$jdZddlZddlZddlZddlmZmZddlmZm Z m Z ddl Z ddl m Z mZmZmZmZmZmZmZmZmZmZmZmZmZmZmZmZmZmZm Z m!Z!m"Z"ej#e$Z%GddeZ&Gd d e&Z'Gd d e&Z(Gd de&Z)Gdde&Z*Gdde&Z+Gdde&Z,Gdde&Z-de&fdZ.dS)an Embeddings abstraction for the memory system. Provides an interface for generating embeddings with different backends. The embedding dimension is auto-detected from the model at initialization. The database schema is automatically adjusted to match the model's dimension. Configuration via environment variables - see hindsight_api.config for all env var names. N)ABCabstractmethod)parse_qsurlparse urlunparse)DEFAULT_EMBEDDINGS_COHERE_MODELDEFAULT_EMBEDDINGS_GEMINI_MODEL DEFAULT_EMBEDDINGS_LITELLM_MODEL$DEFAULT_EMBEDDINGS_LITELLM_SDK_MODEL"DEFAULT_EMBEDDINGS_LOCAL_FORCE_CPUDEFAULT_EMBEDDINGS_LOCAL_MODEL*DEFAULT_EMBEDDINGS_LOCAL_TRUST_REMOTE_CODEDEFAULT_EMBEDDINGS_OPENAI_MODELDEFAULT_EMBEDDINGS_PROVIDERDEFAULT_LITELLM_API_BASEENV_EMBEDDINGS_COHERE_API_KEYENV_EMBEDDINGS_GEMINI_API_KEY"ENV_EMBEDDINGS_LITELLM_SDK_API_KEYENV_EMBEDDINGS_LOCAL_FORCE_CPUENV_EMBEDDINGS_LOCAL_MODEL&ENV_EMBEDDINGS_LOCAL_TRUST_REMOTE_CODEENV_EMBEDDINGS_OPENAI_API_KEYENV_EMBEDDINGS_OPENAI_BASE_URLENV_EMBEDDINGS_OPENAI_MODELENV_EMBEDDINGS_PROVIDERENV_EMBEDDINGS_TEI_URLENV_LLM_API_KEYceZdZdZeedefdZeedefdZ ed dZ ede ede e e fdZ dS) Embeddingsz Abstract base class for embedding generation. The embedding dimension is determined by the model and detected at initialization. The database schema is automatically adjusted to match the model's dimension. returncdS)zFReturn a human-readable name for this provider (e.g., 'local', 'tei').Nselfs g/home/rurouni/.hermes/hermes-agent/venv/lib/python3.11/site-packages/hindsight_api/engine/embeddings.py provider_namezEmbeddings.provider_name8  cdS)z6Return the embedding dimension produced by this model.Nr#r$s r& dimensionzEmbeddings.dimension>r(r)Nc KdS)z Initialize the embedding model asynchronously. This should be called during startup to load/connect to the model and avoid cold start latency on first encode() call. Nr#r$s r& initializezEmbeddings.initializeDs  r)textscdS)z Generate embeddings for a list of texts. Args: texts: List of text strings to encode Returns: List of embedding vectors (each is a list of floats) Nr#)r%r.s r&encodezEmbeddings.encodeNs  r)r!N)__name__ __module__ __qualname____doc__propertyrstrr'intr+r-listfloatr0r#r)r&r r 0s s   ^X  3   ^X    ^   DI  $tE{*;    ^    r)r ceZdZdZddedzdedefdZedefd Zede fd Z dd Z d e ede e e fd ZdS)LocalSTEmbeddingsz Local embeddings implementation using SentenceTransformers. Call initialize() during startup to load the model and avoid cold starts. The embedding dimension is auto-detected from the model. NF model_name force_cputrust_remote_codecZ|pt|_||_||_d|_d|_dS)a Initialize local SentenceTransformers embeddings. Args: model_name: Name of the SentenceTransformer model to use. Default: BAAI/bge-small-en-v1.5 force_cpu: Force CPU mode for local inference. Default: False trust_remote_code: Allow loading models with custom code (security risk). Required for some models with custom architectures. Default: False (disabled for security) N)rr=r>r?_model _dimension)r%r=r>r?s r&__init__zLocalSTEmbeddings.__init__ds2%F(F"!2 &*r)r!cdS)Nlocalr#r$s r&r'zLocalSTEmbeddings.provider_namewswr)c<|jtd|jSN4Embeddings not initialized. Call initialize() first.rB RuntimeErrorr$s r&r+zLocalSTEmbeddings.dimension{! ? "UVV Vr)cK|jdS ddlm}n#t$rtdwxYwtd|jddl}|jrd}tdnd} |j p2t|j do|j j }|rd}n4#t$r'}td |Yd}~nd}~wwxYwt!j5t!jd t& t!jd d t!jd d t)jd}|j}|t(j ||j|ddi|j|_||n#||wxYw dddn #1swxYwY|j|_td|jddS)zLoad the embedding model.Nr)SentenceTransformerzksentence-transformers is required for LocalSTEmbeddings. Install it with: pip install sentence-transformersz3Embeddings: initializing local provider with model cpuzEmbeddings: forcing CPU modempsz/Failed to detect GPU/MPS, falling back to CPU: ignore)categoryz%.*was not found in model state dict.*)messagez.*UNEXPECTED.* transformerslow_cpu_mem_usageF)device model_kwargsr?z-Embeddings: local provider initialized (dim: ))rAsentence_transformersrM ImportErrorloggerinfor=torchr>cuda is_availablehasattrbackendsrO Exceptionwarningwarningscatch_warningsfilterwarnings UserWarninglogging getLoggerlevelsetLevelERRORr? get_sentence_embedding_dimensionrB)r%rMr\rUhas_gpuetransformers_loggeroriginal_levels r&r-zLocalSTEmbeddings.initializes ; " F  A A A A A A A   E    [$/[[\\\   > VF KK6 7 7 7 7 F V*1133ENE22Xu~7I7V7V7X7X"!F V V VTQRTTUUUUUUUU V  $ & & = =  #H{ C C C C  #H6] ^ ^ ^ ^  #H6F G G G G#*"3N"C"C 06N  ( ( 7 7 7 =11O!"5u!=&*&<  $,,^<<<<#,,^<<<<<' = = = = = = = = = = = = = = =*+FFHH VDOVVVWWWWWsL.>AC DC;;DBG6 G9G6G&&G66G:=G:r.c|jtd|j|dd}d|DS)z Generate embeddings for a list of texts. Args: texts: List of text strings to encode Returns: List of embedding vectors NrHTF)convert_to_numpyshow_progress_barc6g|]}|Sr#)tolist.0embs r& z,LocalSTEmbeddings.encode..s 333 333r))rArJr0)r%r. embeddingss r&r0zLocalSTEmbeddings.encodesL ; UVV V[''X]'^^ 33 3333r))NFFr1)r2r3r4r5r7boolrCr6r'r8r+r-r9r:r0r#r)r&r<r<\s++3:++bf++++&sX3X ?X?X?X?XB4DI4$tE{*;444444r)r<c eZdZdZ ddededed ed ef d Zed efd Z ed efdZ deded e j fdZ ddZdeed eeefdZdS)RemoteTEIEmbeddingsaE Remote embeddings implementation using HuggingFace Text Embeddings Inference (TEI) HTTP API. TEI provides a high-performance inference server for embedding models. See: https://github.com/huggingface/text-embeddings-inference The embedding dimension is auto-detected from the server at initialization. >@ ?base_urltimeout batch_size max_retries retry_delayc|d|_||_||_||_||_d|_d|_d|_dS)a Initialize remote TEI embeddings client. Args: base_url: Base URL of the TEI server (e.g., "http://localhost:8080") timeout: Request timeout in seconds (default: 30.0) batch_size: Maximum batch size for embedding requests (default: 32) max_retries: Maximum number of retries for failed requests (default: 3) retry_delay: Initial delay between retries in seconds, doubles each retry (default: 0.5) /N) rstriprrrrr_client _model_idrB)r%rrrrrs r&rCzRemoteTEIEmbeddings.__init__sN$!,,  $&&,0 %)&*r)r!cdS)Nteir#r$s r&r'z!RemoteTEIEmbeddings.provider_namesur)c<|jtd|jSrGrIr$s r&r+zRemoteTEIEmbeddings.dimensionrKr)methodurlc ddl}d}|j}t|jdzD]V} |dkr|jj|fi|}n|jj|fi|}||cS#tj tj tj f$rc} | }||jkrLt d|dzd|jdzd| d|d |||d z}Yd} ~ d} ~ wtj$rv} | jjd krZ||jkrO| }t d |dzd|jdzd| d|d |||d z}nYd} ~ Pd} ~ wwxYw|) z@Make an HTTP request with automatic retries on transient errors.rNGETzTEI request failed (attempt rz): z. Retrying in zs...rizTEI server error (attempt )timerrangerrgetpostraise_for_statushttpx ConnectError ReadTimeout WriteTimeoutrZrbsleepHTTPStatusErrorresponse status_code) r%rrkwargsr last_errordelayattemptrrns r&_request_with_retryz'RemoteTEIEmbeddings._request_with_retrys   T-122  G U??/t|/>>v>>HH0t|0????H))+++&(95;MN    T---NNzw{zzTEUXYEYzz^_zzotzzzJJu%%%QJE(   :)S00Wt?O5O5O!"JNNxWq[xx4CSVWCWxx\]xxmrxxxJJu%%%QJEEEEEE s&AA..&E:AC22E:A+E55E:NcK|jdStd|jt j|j|_ |d|jd}|}| dd|_ d|vrd |vr |d |jd d d gi}|}|r-t|dkrt|d|_ td|j d|j ddS#tj $r}td|jd|d}~wwxYw)z:Initialize the HTTP client and verify server connectivity.Nz)Embeddings: initializing TEI provider at )rrz/infomodel_idunknownmax_input_length model_dtypePOST/embedinputstestjsonrz-Embeddings: TEI provider initialized (model: , dim: rWz#Failed to connect to TEI server at : )rrZr[rrClientrrrrrlenrB HTTPErrorrJ)r%rr[ test_responsetest_embeddingsrns r&r-zRemoteTEIEmbeddings.initialize's < # F O OOPPP|DL999  [//$-7N7N7NOOH==??D!XXj)<z/OpenAIEmbeddings.initialize..s" T T TTQAaD T T Tr) default_queryrrinputz0Embeddings: OpenAI provider initialized (model: rrWr#)rrrrYrrZr[rrrrrr_replaceritemsMODEL_DIMENSIONSrBrzcreatedatar embedding)r%r base_url_msg client_kwargsparsed clean_urlrrs r&r-zOpenAIEmbeddings.initializes < # F n % % % % % % % n n nlmm m n26F-dm---B  e4:eWceefff %)LAQRR = :dm,,F| :&vR'@'@AA ,5 j) T TXfl5K5K5Q5Q5S5S T T T 1> o. ) ,0M j)v.. ..  :. . ."3DJ?DOO|.55jh6H} B"%hmA&6&@"A"A ltzllZ^Zilllmmmmms.r.cf|jtd|sgSg}tdt||jD]p}||||jz}|jj|j|}t|j d}| d|Dq|S)z Generate embeddings using the OpenAI API. Args: texts: List of text strings to encode Returns: List of embedding vectors NrHrrc|jS)N)indexxs r&z)OpenAIEmbeddings.encode..s17r)keycg|] }|j Sr#rrwrns r&ryz+OpenAIEmbeddings.encode..s"I"I"I11;"I"I"Ir)) rrJrrrrzrrsortedrr)r%r.rrrrrs r&r0zOpenAIEmbeddings.encodes < UVV V Iq#e**do66 K KA!a$/112E|.55j6H &hm9J9JKKK   ! !"I"I8H"I"I"I J J J Jr)r1)r2r3r4r5rrr7r8rCr6r'r+r-r9r:r0r#r)r&rrms'#'"&"&5# ++++* +  +  ++++4sX3X )n)n)n)nVDI$tE{*;r)rceZdZdZdddddddZedddd d fd ed ed edzdedzdededefdZ e defdZ e defdZ ddZ deedeeefdZdS)CohereEmbeddingsz Cohere embeddings implementation using the Cohere API. Supports embed-english-v3.0 (1024 dims) and embed-multilingual-v3.0 (1024 dims). The embedding dimension is auto-detected from the model at initialization. iiii)zembed-english-v3.0zembed-multilingual-v3.0zembed-english-light-v3.0zembed-multilingual-light-v3.0zembed-english-v2.0zembed-multilingual-v2.0N`N@search_documentrrroutput_dimensionsrr input_typec||_||_||_||_||_||_||_d|_d|_dS)a Initialize Cohere embeddings client. Args: api_key: Cohere API key model: Cohere embedding model name (default: embed-english-v3.0) base_url: Custom base URL for Cohere-compatible API (e.g., Azure-hosted endpoint) output_dimensions: Optional output embedding dimensions (for Matryoshka-capable models) batch_size: Maximum batch size for embedding requests (default: 96, Cohere's limit) timeout: Request timeout in seconds (default: 60.0) input_type: Input type for embeddings (default: search_document). Options: search_document, search_query, classification, clustering N) rrrrrrrrrB)r%rrrrrrrs r&rCzCohereEmbeddings.__init__sH.    !2$ $ &*r)r!cdS)Ncoherer#r$s r&r'zCohereEmbeddings.provider_name#rr)c<|jtd|jSrGrIr$s r&r+zCohereEmbeddings.dimension'rKr)cK|jdS ddl}n#t$rtdwxYw|jr d|jnd}td|j||j|jd}|jr |j|d<|j di||_|j |j |_ n|j|j vr|j |j|_ nh|j d g|j|j }|jr9t!|jt"rt%|jd|_ td |jd |j d dS)z2Initialize the Cohere client and detect dimension.NrzLcohere is required for CohereEmbeddings. Install it with: pip install cohererrz4Embeddings: initializing Cohere provider with model )rrrrr.rrz0Embeddings: Cohere provider initialized (model: rrWr#)rrrYrrZr[rrrrrrBrembedrrz isinstancer9r)r%rrrrs r&r-zCohereEmbeddings.initialize-s < # F n MMMM n n nlmm m n26F-dm---B  e4:eWceefff%)LT\JJ = 6(, M* %$v}55}55   ! -"4DOO Z40 0 0"3DJ?DOO|))hj?*H " >z(2Et'L'L >"%h&9!&<"="= ltzllZ^Zilllmmmmms,r.c|jtd|sgSg}tdt||jD]}||||jz}|jT|jj||j|j |jdg}| |j j o|j||j|j }| |j |S)z Generate embeddings using the Cohere API. Args: texts: List of text strings to encode Returns: List of embedding vectors NrHrr:)r.rroutput_dimensionembedding_typesr) rrJrrrrv2rrrrrzfloat_)r%r.rrrrs r&r0zCohereEmbeddings.encodeQs < UVV V Iq#e**do66 ; ;A!a$/112E%1.s AgJr)rcg|] }|d Srr#rs r&ryz,LiteLLMEmbeddings.encode..s"L"L"La1[>"L"L"Lr)) rrJrrrrrrrrrr)r%r.rrrrr rs r&r0zLiteLLMEmbeddings.encodes  < UVV V Iq#e**do66 N NA!a$/112E|((=---#zE::)H  % % ' ' ']]__F &fVn:N:NOOO   ! !"L"L;K"L"L"L M M M Mr)r1)r2r3r4r5rr r7r8r:rCr6r'r+r-r9r0r#r)r&rr|s$1"5 +++t+ +  +  ++++6sX3X ^^^^6!DI!$tE{*;!!!!!!r)rceZdZdZedddddfdedededzd edzd ed ed edzfd Ze defdZ e defdZ ddZ de ede e efdZdS)LiteLLMSDKEmbeddingsa LiteLLM SDK embeddings for direct API integration. Supports embeddings via LiteLLM SDK without requiring a proxy server. Supported providers: Cohere, OpenAI, Azure OpenAI, HuggingFace, Voyage AI, Together AI, etc. Example model names: - cohere/embed-english-v3.0 - openai/text-embedding-3-small - together_ai/togethercomputer/m2-bert-80M-8k-retrieval - voyage/voyage-2 Nrrr:rrrrrrencoding_formatc||_||_||_||_||_||_|pd|_d|_d|_dS)a Initialize LiteLLM SDK embeddings client. Args: api_key: API key for the embedding provider model: Model name with provider prefix (e.g., "cohere/embed-english-v3.0") api_base: Custom base URL for API (optional) output_dimensions: Optional output embedding dimensions (provider-dependent) batch_size: Maximum batch size for embedding requests (default: 100) timeout: Request timeout in seconds (default: 60.0) encoding_format: Encoding format for embeddings (default: "float"). Set to None or empty string to omit (needed for Voyage AI, Gemini). N) rrrrrrr_litellmrB)r%rrrrrrrs r&rCzLiteLLMSDKEmbeddings.__init__sN.    !2$ .6$ &*r)r!cdS)N litellm-sdkr#r$s r&r'z"LiteLLMSDKEmbeddings.provider_name s}r)c<|jtd|jSrGrIr$s r&r+zLiteLLMSDKEmbeddings.dimension$rKr)c0K|jdS ddl}||_n#t$rtdwxYw|jr d|jnd}td|j| |jdg|jd}|jr |j|d <|jr |j|d <|j *|j |d <|j d rd g|d <|jj di|d{V}|j r>t|j dkr&t|j dd|_ntd|jn$#t $r}td|d}~wwxYwtd|jd|jddS)z7Initialize the LiteLLM SDK client and detect dimension.NrzRlitellm is required for LiteLLMSDKEmbeddings. Install it with: pip install litellmrrz9Embeddings: initializing LiteLLM SDK provider with model rrrrrr dimensionsopenai/allowed_openai_paramsrz/Unable to detect embedding dimension for model z-Failed to initialize LiteLLM SDK embeddings: z5Embeddings: LiteLLM SDK provider initialized (model: rrWr#)rrrYrrZr[rrrr startswith aembeddingrrrBrJra)r%r api_base_msg embed_kwargsrrns r&r-zLiteLLMSDKEmbeddings.initialize*s) = $ F t NNN#DMM t t trss s t26F-dm---B  jPTPZj\hjjkkk T <L # G262F ./} 9+/= Z(%1-1-C \*:((33K=INL!896T]5EE EEEEEEEEH} cX]!3!3a!7!7"%hmA&6{&C"D"D"#aUYU_#a#abbb  T T TRqRRSS S T  qDJqq_c_nqqqrrrrrs! 3.CE E(E##E(r.c |jtd|sgSg}tdt||jD] }||||jz} |j||jd}|jr |j|d<|jr |j|d<|j *|j |d<|j drdg|d <|jj di|}t|j d }|d |D#t$r?}ddl} t"d |d|d| d}~wwxYw|S)z Generate embeddings using the LiteLLM SDK. Args: texts: List of text strings to encode Returns: List of embedding vectors (one per input text) NrHrrrrrrrc.|ddS)Nrr)rrs r&rz-LiteLLMSDKEmbeddings.encode..squuWVWGXGXr)rcg|] }|d Srr#rs r&ryz/LiteLLMSDKEmbeddings.encode..s&P&P&P!q~&P&P&Pr)z7Error in LiteLLM embedding for batch starting at index rz Traceback: r#)rrJrrrrrrrrrrrrrra tracebackrZerror format_exc) r%r.rrrrrrrnr"s r&r0zLiteLLMSDKEmbeddings.encodeXs = UVV V Iq#e**do66" " A!a$/112E "Z"#|   'K6:6JL!23==/3}L,)5151GL.z,,Y77OAM %<=34=2BB\BB$*(-=X=X#Y#Y#Y %%&P&P?O&P&P&PQQQQ        ;a;;ST;;"+"6"6"8"8;; sB+D E  :EE r1)r2r3r4r5r r7r8r:rCr6r'r+r-r9r0r#r)r&rrs2   :#(,&-++++* + : +  ++t++++BsX3X ,s,s,s,s\7DI7$tE{*;777777r)rceZdZdZedddddddfdededzdedzdedzd edzd edzd ed efd Ze defdZ e defdZ ddZ ddZ ddZdeedeeefdZdS)GeminiEmbeddingsa2 Google embeddings via the google.genai SDK. Supports both: 1. Gemini API (api.generativeai.google.com) with API key authentication 2. Vertex AI with service account or Application Default Credentials (ADC) Uses the embed_content API: client.models.embed_content(model, contents) NrFrrvertexai_project_idvertexai_regionvertexai_service_account_keyoutput_dimensionalityr force_ipv4c ||_||_||_|pd|_||_||_||_||_d|_d|_ d|_ |du|_ d|_ dS)Nz us-central1) rrr'r(r)r*rr+r _httpx_clientrB _is_vertexai _embed_config) r%rrr'r(r)r*rr+s r&rCzGeminiEmbeddings.__init__sv  #6 .?-,H)%:"$$ !&*/t;!r)r!cdS)Ngoogler#r$s r&r'zGeminiEmbeddings.provider_namerr)c<|jtd|jSrGrIr$s r&r+zGeminiEmbeddings.dimensionrKr)c FK|jdSddlm}ddlm}|jr||n||||j | |j|_ |j dgd}|j |j |d<|jj j di|}|jr>DO#'#4CKK)  wi w wRVR\ w weiet w w w     r)cL|jstdd|ji}|jrIddl}|jd|jd|_|d |j |d <|jd i||_t d |j dS)z*Initialize Gemini API client with API key.z.Gemini embeddings provider requires an API keyrrN z0.0.0.0) local_address)r transporti')r httpxClient http_optionsz4Embeddings: initializing Gemini provider with model r#) r ValueErrorr+rr HTTPTransportr- HttpOptionsrrZr[r)r%r4rArrs r&r<zGeminiEmbeddings._init_geminis| OMNN N"DL1 ?  LLL!--%-IFFF"""D -8,C,C .-D--M. ) $u|44m44  W4:WWXXXXXr)c R|jstdd}d}|jrk ddlm}n#t $rt dwxYw|j|jdg}d }t d |j|j d r!|j td d|_ d |j|j d }|||d<|jdi||_t d|jd|j d|j d|d dS)zBInitialize Vertex AI client with project, region, and credentials.zHINDSIGHT_API_EMBEDDINGS_VERTEXAI_PROJECT_ID (or HINDSIGHT_API_LLM_VERTEXAI_PROJECT_ID) is required for Vertex AI embeddings provider.ADCNr)service_accountzdVertex AI service account auth requires 'google-auth' package. Install with: pip install google-authz.https://www.googleapis.com/auth/cloud-platform)scopesrNz1Embeddings: Vertex AI using service account key: zgoogle/T)vertexaiprojectlocation credentialsz5Embeddings: initializing Vertex AI provider (project=z , region=z, model=z, auth=rWr#)r'rIr) google.oauth2rNrY Credentialsfrom_service_account_filerZr[rrrr(rr)r%r4 auth_methodrSrNrs r&r;zGeminiEmbeddings._init_vertexais' A     , q 9999999   !<  *5OO1HIPK,K KKoDLmoo p p p :  + + 6C NN$4$45DJ/,   "+6M- (#u|44m44   70 7 7;?;O 7 7Z 7 7(3 7 7 7     s *Ar.c|jtd|sgSg}tdt||jD]j}||||jz}|j|d}|j |j|d<|jjjd i|}| d|j Dk|j Vddl }| |}|j|dd } d| | dk<|| z }|S) z Generate embeddings using the Google genai SDK. Args: texts: List of text strings to encode Returns: List of embedding vectors NrHrr6r8cg|] }|j Sr#)r@rvs r&ryz+GeminiEmbeddings.encode..?s"K"K"K#3:"K"K"Kr)rT)axiskeepdimsr#)rrJrrrrr/r>r?rrzr*numpyarraylinalgnormru) r%r.rrrrr nparrnormss r&r0zGeminiEmbeddings.encode#s? < UVV V Iq#e**do66 M MA!a$/112E%)ZUCCL!-)-); X&6T\(6FFFFF  ! !"K"K9J"K"K"K L L L L  % 1    ((>**CINN3QN>>E !E%1* !Ek1133Nr)r1)r2r3r4r5r r7r8r{rCr6r'r+r-r<r;r9r:r0r#r)r&r&r&sr5"*.&*37,0 """t"!4Z " t " '*Dj " #Tz""""""2sX3X     BYYYY*+ + + + Z)DI)$tE{*;))))))r)r&r!c ddlm}|}|j}|dkr8|j}|st t dtdt|S|dkr!t|j |j |j S|d krtjt p#tjt"}|s't t d t"dtd tjt$t&}tjt(pd }t+||||j S|dkrF|j}|s t dt"dtdt+||jd|j S|dkrJ|j}|st t4dtdt7||j|j|jS|dkr!t?|j |j!|j"S|dkrP|j#}|st tHdtdtK||j&|j'|j(|j)S|dkrk|j*}|rd }n0|j+}|s't tXd t"dtdt[|j.|||j/|j0|j1|j2St d|d)z Create an Embeddings instance based on configuration. Reads configuration via get_config() to ensure consistency across the codebase. Returns: Configured Embeddings instance r) get_configrz is required when z is 'tei')rrE)r=r>r?rz or z is 'openai'N)rrrr openrouterzRHINDSIGHT_API_EMBEDDINGS_OPENROUTER_API_KEY, HINDSIGHT_API_OPENROUTER_API_KEY, or z is 'openrouter'zhttps://openrouter.ai/api/v1rz is 'cohere')rrrrr)rrrrz is 'litellm-sdk')rrrrrr1zA is 'google' (set VERTEXAI_PROJECT_ID for Vertex AI auth instead))rrr'r(r)r*r+zUnknown embeddings provider: zS. Supported: 'local', 'tei', 'openai', 'cohere', 'google', 'litellm', 'litellm-sdk')3r8rdembeddings_providerlowerembeddings_tei_urlrIrrr}r<embeddings_local_modelembeddings_local_force_cpu"embeddings_local_trust_remote_codeosenvironrrrrrrrembeddings_openai_batch_sizeembeddings_openrouter_api_keyembeddings_openrouter_modelembeddings_cohere_api_keyrrembeddings_cohere_modelembeddings_cohere_base_url#embeddings_cohere_output_dimensionsrembeddings_litellm_api_baseembeddings_litellm_api_keyembeddings_litellm_modelembeddings_litellm_sdk_api_keyrrembeddings_litellm_sdk_modelembeddings_litellm_sdk_api_base(embeddings_litellm_sdk_output_dimensions&embeddings_litellm_sdk_encoding_formatembeddings_vertexai_project_idembeddings_gemini_api_keyrr&embeddings_gemini_modelembeddings_vertexai_region'embeddings_vertexai_service_account_key'embeddings_gemini_output_dimensionalityembeddings_gemini_force_ipv4)rdr8providerrrrrr's r&create_embeddings_from_envrOs $##### Z\\F)//11H5' n 6llJalllmm m"C0000 W   47$G    X  *..!>??b2:>>RaCbCb 0>>o>>/>>>  :<[\\:>>"@AAIT:     \ ! !6 c%cc9Pccc  43:     X  2 x =vvQhvvvww w06$H     Y   751    ] " "7 5ssI`sss $5;$M"I      X  $C  GG6G  4ww/ww3www 0 3"=)/)W"("P:     aH a a a   r))/r5rgrlrcabcrr urllib.parserrrrr8r r r r r rrrrrrrrrrrrrrrrrrhr2rZr r<r}rrrrr&rr#r)r&rs+   ########7777777777 2  8 $ $) ) ) ) ) ) ) ) Xt4t4t4t4t4 t4t4t4nWWWWW*WWWtzDJJJJJzJJJZrrrrr rrrj^^^^^:^^^Bzzzzzzzzzzk Jk k k k k k r)