Teradata Package for LangChain Function Reference - as_retriever - Teradata® Package for LangChain - Look here for syntax, methods and examples for the functions included in the Teradata langchain-teradata package.

Teradata® Package for LangChain Function Reference

Deployment
VantageCloud
Edition
Enterprise
Product
Teradata® Package for LangChain
Release Number
20.00.00.01
Published
December 2025
ft:locale
en-US
ft:lastEdition
2025-12-19
dita:id
Langchain-Teradata_FxRef_Lake
Product Category
Teradata Vantage
libs.teradata.langchain_teradata.TeradataVectorStore.as_retriever = as_retriever(self, **kwargs)
DESCRIPTION:
    Creates and returns a TeradataVectorStoreRetriever 
    instance that can be used to retrieve relevant documents 
    from the vector store based on similarity search.
    Note: Applicable for content-based , file-based and 
          embedding-based(only if metadata_columns is specified)
          vector stores.
 
PARAMETERS:
    search_type
        Optional Argument
        Specifies the type of search that the Retriever should perform.
        Default Value: "similarity"
        Permitted Values: "similarity", "similarity_score_threshold"
        Types: str
        Note:
            * "similarity_score_threshold" will be supported in the future release.
    
    search_kwargs
        Optional Argument
        Specifies additional parameters for the search operation.
        Includes the following keys:
            * top_k
                Optional Argument
                Specifies the number of top clusters to be considered while searching
                Types: int
 
            * score_threshold
                Optional Argument. Required when search_type is "similarity_score_threshold".
                Specifies the threshold value to consider for matching tables/views
                while searching. A higher threshold value limits responses 
                to the top matches only.
                Types: float          
 
            * search_numcluster:
                Optional Argument.
                Number of clusters or fraction of train_numcluster to be considered while searching.
                Notes:
                    Applicable when "search_algorithm" is 'KMEANS'.
                    If you want to pass a fraction of train_numcluster to be used for searching,
                    the supported range is (0, 1.0].
                    If you want to pass the exact number of clusters to be used for searching,
                    the supported range is [1, train_numcluster].
                Types: int or float
 
            * ef_search:
                Optional Argument.
                Specify the number of neighbors to consider during search in HNSW graph.
                Note:
                    Applicable when "search_algorithm" is 'HNSW'.
                Permitted Values: [1 - 1024]
                Types: int
 
            * filter
                Specifies the filter conditions to be applied to the document metadata.
                Types: str
        Types: dict
 
RETURNS:
    TeradataVectorStoreRetriever.
 
RAISES:
    ValueError.
 
EXAMPLES:
    # Load necessary imports and data
    >>> from langchain_teradata import TeradataVectorStore
    >>> from teradatagenai import load_data
    >>> amazon_data = load_data("amazon", "amazon_reviews_25")
 
    # Note this step is not needed if vector store already exists
 
    # Create an instance of a content-based vector store for
    # the data in table 'amazon_reviews_25'.
    >>> vs = TeradataVectorStore.from_datasets(name = "test_vs",
                                               data = "amazon_reviews_25",
                                               data_columns = "rev_text",
                                               key_columns = ["rev_id", "aid"],
                                               metadata_columns = ["rev_name"])
 
    # Example 1: Create a basic instance of the TeradataVectorStoreRetriever.
    #            Instantiate an already present vector store.
    >>> td_vs = TeradataVectorStore(name="test_vs")
    >>> retriever = td_vs.as_retriever()
 
    # Example 2: Create an instance of the TeradataVectorStoreRetriever 
    #            with a search_type of "similarity_score_threshold",
    #            a threshold of 0.8.
    >>> td_vs = TeradataVectorStore(name="test_Vs")
    >>> retriever = td_vs.as_retriever(search_type="similarity_score_threshold",
                                       search_kwargs={'score_threshold': 0.8})
 
    # Example 3: Create an instance of the TeradataVectorStoreRetriever
    #            with a search_type of "similarity", a top_k of 5,
    #            and a filter condition on metadata.
    >>> td_vs = TeradataVectorStore(name="test_Vs")
    >>> retriever = td_vs.as_retriever(search_type="mmr",
                                       search_kwargs={"top_k":5, 
                                                      "filter" : "rev_name LIKE 'A%'"})