k i:dZddlZddlmZmZmZmZmZddlm Z GddZ ddded ee d e fd Z Gd d e Z dded ee d e fdZdS)aAsync embedding vectorizer adapter for Redis middleware. This module provides vectorizer adapters for Redis middleware: 1. AsyncEmbeddingVectorizer: A simple async wrapper for embedding functions 2. LangChainVectorizer: A redisvl-compatible BaseVectorizer that wraps LangChain embeddings The LangChainVectorizer is the recommended approach for using LangChain embeddings (OpenAI, Azure OpenAI, etc.) with SemanticCache and SemanticCacheMiddleware. N)AnyCallableListOptionalUnion)BaseVectorizerc leZdZdZdddeeegeeefdede eegeefddfdZ e defd Z d e eeefdeeefd Zd e eeefdeeefd Zd edeefdZd edeefdZdS)AsyncEmbeddingVectorizeraLAsync wrapper for synchronous embedding functions. This class wraps synchronous embedding functions (like those from langchain-openai) for use in async contexts using asyncio.to_thread(). Example: ```python from langchain_openai import OpenAIEmbeddings embeddings = OpenAIEmbeddings() vectorizer = AsyncEmbeddingVectorizer( embed_fn=embeddings.embed_documents, dims=1536 # OpenAI ada-002 dimensions ) # Use in async context vectors = await vectorizer.aembed(["hello world"]) ``` N)embed_query_fnembed_fndimsr returnc0||_||_||_dS)aYInitialize the async vectorizer. Args: embed_fn: Synchronous function that embeds a list of texts. dims: Dimensionality of the embedding vectors. embed_query_fn: Optional separate function for single query embedding. If not provided, embed_fn is used with a single-item list. N) _embed_fn_embed_query_fn_dims)selfr r r s C:\Users\Dell Inspiron 16\Desktop\tws\AgrotaPowerBi\back-agrota-powerbi\mcp-client-agrota\venv\Lib\site-packages\langgraph/middleware/redis/vectorizer.py__init__z!AsyncEmbeddingVectorizer.__init__'s"- c|jS)z Return the embedding dimensions.)r)rs rr zAsyncEmbeddingVectorizer.dims:s zrtextsc\t|tr|g}||S)zSynchronously embed texts. Args: texts: A single text or list of texts to embed. Returns: List of embedding vectors. ) isinstancestrrrrs rembedzAsyncEmbeddingVectorizer.embed?s/ eS ! ! GE~~e$$$rcvKt|tr|g}tj|j|d{VS)zAsynchronously embed texts using asyncio.to_thread(). Args: texts: A single text or list of texts to embed. Returns: List of embedding vectors. N)rrasyncio to_threadrrs raembedzAsyncEmbeddingVectorizer.aembedLsH eS ! ! GE&t~u=========rtextcr|j||S||gdS)zSynchronously embed a single query text. Args: text: The query text to embed. Returns: The embedding vector. Nr)rr)rr"s r embed_queryz$AsyncEmbeddingVectorizer.embed_queryYs:   +''-- -~~tf%%a((rcK|j tj|j|d{VStj|j|gd{V}|dS)zAsynchronously embed a single query text. Args: text: The query text to embed. Returns: The embedding vector. Nr)rrr r)rr"results r aembed_queryz%AsyncEmbeddingVectorizer.aembed_queryfsk   + *4+?FFFFFFFF F($@@@@@@@@ayr)__name__ __module__ __qualname____doc__rrrfloatintrrpropertyr rrr!r$r'rrr r sz2BF DI;T%[(99: !3%e*   &cX %5d3i0 %T$u+5F % % % % >%T#Y"7 >De > > > ) )U ) ) ) ) s tE{      rr )r embeddingsr rcL|'dD]$}t||rt||}n%|?t|ddpt|dd}|rdddd}||}|td|j}t|d d}t ||| S) a_Create an AsyncEmbeddingVectorizer from a LangChain Embeddings object. Args: embeddings: A LangChain Embeddings object (e.g., OpenAIEmbeddings). dims: Optional dimensions override. If not provided, will attempt to determine from the embeddings object or use a default. Returns: An AsyncEmbeddingVectorizer wrapping the LangChain embeddings. Example: ```python from langchain_openai import OpenAIEmbeddings embeddings = OpenAIEmbeddings() vectorizer = create_vectorizer_from_langchain(embeddings, dims=1536) ``` N dimensionsr embedding_dim dimensionmodel model_name ztext-embedding-ada-002ztext-embedding-3-smallztext-embedding-3-largeSCould not determine embedding dimensions. Please provide dims parameter explicitly.r$)r r r )hasattrgetattrget ValueErrorembed_documentsr )r0r attrr6 model_dimsr r s r create_vectorizer_from_langchainrCus, |H  Dz4(( z400   | GT22 g  d7 7   )+/*.*.J >>%((D | 8   )HZ==N # %   rceZdZUdZdZeed<ddiZdededed dffd Z dd e d e e de ded eeeeff dZ ddee d e e dede ded eeeeeeff dZ dd e d e e de ded eeeeff dZ ddee d e e dede ded eeeeeeff dZxZS)LangChainVectorizera`A redisvl-compatible vectorizer that wraps LangChain Embeddings. This class bridges LangChain embeddings (OpenAI, Azure OpenAI, HuggingFace, etc.) with redisvl's SemanticCache by implementing the BaseVectorizer interface. Unlike AsyncEmbeddingVectorizer, this class properly extends redisvl's BaseVectorizer (a Pydantic model), making it compatible with SemanticCache's type validation. Example: ```python from langchain_openai import OpenAIEmbeddings from langgraph.middleware.redis import LangChainVectorizer, SemanticCacheConfig # Create LangChain embeddings embeddings = OpenAIEmbeddings(model="text-embedding-3-small") # Wrap in LangChainVectorizer for redisvl compatibility vectorizer = LangChainVectorizer(embeddings=embeddings, dims=1536) # Use with SemanticCacheMiddleware config = SemanticCacheConfig( redis_url="redis://localhost:6379", vectorizer=vectorizer, ) ``` For Azure OpenAI: ```python from langchain_openai import AzureOpenAIEmbeddings embeddings = AzureOpenAIEmbeddings( azure_deployment="your-embedding-deployment", openai_api_version="2024-02-15-preview", ) vectorizer = LangChainVectorizer(embeddings=embeddings, dims=1536) ``` N _embeddingsarbitrary_types_allowedTr0r kwargsrc xtjdd|d|t|d|dS)aNInitialize the LangChain vectorizer. Args: embeddings: A LangChain Embeddings object (e.g., OpenAIEmbeddings, AzureOpenAIEmbeddings, HuggingFaceEmbeddings). dims: The dimensionality of the embedding vectors. **kwargs: Additional arguments passed to BaseVectorizer. langchain)r6r rFNr/)superrobject __setattr__)rr0r rH __class__s rrzLangChainVectorizer.__init__sG @{@@@@@4 ;;;;;rFr" preprocess as_bufferc | ||}t|jdr|j|}n!|j|gd}|r||d|jS|S)aEmbed a single text using LangChain embeddings. Args: text: The text to embed. preprocess: Optional preprocessing function (applied if provided). as_buffer: If True, return as bytes buffer. **kwargs: Additional arguments (ignored). Returns: The embedding vector as a list of floats, or bytes if as_buffer=True. Nr$rTrPdtype)r<rFr$r@_process_embeddingrS)rr"rOrPrH embeddings rrzLangChainVectorizer.embeds$  !:d##D 4#] 3 3 D(44T::II(88$@@CI  X**9DJ*WW Wrr batch_sizec g}|||D]A}j|}|rfd|D}||B|S)aEmbed multiple texts using LangChain embeddings. Args: texts: List of texts to embed. preprocess: Optional preprocessing function (applied to each text). batch_size: Batch size for processing (used for batching). as_buffer: If True, return as bytes buffers. **kwargs: Additional arguments (ignored). Returns: List of embedding vectors, or list of bytes if as_buffer=True. cJg|]}|dj STrRrTrS.0ers r z2LangChainVectorizer.embed_many..$@++ATZ+PPr)batchifyrFr@extend rrrOrWrPrHall_embeddingsbatchr0s ` r embed_manyzLangChainVectorizer.embed_many s(-/]]5*jAA . .E)99%@@J '   ! !* - - - -rcK| ||}t|jdr!|j|d{V}nat|jdr*|j|gd{V}|d}n"t j|j|ddd{V}|r||d|jS|S)akAsynchronously embed a single text. Args: text: The text to embed. preprocess: Optional preprocessing function. as_buffer: If True, return as bytes buffer. **kwargs: Additional arguments (ignored). Returns: The embedding vector as a list of floats, or bytes if as_buffer=True. Nr'aembed_documentsrFTrR) r<rFr'rhrr rrTrS)rr"rOrPrHrUr0s rr!zLangChainVectorizer.aembed,s$  !:d##D 4#^ 4 4 O".;;DAAAAAAAAII T%'9 : : O#/@@$HHHHHHHHJ"1 II&/ D$NNNNNNNNI  X**9DJ*WW WrcDKg}|||D]}tjdr!j|d{V}n%t jjj|d{V}|rfd|D}|||S)aAsynchronously embed multiple texts. Args: texts: List of texts to embed. preprocess: Optional preprocessing function. batch_size: Batch size for processing. as_buffer: If True, return as bytes buffers. **kwargs: Additional arguments (ignored). Returns: List of embedding vectors, or list of bytes if as_buffer=True. rhNcJg|]}|dj SrZr[r\s rr_z3LangChainVectorizer.aembed_many..pr`r)rar<rFrhrr r@rbrcs ` r aembed_manyzLangChainVectorizer.aembed_manyOs(-/]]5*jAA . .Et');<< #'#3#D#DU#K#KKKKKKK $+#4$4e$$  '   ! !* - - - -r)NF)NrVF)r(r)r*r+rFr__annotations__ model_configr-rrrrboolrrr,bytesrrfr!rk __classcell__)rNs@rrErEsI%%PK-t4L <3 r?rE)r0r rAvalr6rBs rcreate_langchain_vectorizerrsys> |H  Dz4(( j$//?DE | GT22 g  d7 7   )+/*.*.J >>%((D | 8   *4 @ @ @@r)N)r+rtypingrrrrrredisvl.utils.vectorize.baserr r-rCrErsr/rrrvsH  77777777777777777777````````H/39999&sm99999xEEEEE.EEER,0<A<A<A#C=<A<A<A<A<A<A<Ar