o Zh@s6UddlmZddlZddlZddlZddlZddlmZddlm Z m Z m Z m Z m Z mZmZmZmZmZmZmZmZddlZddlmZddlmZddlmZddlmZdd l m!Z!e rcddl"Z"ed Z#e$ee#Z%d e&d <ed Z'e$ee'Z(de&d<dZ)dZ*dZ+e+dZ,gdZ-gdZ.ddigZ/e0e1Z2ddd Z3dd#d$Z4dd&d'Z5ddd.d/Z6ed0d1d2d3Gd4d5d5eZ7       dddBdCZ8       D  dddMdNZ9ddSdTZ:dUgdfddXdYZ;dd\d]ZddhdiZ?ddmdnZ@ddrdsZA D DdddwdxZBddzd{ZCdd~dZDdS)) annotationsN)deepcopy) TYPE_CHECKINGAnyCallableDictIterableListLiteralOptionalSizedTupleTypeUnionget_args) deprecated)Document) Embeddings) VectorStore)maximal_marginal_relevance)L2ZIPzList[DISTANCE_METRICS]AVAILABLE_DISTANCE_METRICS)Z TileDBDenseZ TileDBSparse FaissFlatZ FaissIVFFlatZFlinngz List[ENGINES]AVAILABLE_ENGINESZ langchain ) _distanceidcontent)rrblobzMissing propertyxryx_namestry_namereturnNonec CsZt|tr+t|tr+t|t|kr+t|d|d|dt|d|dt| dS)z Check that sizes of two variables are the same Args: x: Variable to compare y: Variable to compare x_name: Name for variable x y_name: Name for variable y z and z% expected to be equal length but len(z)=z and len(N) isinstancer len ValueError)r!r"r#r%r+\/var/www/html/lang_env/lib/python3.10/site-packages/langchain_community/vectorstores/vdms.py_len_check_if_sized?s$  r-resultsList[Document]cCsddt|DS)NcSg|]\}}|qSr+r+.0doc_r+r+r, Rz$_results_to_docs..)_results_to_docs_and_scores)r.r+r+r,_results_to_docsQsr8List[Tuple[Document, float]]c Csg}z^|d\}}t|dkrUd|dvrXd|ddvr[|ddd}|D]6}t|dd}|d}tD] }||vr?||=q6dd|D} |t|| d |fq'W|SW|SW|SW|Sty{} ztd | WYd} ~ |Sd} ~ ww) NrFindDescriptorentitiesr rcSsi|] \}}|tvr||qSr+)INVALID_METADATA_VALUE)r2ZmkeyZmvalr+r+r, fs z/_results_to_docs_and_scores..) page_contentmetadataz2No results returned. Error while parsing results: ) r)roundINVALID_DOC_METADATA_KEYSitemsappendr Exceptionloggerwarning) r.Z final_res responsesZblobsZresult_entitiesentZdistanceZ txt_contentsppropser+r+r,r7UsF    r7 localhosthostportint vdms.vdmscCs<zddl}Wn tytdw|}||||S)zVDMS client for the VDMS server. Args: host: IP or hostname of VDMS server port: Port to connect to VDMS server rNzOCould not import vdms python package. Please install it with `pip install vdms.)vdms ImportErrorconnect)rOrPrSclientr+r+r, VDMS_Clientws   rWz0.3.18z1.0.0zlangchain_vdms.VDMS)ZsinceZremovalZalternative_importc@seZdZdZdeddddddddZedddZdddZdd"d#Z dd%d&Z dd)d*Z dd,d-Z e edfdd5d6Z  ddd=d>Z  ddd@dAZ  dddFdGZ    dddOdPZ Q QdddUdVZgdQfdd]d^Z ddd`daZddddeZddedffddjdkZ    l fdddmdnZddefddpdqZdefddrdsZddudvZddwdxZdd{d|Z   ddd~dZ   QddddZ!e eddddQfdddZ"dddZ#e$ddeefdddZ%e$dddeefdddZ&dddHgfdddZ'e eddfdddZ(e eddfdddZ)e eddfdddZ*e eddfdddZ+dde edddQfdddZ,e edfdddZ-e edfdddZ.e edfdddZ/e edfdddZ0dddZ1dddZ2dS)VDMSaIntel Lab's VDMS for vector-store workloads. To use, you should have both: - the ``vdms`` python package installed - a host (str) and port (int) associated with a deployed VDMS Server Visit https://github.com/IntelLabs/vdms/wiki more information. IT IS HIGHLY SUGGESTED TO NORMALIZE YOUR DATA. Args: client: VDMS Client used to connect to VDMS server collection_name: Name of data collection [Default: langchain] distance_strategy: Method used to calculate distances. VDMS supports "L2" (euclidean distance) or "IP" (inner product) [Default: L2] engine: Underlying implementation for indexing and computing distances. VDMS supports TileDBDense, TileDBSparse, FaissFlat, FaissIVFFlat, and Flinng [Default: FaissFlat] embedding: Any embedding function implementing `langchain_core.embeddings.Embeddings` interface. relevance_score_fn: Function for obtaining relevance score Example: .. code-block:: python from langchain_huggingface import HuggingFaceEmbeddings from langchain_community.vectorstores.vdms import VDMS, VDMS_Client model_name = "sentence-transformers/all-mpnet-base-v2" vectorstore = VDMS( client=VDMS_Client("localhost", 55555), embedding=HuggingFaceEmbeddings(model_name=model_name), collection_name="langchain-demo", distance_strategy="L2", engine="FaissFlat", ) Nrr) embeddingcollection_namedistance_strategyenginerelevance_score_fnembedding_dimensionsrVrRrYOptional[Embeddings]rZr$r[DISTANCE_METRICSr\ENGINESr]"Optional[Callable[[float], float]]r^ Optional[int]r&r'cCsD||_||_||_||_|||||_|j||j|jd|_dS)Nr\metric)_clientsimilarity_search_enginer[rY_check_required_inputsoverride_relevance_score_fnadd_set_collection_name)selfrVrYrZr[r\r]r^r+r+r,__init__s   z VDMS.__init__cCs|jSN)rYrlr+r+r, embeddingsszVDMS.embeddingstexts List[str]List[List[float]]cCs,t|jtr |j|Sd}|d7}t|)Nz*Must provide `embedding` which is expectedz to be an Embeddings object)r(rYrZembed_documentsr*)rlrqp_strr+r+r,_embed_documentss  zVDMS._embed_documentspathskwargsrcKs4|jdurt|jdr|jjdd|i|Std)N embed_videorvz:Must provide `embedding` which has attribute `embed_video`r+)rYhasattrrxr*)rlrvrwr+r+r, _embed_videos zVDMS._embed_videouriscCs,|jdurt|jdr|jj|dStd)N embed_imager{z:Must provide `embedding` which has attribute `embed_image`)rYryr|r*)rlr{r+r+r, _embed_images zVDMS._embed_imagetext List[float]cCs t|jtr |j|Std)NzEMust provide `embedding` which is expected to be an Embeddings object)r(rYr embed_queryr*)rlrr+r+r, _embed_querys  zVDMS._embed_queryCallable[[float], float]cCs8|jdur|jS|jdvrddStd|jd)a8 The 'correct' relevance function may differ depending on a few things, including: - the distance / similarity metric used by the VectorStore - the scale of your embeddings (OpenAI's are unit normed. Many others are not!) - embedding dimensionality - etc. N)ipl2cSs|Srnr+)r!r+r+r,sz1VDMS._select_relevance_score_fn..z=No supported normalization function for distance_strategy of z;.Consider providing relevance_score_fn to VDMS constructor.)rir[lowerr*ror+r+r,_select_relevance_score_fns zVDMS._select_relevance_score_fnquerykrQfetch_kfilterOptional[Dict[str, Any]]r9c Ksp|jdur d|d<|jd||||d|}g}|D]\}} |jdur+||| fq|||| fq|S)z?Return docs and their similarity scores on a scale from 0 to 1.NTnormalize_distance)rrrrr+)risimilarity_search_with_scorerD) rlrrrrrwdocs_and_scoresZdocs_and_rel_scoresr3Zscorer+r+r,(_similarity_search_with_relevance_scores s(   z-VDMS._similarity_search_with_relevance_scoresrp metadatas1Optional[Union[List[None], List[Dict[str, Any]]]]idsOptional[List[str]]r cCst||dd|dur |ndd|D}t||dd|dur!|ndd|D}t||ddg}g}g}t||||D]%\} } } } |j|| | | | d\} }|dura|| |||| q<|||\}}|S) NrqrpcSg|]}dqSrnr+r2r4r+r+r,r54zVDMS.add..rcSg|]}ttqSr+r$uuiduuid4rr+r+r,r57rr@rYdocumentr)r-zip_VDMS__get_add_queryrD_VDMS__run_vdms_query)rlrZrqrprr all_queries all_blobs inserted_idsmetaembr3rrr responseresponse_arrayr+r+r,add*s&     zVDMS.addrec CsRtd||jt|d|t|d|d}||g\}}d|dvr'td||S)NAddDescriptorSetvaluerdZ FailedCommandrzFailed to add collection )_add_descriptorsetembedding_dimensiongetattrrr*)rlrZr\rerrr4r+r+r,rjKs   z VDMS.add_setUnion[None, List[str]] constraintsUnion[None, Dict[str, Any]]boolc Csg}g}||}d|i}|durdddgi}nddg|d<|dur*d|dg|d<td|ddddd||d }|||||\} } td |d d }||g|\} } d| dvS) zA Deletes entire collection if id is not provided listN _deletion==rrr:labelrefrKlink k_neighborsrr.FindDescriptorSetT) storeIndex)_VDMS__get_properties_add_descriptorrDrr) rlrZrrrrcollection_propertiesr.rrrZ responseSetr4r+r+r,Z__delete`s6     z VDMS.__deleter@ Optional[Any]Union[List[float], None]rr Optional[str]4Tuple[Dict[str, Dict[str, Any]], Union[bytes, None]]c Cs|duri}n5d|i}t|j||\}}|rsz(VDMS.__get_add_query..r:rz[!] Embedding with id (z) exists in DB;z#Therefore, skipped and not insertedz Skipped values are: )NrZ AddDescriptorr) _check_descriptor_exists_by_idrfrCprintupdatekeysrrDrembedding2bytes) rlrZr@rYrrrKZ id_existsrZ skipped_valuepstrrr r+r+r,Z__get_add_querysR      zVDMS.__get_add_queryF unique_entityOptional[bool]deletioncCsLt|||d}||g\}}t|dkr t|dd}|Stt}|S)N)rrr,)_find_property_entityrr) _bytes2strsplitrDEFAULT_PROPERTIES)rlrZrrZ find_queryrZ response_blobrr+r+r,Z__get_propertiess zVDMS.__get_propertiesr List[Dict]rOptional[List]print_last_responseTuple[Any, Any]cCs2|j||\}}t||}|r|j||fSrn)rfr_check_valid_responser)rlrrrrrr4r+r+r,Z__run_vdms_querys   zVDMS.__run_vdms_query documentsc Cst||ddt||dd|dur|ndd|D}t||dd||}g}t||||D]O\}} } } d|ji} d d d gi} | durKd | g| d <td |ddddd| | d }||g\}}|j||| | | d\}}|dur||g|g\}}|| q0||||jdS)z Updates (find, delete, add) a collection based on id. If more than one collection returned with id, error occuers rrrpNcSrrnr+rr+r+r,r5rz!VDMS.__update..rrrrrrr:rr) r-rrrrrrrD_VDMS__update_properties)rlrZrrrpr orig_propsZ updated_idsrrr3rr.rrrrr r+r+r,Z__updatesJ        z VDMS.__updatecurrent_collection_propertiesnew_collection_propertiesc Csd|dur.t|}|D] }||vr||q ||kr0t|d|d\}}|||g\}} dSdSdS)Nr) command_typeall_properties)rrD_build_property_queryr) rlrZrrZold_collection_propertiesproprblob_arrrr4r+r+r,Z__update_propertiess  zVDMS.__update_propertiesTOptional[List[dict]] batch_sizeadd_pathc  sfdd|D}|r|rt|D] \}} | ||d<qn|r-g}|D] } |d| iq#|dur3|ndd|D}j|d} |durLdd|D}ndd|D}jd || |||d ||S) aRun more images through the embeddings and add to the vectorstore. Images are added as embeddings (AddDescriptor) instead of separate entity (AddImage) within VDMS to leverage similarity search capability Args: uris: List of paths to the images to add to the vectorstore. metadatas: Optional list of metadatas associated with the images. ids: Optional list of unique IDs. batch_size (int): Number of concurrent requests to send to the server. add_path: Bool to add image path as metadata Returns: List of ids from adding images into the vectorstore. csg|]}j|dqS)) image_path) encode_image)r2uriror+r,r5Jrz#VDMS.add_images..rNcSrr+rrr+r+r,r5Urr}cSg|]}iqSr+r+rr+r+r,r5[rcSg|]}t|qSr+_validate_vdms_propertiesr2mr+r+r,r5]r6rqrprrrr+) enumeraterDr~add_from) rlr{rrrrrwZ b64_textsmidxrrpr+ror, add_images1s0 zVDMS.add_imagesrc Ks|dur dd|D}|r|rt|D] \}} | ||d<qn|r/g}|D] } |d| iq%|dur5|ndd|D}|jd d|i|} |durQdd|D}|jd || |||d||S) aRun videos through the embeddings and add to the vectorstore. Videos are added as embeddings (AddDescriptor) instead of separate entity (AddVideo) within VDMS to leverage similarity search capability Args: paths: List of paths to the videos to add to the vectorstore. metadatas: Optional list of text associated with the videos. metadatas: Optional list of metadatas associated with the videos. ids: Optional list of unique IDs. batch_size (int): Number of concurrent requests to send to the server. add_path: Bool to add video path as metadata Returns: List of ids from adding videos into the vectorstore. NcSr)rr+rr+r+r,r5rz#VDMS.add_videos..Z video_pathcSrr+rrr+r+r,r5rrvcSrr+r+rr+r+r,r5rrr+)rrDrzr) rlrvrqrrrrrwrpathrpr+r+r, add_videosis0zVDMS.add_videos Iterable[str]cKsnt|}|durdd|D}||}|dur dd|D}ndd|D}|jd|||||d|}|S)aRun more texts through the embeddings and add to the vectorstore. Args: texts: List of strings to add to the vectorstore. metadatas: Optional list of metadatas associated with the texts. ids: Optional list of unique IDs. batch_size (int): Number of concurrent requests to send to the server. Returns: List of ids from adding the texts into the vectorstore. NcSrr+rrr+r+r,r5rz"VDMS.add_texts..cSrr+r+rr+r+r,r5rcSrr+rrr+r+r,r5r6rr+)rrur)rlrqrrrrwrprr+r+r, add_textss" zVDMS.add_textscKs||j}g}tdt||D]5} t| |t|} || | } || | } || | } |r5|| | }|j|j| | || d}||q||j||j|S)Nr)rprqrr) rrkranger)minrextendrr)rlrqrprrrrwrrZ start_idxZend_idxZ batch_textsZbatch_embedding_vectorsZ batch_idsZbatch_metadatasresultr+r+r,rs*       z VDMS.add_fromUnion[int, None]cCs|js td|jtvrtd|jtvrtd|jdur$td|dur,||_nD|jdur@t |jdr@t | d|_n0|jdurpt |jdsQt |jdrpt |jd rlz |jj j j|_Wntyktd wtd ||}t |d r|j|dS||_dS) Nz_VDMS client must be connected to a VDMS server.Please use VDMS_Client to establish a connectionz-distance_strategy must be either 'L2' or 'IP'z]engine must be either 'TileDBDense', 'TileDBSparse', 'FaissFlat', 'FaissIVFFlat', or 'Flinng'Must provide embedding functionrzThis is a sample sentence.r|rxmodelz>Embedding dimension needed. Please define embedding_dimensionsr)rfZ is_connectedr*r[rrgrrYrryr)rrZtoken_embeddingZ embedding_dimrrr)rlrZr^Z current_propsr+r+r,rhsN             zVDMS._check_required_inputsc CsXg}g}ddgd}td|dddddd|d }|||||\}}|dddS)Nrr)countrr:rrreturned)rrDr)rlrZrrr.rrrr+r+r,r's"  z VDMS.count base64_imagebytescCs t|Srn)base64 b64decode)rlrr+r+r, decode_image= zVDMS.decode_imageOptional[Dict]cKs"|dur|n|j}|j|||dS)zDelete by ID. These are the IDs in the vectorstore. Args: ids: List of ids to delete. Returns: Optional[bool]: True if deletion is successful, False otherwise, None if not implemented. N)rr)rk _VDMS__delete)rlrrZrrwnamer+r+r,delete@sz VDMS.deletesetnamer. normalize(Tuple[List[Dict[str, Any]], List, float]c Cs\d}d}t||||d}||g|\} } |r)|| dvr)| d|ddd}| | |fS)Nrr:)rr.rr;rr)rr) rlr rr.rrmax_dist command_strrrrr+r+r,get_k_candidatesSs zVDMS.get_k_candidatesrrOptional[dict]query_embeddingOptional[List[float]]r!Tuple[List[Dict[str, Any]], List]c Cs g} t|} | dur| | |dur!|j|||| |d\} } } n|dur+ddgi}nd|vr5dg|d<n d|dvrB|ddt||||d}||g\} } || dvrp| d|ddkrpdd| d|d D}nggfS|j|||| |d\} } } || dvs|| dvr| d|ddkrggfSg}| d|d D]}|d|vr||t||krnq|| d|d <t|| d|d<t||krd }t||r | dtjfvrd n| } t | d|d D]\}}|d | |d <|d | d|d |d <q| | fS) N)rrrrr.rrcSsg|]}|dqS)rr+)r2rIr+r+r,r5sz0VDMS.get_descriptor_response..r;z4Returned items < k_neighbors; Try increasing fetch_kg?r) rrDrrrr)rnpinfr)rlrr rrrr.rrrr rrrrZids_of_interestZ new_entitiesrIrtZent_idxr+r+r,get_descriptor_responsejsh              zVDMS.get_descriptor_responsercCsFt|d}|}t|dWdS1swYdS)Nrbzutf-8)openreadr b64encodedecode)rlrfr r+r+r,rs $zVDMS.encode_imagecls Type[VDMS]r/c Ks4|d}|j|dd|Ddd|D||||dS)aCreate a VDMS vectorstore from a list of documents. Args: collection_name (str): Name of the collection to create. documents (List[Document]): List of documents to add to vectorstore. embedding (Embeddings): Embedding function. Defaults to None. ids (Optional[List[str]]): List of document IDs. Defaults to None. batch_size (int): Number of concurrent requests to send to the server. Returns: VDMS: VDMS vectorstore. rVcSg|]}|jqSr+r?r2r3r+r+r,r5z'VDMS.from_documents..cSr#r+)r@r%r+r+r,r5r&)rVrqrrYrrrZ) from_texts)r!rrYrrrZrwrVr+r+r,from_documentss  zVDMS.from_documentsc KsB|d}||||d} |durdd|D}| j||||d| S)aHCreate a VDMS vectorstore from a raw documents. Args: texts (List[str]): List of texts to add to the collection. embedding (Embeddings): Embedding function. Defaults to None. metadatas (Optional[List[dict]]): List of metadatas. Defaults to None. ids (Optional[List[str]]): List of document IDs. Defaults to None. batch_size (int): Number of concurrent requests to send to the server. collection_name (str): Name of the collection to create. Returns: VDMS: VDMS vectorstore. rV)rZrYrVNcSrr+rrr+r+r,r5rz#VDMS.from_texts..)rqrrr)r) r!rqrYrrrrZrwrVZvdms_collectionr+r+r,r'szVDMS.from_textslimitincludec Cs~g}g}ddi}|dur||d<d|vr||}||d<d|vr%d|d <td |d||d } || |||\} } | | fS) aGets the collection. Get embeddings and their associated data from the data store. If no constraints provided returns all embeddings up to limit. Args: constraints: A dict used to filter results by. E.g. `{"color" : ["==", "red"], "price": [">", 4.00]}`. Optional. limit: The number of documents to return. Optional. include: A list of what to include in the results. Can contain `"embeddings"`, `"metadatas"`, `"documents"`. Ids are always included. Defaults to `["metadatas", "documents"]`. Optional. rrNr)r@rrpTr r:)rrr.)rrrDr) rlrZrr)r*rrr.rrrrr+r+r,gets(  zVDMS.getg? lambda_multfloatOptional[Dict[str, List]]c K|jdur tdtj|st|jdr||}n>tj|r1t|jdr1|j|gdd}n(tj|rGt|jdrG|j|gdd}nd |d }|d 7}|d 7}t||j |||||d } | SaRReturn docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: query (str): Query to look up. Text or path for image or video. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List of Documents selected by maximal marginal relevance. NzBFor MMR search, you must specify an embedding function oncreation.rr|r}rrxrv(Could not generate embedding for query ''.9If using path for image or video, verify embedding model 6has callable functions 'embed_image' or 'embed_video'.)r,r) rYr*osrisfileryrr~rz'max_marginal_relevance_search_by_vector) rlrrrr,rrwZembedding_vector error_msgdocsr+r+r,max_marginal_relevance_search9s,   z"VDMS.max_marginal_relevance_searchc  |j|g||gdd}t|dddkrgSdd|ddD}ttj|tjd|||dt|} fd dt| D} | S) HReturn docs selected using the maximal marginal relevance. Maximal marginal relevance optimizes for similarity to query AND diversity among selected documents. Args: embedding: Embedding to look up documents similar to. k: Number of Documents to return. Defaults to 4. fetch_k: Number of Documents to fetch to pass to MMR algorithm. lambda_mult: Number between 0 and 1 that determines the degree of diversity among the results with 0 corresponding to maximum diversity and 1 to minimum diversity. Defaults to 0.5. filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List of Documents selected by maximal marginal relevance. rrZ distancesrpquery_embeddings n_resultsrr*rrcSg|]}tt|qSr+r_bytes2embeddingr2rr+r+r,r5 z@VDMS.max_marginal_relevance_search_by_vector..Zdtyperr,csg|] \}}|vr|qSr+r+)r2irZ mmr_selectedr+r,r5s)query_collection_embeddingsr)rrarrayfloat32r8r rlrYrrr,rrwr.Zembedding_list candidatesZselected_resultsr+rKr,r8o,  z,VDMS.max_marginal_relevance_search_by_vectorc Kr/r0) rYr*r6rr7ryrr~rz2max_marginal_relevance_search_with_score_by_vector) rlrrrr,rrwrYr9r:r+r+r,(max_marginal_relevance_search_with_scores,   z-VDMS.max_marginal_relevance_search_with_scorec  r<) r=r>r?rrcSrBr+rCrEr+r+r,r5rFzKVDMS.max_marginal_relevance_search_with_score_by_vector..rGrHcs$g|]\}\}}|vr||fqSr+r+)r2rIrJsrKr+r,r5 s)rLr)rrrMrNr7rrOr+rKr,rRrQz7VDMS.max_marginal_relevance_search_with_score_by_vectorr@Optional[List[List[float]]]rA!List[Tuple[Dict[str, Any], List]]c Ksg} |dur |j}|dur| S|ddg} |dur&d| vr&|jd| vd}|D]} |jd||||||| d\} } | | | gq(| S)Nr*rrp)rr r:)rrrr.rr)rkr+rrrD)rlr@rZrArrr.rrwZ all_responsesr*Zqembrrr+r+r,rLs.   z VDMS.query_collection_embeddingscKs(|j|f|||d|}dd|DS)aRun similarity search with VDMS. Args: query (str): Query to look up. Text or path for image or video. k (int): Number of results to return. Defaults to 3. fetch_k (int): Number of candidates to fetch for knn (>= k). filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List[Document]: List of documents most similar to the query text. )rrrcSr0r+r+r1r+r+r,r5Pr6z*VDMS.similarity_search..)r)rlrrrrrwrr+r+r,similarity_search:szVDMS.similarity_searchcK$|jd|g|||d|}t|S)aReturn docs most similar to embedding vector. Args: embedding (List[float]): Embedding to look up documents similar to. k (int): Number of Documents to return. Defaults to 3. fetch_k (int): Number of candidates to fetch for knn (>= k). filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List of Documents most similar to the query vector. r@rArrNr+)rLr8rlrYrrrrwr.r+r+r,similarity_search_by_vectorRsz VDMS.similarity_search_by_vectorc Ks|jdur tdtj|st|jdr||}n>tj|r1t|jdr1|j|gdd}n(tj|rGt|jdrG|j|gdd}nd |d }|d 7}|d 7}t||j d|g|||d |}t |S)aERun similarity search with VDMS with distance. Args: query (str): Query to look up. Text or path for image or video. k (int): Number of results to return. Defaults to 3. fetch_k (int): Number of candidates to fetch for knn (>= k). filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List[Tuple[Document, float]]: List of documents most similar to the query text and cosine distance in float for each. Lower score represents more similarity. Nrrr|r}rrxr1r2r3r4r5rYr+) rYr*r6rr7ryrr~rzrLr7) rlrrrrrwrr9r.r+r+r,rms*   z!VDMS.similarity_search_with_scorecKrX)a6 Return docs most similar to embedding vector and similarity score. Args: embedding (List[float]): Embedding to look up documents similar to. k (int): Number of Documents to return. Defaults to 3. fetch_k (int): Number of candidates to fetch for knn (>= k). filter (Optional[Dict[str, str]]): Filter by metadata. Defaults to None. Returns: List[Tuple[Document, float]]: List of documents most similar to the query text. Lower score represents more similarity. rYNr+)rLr7rZr+r+r,&similarity_search_with_score_by_vectorsz+VDMS.similarity_search_with_score_by_vector document_idrcCs|||g|gS)zUpdate a document in the collection. Args: document_id (str): ID of the document to update. document (Document): Document to update. )update_documents)rlrZr]rr+r+r,update_documents zVDMS.update_documentcCs>dd|D}dd|D}||}|j|||||ddS)zUpdate a document in the collection. Args: ids (List[str]): List of ids of the document to update. documents (List[Document]): List of documents to update. cSr#r+r$r2rr+r+r,r5r&z)VDMS.update_documents..cSsg|]}t|jqSr+)rr@r`r+r+r,r5s )rrprN)ru _VDMS__update)rlrZrrrr@rpr+r+r,r^s   zVDMS.update_documents)rVrRrYr_rZr$r[r`r\rar]rbr^rcr&r')r&r_)rqrrr&rs)rvrrrwrr&rs)r{rrr&rs)rr$r&r)r&r) rr$rrQrrQrrrwrr&r9)NN) rZr$rqrrrprsrrrrr&r )rr)rZr$r\rarer`r&r$)rZr$rrrrr&r)NNNN) rZr$r@rrYrrrrrr&rFF)rZr$rrrrr&rr)rrrrrrr&rrn) rZr$rrrrrrrprsrrr&r')rZr$rr rrr&r')r{rrrrrrrrQrrrwrr&rr)NNNrT)rvrrrqrrrrrrrQrrrwrr&rr) rqrrrrrrrQrwrr&rr)rqrrrprsrrrrrrrQrwrr&rr)rZr$r^rr&r')rZr$r&rQ)rr$r&r)NNN) rrrZrrr rwrr&r)NNF) r r$rrcr.rrrrrr&r)rr$r r$rrQrrQrrr.rrrrrr&r)rr$r&r$)r!r"rr/rYr_rrrrQrZr$rwrr&rX)r!r"rqrrrYr_rrrrrrQrZr$rwrr&rX) rZr$rr r)rcr*rrr&r)rr$rrQrrQr,r-rr.rwrr&r/)rYrrrQrrQr,r-rr.rwrr&r/)rr$rrQrrQr,r-rr.rwrr&r9)rYrrrQrrQr,r-rr.rwrr&r9)r@rUrZrrArQrrQrrr.rrrrwrr&rV) rr$rrQrrQrr.rwrr&r/) rYrrrQrrQrr.rwrr&r/) rr$rrQrrQrr.rwrr&r9) rYrrrQrrQrr.rwrr&r9)rZr$r]r$rrr&r')rZr$rrrrr/r&r')3__name__ __module__ __qualname____doc__DEFAULT_COLLECTION_NAMErmpropertyrprurzr~rr DEFAULT_KDEFAULT_FETCH_Krrrjr rrrrarDEFAULT_INSERT_BATCH_SIZErrrrrhrrr rrr classmethodr(r'r+r;r8rSrRrLrWr[rr\r_r^r+r+r+r,rXs*       % $  0 ;   ; ; ; . % 8    L"+ 5 9 : 7 9 ,   1 ! rXrr rrrrcrKrrrrDict[str, Dict[str, Any]]c Csd|i} d|vr|r|| d<|dur|| d<|tvr|| d<d|vr*|dur*|| d<d|vr8|dur8t|| d<d|vrD|tvrD|| d <d|vrP|tvrP|| d <|| i} | S) NsetAddr_ref propertiesrZFindrrr.)r=rQ) rr rrrKrrrr.entityrr+r+r,rs"  rFr num_dimsr\rer rrDict[str, Any]c Cs|dkr?tdd||fDr?||d} |dur|| d<|dur$|| d<|dur,|| d<|difvr6|| d<|dur>|| d <n(|d kr`d |i} |rM|| d <| difvrW| | d <| dur_| | d<ntd||| i} | S)Nrcss|]}|duVqdSrnr+)r2varr+r+r, s z%_add_descriptorset..)r  dimensionsr\rerprqrrrnrrr.zUnknown command: )allr*) rr rsr\rerrKrrrr.rrrr+r+r,rs:   rrZrr Tuple[Dict[str, Any], bytes]cCsnt|dkr d|nd}d}i}d|d<d|d<d |i}d |d <||d <||d<t|}i}|||<||fS) NrrrZ AddEntityrqclassTr r zqueryable propertiestyper)r)join _str2bytes)rZrZall_properties_str querytyperrrK byte_datarr+r+r,_add_entity_with_blob?srfindrrc Csg}g}gd}||vrtdd||dkr-t|dd}||||fS|dkrHt||\}}||||||fS|dkrjt|dd }||t||\}}||||||fS) N)rrrz"[!] Invalid type. Choices are : {}rrT)rrr)r)rr*formatr|rrDr) rZrrrrrchoicesrrr+r+r,rUs*           rr rcCstj|dd}|S)NrNrG)rZ frombuffer)r rr+r+r,rDxsrDin_bytescCs|Srn)r)rr+r+r,r}srrrrrcCsttdd|DS)NcSsg|] }|D]}|qqSr+)r)r2qrr+r+r,r5sz(_get_cmds_from_query..)rrn)rr+r+r,_get_cmds_from_querysr List[dict]rcs,t|}ttotfdd|D}|S)Nc3s@|]}|dvodd|vod|ddkVqdS)rrNr+)r2cmdrr+r,rvs  z(_check_valid_response..)rr(rany)rrZcmd_list valid_resr+rr,rs rrVrTuple[bool, Any]c CsJdd|gi}td||dgddd}|g}||\}}t||}||fS)Nrrr:r)rrr)rrr) rVr rrZfindDescriptorrresr4rr+r+r,rs   rrYrUnion[bytes, None]cCs&d}|durtj|dd}|}|S)zConvert embedding to bytes.NrNrG)rrMtobytes)rYr rr+r+r,rs rrrrcCszd}i}d|d<|r||d<i}d|d<d|d<d g|d <||d <i}|r+d d g|d<d |g|d<||d<i}|||<|S)NZ FindEntityrqrzuniqueTr rrrrr.rrrr rr+)rZrrr~rrr.rrr+r+r,rs$   rin_strcCs t|Srn)r$encode)rr+r+r,r}rr}r@rcCs0i}|D]\}}t|ts||t|<q|Srn)rCr(rr$)r@Z new_metadatakeyrr+r+r,rs   r) r!rr"rr#r$r%r$r&r')r.rr&r/)r.rr&r9)rMrN)rOr$rPrQr&rR)NNNNNNN)rr$r r$rrrrcrKrrrrrcrrr.rr&rm) NNNNNNFNN)rr$r r$rsrcr\rrerrrcrKr rr rrrr r.r r&rt)rZr$rr r&ry) rZr$rr$rr rrcr&r)r rr&r)rrr&r$)rrr&rr)rrrrr&r)rVrRr r$rr$r&r)rYrr&rrb)rZr$rrrrr&rm)rr$r&r)r@rtr&r)E __future__rrloggingr6rcopyrtypingrrrrrr r r r r rrrnumpyrZlangchain_core._api.deprecationrZlangchain_core.documentsrZlangchain_core.embeddingsrZlangchain_core.vectorstoresrZ&langchain_community.vectorstores.utilsrrSr`rr__annotations__rarrgrkrirjrrBr= getLoggerrcrFr-r8r7rWrXrrrrrDrrrrrrr}rr+r+r+r,s <           " c ) 7 #