U h,$@sddlmZddlZddlZddlmZddlmZmZm Z m Z m Z m Z ddl mZddlmZddlmZddlmZdd lmZmZmZGd d d eZdS) ) annotationsN)Path)AnyDictListOptionalTupleUnion)CallbackManagerForRetrieverRun)Document) SecretStr) BaseRetriever)convert_to_secret_strget_from_dict_or_envpre_initc@seZdZUdZded<dZded<GdddZeddd ddddZ e dddddZ d?ddddd dddZd d d!d"d#Zd$d%d d&d'd(Zd)d d*d+d,Zd$d$d d-d.d/Zd0d d1d2d3Zd$d4dd5d6d7d8Zd$d d9d:d;ZdS)@NeuralDBRetrieverz0Document retriever that uses ThirdAI's NeuralDB.r thirdai_keyNrdbc@seZdZdZdZdS)zNeuralDBRetriever.ConfigZforbidTN)__name__ __module__ __qualname__extraZunderscore_attrs_are_privaterrS/tmp/pip-unpacked-wheel-9gdii04g/langchain_community/retrievers/thirdai_neuraldb.pyConfigsrz Optional[str]None)rreturncCsRz0ddlm}tjd||p*tdWntk rLtdYnXdS)Nr) licensingzthirdai.neural_db THIRDAI_KEYz{Could not import thirdai python package and neuraldb dependencies. Please install it with `pip install thirdai[neural_db]`.) thirdair importlibutil find_specactivateosgetenv ImportError)rrrrr_verify_thirdai_librarys  z)NeuralDBRetriever._verify_thirdai_librarydict)r model_kwargsrcKs*t|ddlm}|||jf|dS)a Create a NeuralDBRetriever from scratch. To use, set the ``THIRDAI_KEY`` environment variable with your ThirdAI API key, or pass ``thirdai_key`` as a named parameter. Example: .. code-block:: python from langchain_community.retrievers import NeuralDBRetriever retriever = NeuralDBRetriever.from_scratch( thirdai_key="your-thirdai-key", ) retriever.insert([ "/path/to/doc.pdf", "/path/to/doc.docx", "/path/to/doc.csv", ]) documents = retriever.invoke("AI-driven music therapy") r neural_dbrr)rr'rr+NeuralDB)clsrr)ndbrrr from_scratch*s  zNeuralDBRetriever.from_scratchzUnion[str, Path]) checkpointrrcCs*t|ddlm}|||j|dS)a! Create a NeuralDBRetriever with a base model from a saved checkpoint To use, set the ``THIRDAI_KEY`` environment variable with your ThirdAI API key, or pass ``thirdai_key`` as a named parameter. Example: .. code-block:: python from langchain_community.retrievers import NeuralDBRetriever retriever = NeuralDBRetriever.from_checkpoint( checkpoint="/path/to/checkpoint.ndb", thirdai_key="your-thirdai-key", ) retriever.insert([ "/path/to/doc.pdf", "/path/to/doc.docx", "/path/to/doc.csv", ]) documents = retriever.invoke("AI-driven music therapy") rr*r,)rr'rr+r-from_checkpoint)r.r1rr/rrrr2Ls  z!NeuralDBRetriever.from_checkpointr)valuesrcCstt|dd|d<|S)z'Validate ThirdAI environment variables.rr)rr)r.r3rrrvalidate_environmentsosz'NeuralDBRetriever.validate_environmentsTz List[Any]bool)sourcestrain fast_modekwargsrcKs(||}|jjf|||d|dS)asInserts files / document sources into the retriever. Args: train: When True this means that the underlying model in the NeuralDB will undergo unsupervised pretraining on the inserted files. Defaults to True. fast_mode: Much faster insertion with a slight drop in performance. Defaults to True. )r6r7Zfast_approximationN)_preprocess_sourcesrinsert)selfr6r7r8r9rrrr;{s zNeuralDBRetriever.insertlist)r6rcCsddlm}|s|Sg}|D]}t|ts6||q|drV|||q|drv|||q|dr|| |qt d|dq|S)zChecks if the provided sources are string paths. If they are, convert to NeuralDB document objects. Args: sources: list of either string paths to PDF, DOCX or CSV files, or NeuralDB document objects. rr*z.pdfz.docxz.csvzCould not automatically load z. Only files with .pdf, .docx, or .csv extensions can be loaded automatically. For other formats, please use the appropriate document object from the ThirdAI library.) rr+ isinstancestrappendlowerendswithZPDFZDOCXZCSV RuntimeError)r<r6r/Zpreprocessed_sourcesdocrrrr:s"    z%NeuralDBRetriever._preprocess_sourcesr?int)query document_idrcCs|j||dS)a!The retriever upweights the score of a document for a specific query. This is useful for fine-tuning the retriever to user behavior. Args: query: text to associate with `document_id` document_id: id of the document to associate query with. N)rZtext_to_result)r<rFrGrrrupvoteszNeuralDBRetriever.upvotezList[Tuple[str, int]])query_id_pairsrcCs|j|dS)aGiven a batch of (query, document id) pairs, the retriever upweights the scores of the document for the corresponding queries. This is useful for fine-tuning the retriever to user behavior. Args: query_id_pairs: list of (query, document id) pairs. For each pair in this list, the model will upweight the document id for the query. N)rZtext_to_result_batch)r<rIrrr upvote_batchs zNeuralDBRetriever.upvote_batch)sourcetargetrcCs|j||dS)a=The retriever associates a source phrase with a target phrase. When the retriever sees the source phrase, it will also consider results that are relevant to the target phrase. Args: source: text to associate to `target`. target: text to associate `source` to. N)r associate)r<rKrLrrrrMs zNeuralDBRetriever.associatezList[Tuple[str, str]]) text_pairsrcCs|j|dS)a.Given a batch of (source, target) pairs, the retriever associates each source phrase with the corresponding target phrase. Args: text_pairs: list of (source, target) text pairs. For each pair in this list, the source will be associated with the target. N)rassociate_batch)r<rNrrrrOsz!NeuralDBRetriever.associate_batchr zList[Document])rF run_managerr9rc Ksnz6d|krd|d<|jjfd|i|}dd|DWStk rh}ztd||W5d}~XYnXdS)zRetrieve {top_k} contexts with your retriever for a given query Args: query: Query to submit to the model top_k: The max number of context results to retrieve. Defaults to 10. Ztop_k rFc Ss8g|]0}t|j|j|j|j|j|j|dddqS))id upvote_idsrKmetadatascorecontext)Z page_contentrU)r textrSrTrKrUrVrW).0refrrr s z=NeuralDBRetriever._get_relevant_documents..z"Error while retrieving documents: N)rsearch Exception ValueError)r<rFrPr9Z referenceserrr_get_relevant_documentss  z)NeuralDBRetriever._get_relevant_documents)pathrcCs|j|dS)zSaves a NeuralDB instance to disk. Can be loaded into memory by calling NeuralDB.from_checkpoint(path) Args: path: path on disk to save the NeuralDB instance to. N)rsave)r<rarrrrbszNeuralDBRetriever.save)N)N)N)TT)rrr__doc____annotations__rr staticmethodr' classmethodr0r2rr4r;r:rHrJrMrOr`rbrrrrrs0   !"     r) __future__rr r$pathlibrtypingrrrrrr Zlangchain_core.callbacksr Zlangchain_core.documentsr Zlangchain_core.pydantic_v1r Zlangchain_core.retrieversr Zlangchain_core.utilsrrrrrrrrs