Ng|ddlmZddlZddlmZmZmZmZmZm Z m Z m Z ddl m Z erddlZejeZe eeeeeefZdd Zdd ZGd de dZGddZdS)) annotationsN) TYPE_CHECKINGAnyCallableListLiteralOptionalSequenceUnion) TypedDictXMatrixYreturn np.ndarraycddl}t|dkst|dkr|jgS|j|}|j|}|jd|jdkr t d|jd|jd ddl}|j||j}|j||j}d|||d z }t|tr|j|gS|j|S#t$rt d |j |d }|j |d }|jd d 5|j||j|j||z }dddn #1swxYwYd||j||j|z<|cYSwxYw)z*>>? s>A2D DA.G2 )F>2 G2>G G2G )G21G24Callable[[Sequence[str]], Sequence[Sequence[float]]]c^ ddlmn#t$rtdwxYwd fd }|S) zGet the OpenAI GPT-3 encoder.r)ClientzTHe default encoder for the EmbeddingDistance class uses the OpenAI API. Please either install the openai library with `pip install openai` or provide a custom encoder function (Callable[[str], Sequence[float]]).texts Sequence[str]rSequence[Sequence[float]]c}|jt|d}d|jDS)Nztext-embedding-3-small)inputmodelcg|] }|j S) embedding).0ds r7 z<_get_openai_encoder..encode_text..Os333 333) embeddingscreatelistdata)r<clientresponse OpenAIClients r7 encode_textz(_get_openai_encoder..encode_textJsO$++u++%=,  43X]3333rH)r<r=rr>)openair;r&)rPrOs @r7_get_openai_encoderrR?sv 1111111     T    444444 s $c$eZdZUded<ded<dS)EmbeddingConfigz0Callable[[List[str]], Sequence[Sequence[float]]]encoderzCLiteral['cosine', 'euclidean', 'manhattan', 'chebyshev', 'hamming']rN)__name__ __module__ __qualname____annotations__rCrHr7rTrTTs*====OOOOOOrHrTF)totalceZdZ dddZdd ZddZeddZeddZeddZ eddZ eddZ dS)EmbeddingDistanceNconfigOptional[EmbeddingConfig]c|pi}|dpd|_|dp t|_dS)NrrrU)getdistancerRrU)selfr]s r7__init__zEmbeddingDistance.__init__ZsH2 8,,8 zz),,E0C0E0E rH predictionstr referencerr%c ddl}n#t$rtdwxYw|||g}|j|}||d|dS)NrzWThe EmbeddingDistance class requires NumPy. Please install it with `pip install numpy`.r)rr&rUr_compute_distanceitem)rbrdrfr1rIvectors r7evaluatezEmbeddingDistance.evaluatebs          '   \\:y"9:: *%%%%fQi;;@@BBBs!arb np.floatingcz|jdkr|||S|jdkr|||S|jdkr|||S|jdkr|||S|jdkr|||St d|j)Nr euclidean manhattan chebyshevhammingzInvalid distance metric: )ra_cosine_distance_euclidean_distance_manhattan_distance_chebyshev_distance_hamming_distancer )rbrlrms r7rhz#EmbeddingDistance._compute_distancers =H $ $((A.. . ]k ) )++Aq11 1 ]k ) )++Aq11 1 ]k ) )++Aq11 1 ]i ' '))!Q// /HHHII IrHc,dt|g|gz S)zCompute the cosine distance between two vectors. Args: a (np.ndarray): The first vector. b (np.ndarray): The second vector. Returns: np.ndarray: The cosine distance. g?)r8rlrms r7rtz"EmbeddingDistance._cosine_distances&sQC0000rHcFtj||z S)zCompute the Euclidean distance between two vectors. Args: a (np.ndarray): The first vector. b (np.ndarray): The second vector. Returns: np.floating: The Euclidean distance. )r1r)r*rzs r7ruz%EmbeddingDistance._euclidean_distancesy~~a!e$$$rHcTtjtj||z S)zCompute the Manhattan distance between two vectors. Args: a (np.ndarray): The first vector. b (np.ndarray): The second vector. Returns: np.floating: The Manhattan distance. )r1sumabsrzs r7rvz%EmbeddingDistance._manhattan_distance vbfQUmm$$$rHcTtjtj||z S)zCompute the Chebyshev distance between two vectors. Args: a (np.ndarray): The first vector. b (np.ndarray): The second vector. Returns: np.floating: The Chebyshev distance. )r1maxr~rzs r7rwz%EmbeddingDistance._chebyshev_distancerrHc2tj||kS)zCompute the Hamming distance between two vectors. Args: a (np.ndarray): The first vector. b (np.ndarray): The second vector. Returns: np.floating: The Hamming distance. )r1meanrzs r7rxz#EmbeddingDistance._hamming_distanceswqAvrH)N)r]r^)rdrerfrerr%)rlrrmrrrn)rlrrmrrr) rVrWrXrcrkrh staticmethodrtrurvrwrxrCrHr7r\r\Ys-1FFFFFCCCC J J J J 1 1 1\ 1 % % %\ % % % %\ % % % %\ %   \   rHr\)r rrrrr)rr9) __future__rloggingtypingrrrrrr r r typing_extensionsr rr1 getLoggerrVr'r%rr8rRrTr\rCrHr7rs}""""""                    ('''''  8 $ $ tDK $s)S0 1""""J*PPPPPiuPPPP ffffffffffrH