmbeddingStore.Builder port(Integer port)
      Parameters:
      port - The port of the self-managed Milvus instance. Default value: 19530.
      Returns:
      builder

    • collectionName

      public MilvusV2EmbeddingStore.Builder collectionName(String collectionName)
      Parameters:
      collectionName - The name of the Milvus collection. If there is no such collection yet, it will be created automatically. Default value: "default".
      Returns:
      builder

    • dimension

      public MilvusV2EmbeddingStore.Builder dimension(Integer dimension)
      Parameters:
      dimension - The dimension of the embedding vector. (e.g. 384) Mandatory if a new collection should be created.
      Returns:
      builder

    • indexType

      public MilvusV2EmbeddingStore.Builder indexType(io.milvus.v2.common.IndexParam.IndexType indexType)
      Parameters:
      indexType - The type of the index. Default value: FLAT.
      Returns:
      builder

    • metricType

      public MilvusV2EmbeddingStore.Builder metricType(io.milvus.v2.common.IndexParam.MetricType metricType)
      Parameters:
      metricType - The type of the metric used for similarity search. Default value: COSINE.
      Returns:
      builder

    • uri

      public MilvusV2EmbeddingStore.Builder uri(String uri)
      Parameters:
      uri - The URI of the managed Milvus instance. (e.g. "https://xxx.api.gcp-us-west1.zillizcloud.com")
      Returns:
      builder

    • token

      public MilvusV2EmbeddingStore.Builder token(String token)
      Parameters:
      token - The token (API key) of the managed Milvus instance.
      Returns:
      builder

    • username

      public MilvusV2EmbeddingStore.Builder username(String username)
      Parameters:
      username - The username. See details here.
      Returns:
      builder

    • password

      public MilvusV2EmbeddingStore.Builder password(String password)
      Parameters:
      password - The password. See details here.
      Returns:
      builder

    • consistencyLevel

      public MilvusV2EmbeddingStore.Builder consistencyLevel(io.milvus.v2.common.ConsistencyLevel consistencyLevel)
      Parameters:
      consistencyLevel - The consistency level used by Milvus. Default value: EVENTUALLY.
      Returns:
      builder

    • retrieveEmbeddingsOnSearch

      public MilvusV2EmbeddingStore.Builder retrieveEmbeddingsOnSearch(Boolean retrieveEmbeddingsOnSearch)
      Parameters:
      retrieveEmbeddingsOnSearch - During a similarity search in Milvus (when calling search()), the embedding itself is not retrieved. To retrieve the embedding, an additional query is required. Setting this parameter to "true" will ensure that embedding is retrieved. Be aware that this will impact the performance of the search. Default value: false.
      Returns:
      builder

    • autoFlushOnInsert

      public MilvusV2EmbeddingStore.Builder autoFlushOnInsert(Boolean autoFlushOnInsert)
      Parameters:
      autoFlushOnInsert - Whether to automatically flush after each insert (add(...) or addAll(...) methods). Default value: false. More info can be found here.
      Returns:
      builder

    • databaseName

      public MilvusV2EmbeddingStore.Builder databaseName(String databaseName)
      Parameters:
      databaseName - Milvus name of database. Default value: null. In this case default Milvus database name will be used.
      Returns:
      builder

    • idFieldName

      public MilvusV2EmbeddingStore.Builder idFieldName(String idFieldName)
      Parameters:
      idFieldName - the name of the field where the ID of the Embedding is stored. Default value: "id".
      Returns:
      builder

    • textFieldName

      public MilvusV2EmbeddingStore.Builder textFieldName(String textFieldName)
      Parameters:
      textFieldName - the name of the field where the text of the TextSegment is stored. Default value: "text".
      Returns:
      builder

    • metadataFieldName

      public MilvusV2EmbeddingStore.Builder metadataFieldName(String metadataFieldName)
      Parameters:
      metadataFieldName - the name of the field where the Metadata of the TextSegment is stored. Default value: "metadata".
      Returns:
      builder

    • vectorFieldName

      public MilvusV2EmbeddingStore.Builder vectorFieldName(String vectorFieldName)
      Parameters:
      vectorFieldName - the name of the field where the Embedding is stored. Default value: "vector".
      Returns:
      builder

    • sparseVectorFieldName

      public MilvusV2EmbeddingStore.Builder sparseVectorFieldName(String sparseVectorFieldName)
      Parameters:
      sparseVectorFieldName - the name of the field where the Embedding is stored. Default value: "sparse_vector".
      Returns:
      builder

    • ranker

      public MilvusV2EmbeddingStore.Builder ranker(MilvusV2Ranker ranker)
      Parameters:
      ranker - the component that combines and reorders the similarity scores from multiple ANN sub-searches (e.g., dense and sparse) into a single final ranking. Must be one of the ranker classes from io.milvus.v2.service.vector.request.ranker (e.g. new RRFRanker(60) or new WeightedRanker(...)). Default value: new RRFRanker(60).
      Returns:
      builder

    • sparseMetricType

      public MilvusV2EmbeddingStore.Builder sparseMetricType(io.milvus.v2.common.IndexParam.MetricType sparseMetricType)
      Parameters:
      sparseMetricType - The type of the metric used for sparse vector similarity search. Default value: IP.
      Returns:
      builer

    • sparseIndexType

      public MilvusV2EmbeddingStore.Builder sparseIndexType(io.milvus.v2.common.IndexParam.IndexType sparseIndexType)
      Parameters:
      sparseIndexType - The type of the index. Default value: SPARSE_INVERTED_INDEX.
      Returns:
      builder

    • sparseMode

      public MilvusV2EmbeddingStore.Builder sparseMode(MilvusV2EmbeddingStore.MilvusSparseMode sparseMode)
      Parameters:
      sparseMode - The mode of sparse vector generation. BM25 - use Milvus built-in sparse vector generation from text (query in search request must be provided). CUSTOM - user provides sparse vector (sparseEmbedding in search request must be provided). Default value: BM25.
      Returns:
      builder

    • searchMode

      public MilvusV2EmbeddingStore.Builder searchMode(MilvusV2EmbeddingStore.SearchMode searchMode)
      Parameters:
      searchMode - The search mode for this store. VECTOR - dense vector search only (default). HYBRID - hybrid dense + sparse search. Default value: VECTOR.
      Returns:
      builder

    • build

      public MilvusV2EmbeddingStore build()