7|hddlmZddlmZmZmZmZmZddlZ ddl m Z ddl m Z ddlmZddlmZddlmZdd lmZdd lmZGd d eeZGd deZy))Enum)AnyDictListOptionalUnionN)CallbackManagerForRetrieverRun)Document) Embeddings) BaseRetriever) get_fields) ConfigDict)maximal_marginal_relevanceceZdZdZdZdZy) SearchTypez-Enumerator of the types of search to perform. similaritymmrN)__name__ __module__ __qualname____doc__rrf/var/www/html/test/engine/venv/lib/python3.12/site-packages/langchain_community/retrievers/docarray.pyrrs7J Crrc heZdZUdZdZeed<eed<eed<eed<e jZ e ed<dZ e ed <dZeeed <ed Zd ededeefdZdej.d e deeeeefeffdZdej.deefdZdej.deefdZdeeeefefdefdZy)DocArrayRetrievera`DocArray Document Indices` retriever. Currently, it supports 5 backends: InMemoryExactNNIndex, HnswDocumentIndex, QdrantDocumentIndex, ElasticDocIndex, and WeaviateDocumentIndex. Args: index: One of the above-mentioned index instances embeddings: Embedding model to represent text as vectors search_field: Field to consider for searching in the documents. Should be an embedding/vector/tensor. content_field: Field that represents the main content in your document schema. Will be used as a `page_content`. Everything else will go into `metadata`. search_type: Type of search to perform (similarity / mmr) filters: Filters applied for document retrieval. top_k: Number of documents to return Nindex embeddings search_field content_field search_typetop_kfiltersT)arbitrary_types_allowedquery run_managerreturncPtj|jj|}|jt j k(r|j|}|S|jt jk(r|j|}|Std|jd)zGet documents relevant for a query. Args: query: string to find relevant documents for Returns: List of relevant documents z Search type z5 does not exist. Choose either 'similarity' or 'mmr'.) nparrayr embed_queryr!rr_similarity_searchr _mmr_search ValueError)selfr&r' query_embresultss r_get_relevant_documentsz)DocArrayRetriever._get_relevant_documents5sHHT__88?@   z44 4--i8G   /&&y1G t//0178 rr1c^ddlm}m}i}|j}t |j |r|j |d<d}n5t |j |r|j |d<n|j |d<|j r|j jj||jd i|j|}|j j|}t|d r |j}|d |}|S|j j||| j}|S) a Perform a search using the query embedding and return top_k documents. Args: query_emb: Query represented as an embedding top_k: Number of documents to return Returns: A list of top_k documents matching the query r)ElasticDocIndexWeaviateDocumentIndex where_filterr& filter_query)r&r)limit documentsN)r&rr:r)docarray.indexr5r6r isinstancerr$ build_queryfindfilterbuild execute_queryhasattrr;) r0r1r#r5r6 filter_argsrr&docss r_searchzDocArrayRetriever._searchQs& J (( djj"7 8*.,,K 'L  O 4#'<c 4|j|d}t||Dcgc]7}t|tr||jnt ||j9c}|j }|Dcgc]}|j||}}|Scc}wcc}w)z Perform a maximal marginal relevance (mmr) search. Args: query_emb: Query represented as an embedding Returns: A list of diverse documents related to the query rH)k)rFrr=dictrgetattrr#rI)r0r1rErJ mmr_selectedidxr2s rr.zDocArrayRetriever._mmr_searchs||ir|:1    c4(D%%&S$"3"345  jj  JVV#42249=VV Ws content_field metadata <-> all other fields excluding tensors and embeddings (so float, int, string) Args: doc: DocArray document Returns: Document in langchain format Raises: ValueError: If the document doesn't contain the content field z.Document does not contain the content field - .) page_content) r=rNkeysr r r/r rOstrintfloatboolmetadata)r0rJfieldslc_docnamevalues rrIz,DocArrayRetriever._docarray_to_langchain_docs, *#t4*S/   V +@ASAS@TTUV #t$T//0d001   .D!+C!6CIGC RZZDN2*eDcNC4G.H*X*rr)enumrtypingrrrrrnumpyr*langchain_core.callbacksr langchain_core.documentsr langchain_core.embeddingsr langchain_core.retrieversr langchain_core.utils.pydanticr pydanticr&langchain_community.vectorstores.utilsrrVrrrrrrlsB33C-034Mdz zr