NgddlmZddlmZmZmZmZmZddlZ ddl m Z ddl m Z ddlmZddlmZddlmZdd lmZdd lmZGd d eeZGd deZdS))Enum)AnyDictListOptionalUnionN)CallbackManagerForRetrieverRun)Document) Embeddings) BaseRetriever) get_fields) ConfigDict)maximal_marginal_relevanceceZdZdZdZdZdS) SearchTypez-Enumerator of the types of search to perform. similaritymmrN)__name__ __module__ __qualname____doc__rrc/var/www/html/ai-engine/env/lib/python3.11/site-packages/langchain_community/retrievers/docarray.pyrrs77J CCCrrc ~eZdZUdZdZeed<eed<eed<eed<e j Z e ed<dZ e ed <dZeeed <ed Zd ededeefdZdejd e deeeeefeffdZdejdeefdZdejdeefdZdeeeefefdefdZdS)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_managerreturnc:tj|j|}|jt jkr||}nC|jt jkr| |}ntd|jd|S)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_documents5sHT_88??@@  z4 4 4--i88GG   / /&&y11GG8t/888  rr1c$ddlm}m}i}|j}t |j|r |j|d<d}n*t |j|r |j|d<n |j|d<|jr|j||j d i| |}|j |}t|d r|j }|d |}n"|j||| j }|S) a Perform a search using the query embedding and return top_k documents. Args: query_emb: Query represented as an embedding top_k: Number of documents to return Returns: A list of top_k documents matching the query r)ElasticDocIndexWeaviateDocumentIndex where_filterr& filter_query)r&r)limit documentsN)r&rr:r)docarray.indexr5r6r isinstancerr$ build_queryfindfilterbuild execute_queryhasattrr;) r0r1r#r5r6 filter_argsrr&docss r_searchzDocArrayRetriever._searchQsZ JIIIIIII ( dj"7 8 8 7*.,K 'LL  O 4 4 7#'z8DocArrayRetriever._similarity_search..s'HHHC422377HHHr)rFr#)r0r1rEr2s` rr-z$DocArrayRetriever._similarity_searchs:||itz|BBHHHH4HHHrc|dt|fdDj}fd|D}|S)z Perform a maximal marginal relevance (mmr) search. Args: query_emb: Query represented as an embedding Returns: A list of diverse documents related to the query rHc~g|]9}t|tr |jnt|j:Sr)r=dictrgetattrrLs rrOz1DocArrayRetriever._mmr_search..sW   c4((5D%&&S$"344   r)kcFg|]}|SrrJ)rMidxrEr0s rrOz1DocArrayRetriever._mmr_search..s+VVV#42249==VVVr)rFrr#)r0r1 mmr_selectedr2rEs` @rr.zDocArrayRetriever._mmr_searchs||ir|::1          j    WVVVVVVVrrNc t|tr|nt|}|j|vrt d|jdt t|tr ||jnt||j}|D]l}t|tr||nt||}t|ttttfr||jkr ||j |<m|S)a; Convert a DocArray document (which also might be a dict) to a langchain document format. DocArray document can contain arbitrary fields, so the mapping is done in the following way: page_content <-> 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=rSkeysr r r/r rTstrintfloatboolmetadata)r0rNfieldslc_docnamevalues rrKz,DocArrayRetriever._docarray_to_langchain_docs, *#t44I*S//  V + +VASVVV #t$$2T/00d011     . .D!+C!6!6NCIIGC    RZDN2*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.utilsrr]rrrrrrss533333333333333CCCCCC------000000333333444444MMMMMMdzzzzz zzzzzr