U h0@sfddlZddlmZmZmZmZddlZddlmZddl m Z ddl m Z m Z Gddde eZdS)N)Dict GeneratorListOptional) Embeddings) BaseModel)get_from_dict_or_envpre_initc@seZdZUdZdZeed<dZeed<dZeed<dZ eed<dZ eed<iZ e ed<d Z eed <eeed d d ZedddZeedddZeeeedddZdeeeeeeedddZeeedddZdS)SambaStudioEmbeddingsa3SambaNova embedding models. To use, you should have the environment variables ``SAMBASTUDIO_EMBEDDINGS_BASE_URL``, ``SAMBASTUDIO_EMBEDDINGS_BASE_URI`` ``SAMBASTUDIO_EMBEDDINGS_PROJECT_ID``, ``SAMBASTUDIO_EMBEDDINGS_ENDPOINT_ID``, ``SAMBASTUDIO_EMBEDDINGS_API_KEY`` set with your personal sambastudio variable or pass it as a named parameter to the constructor. Example: .. code-block:: python from langchain_community.embeddings import SambaStudioEmbeddings embeddings = SambaStudioEmbeddings(sambastudio_embeddings_base_url=base_url, sambastudio_embeddings_base_uri=base_uri, sambastudio_embeddings_project_id=project_id, sambastudio_embeddings_endpoint_id=endpoint_id, sambastudio_embeddings_api_key=api_key, batch_size=32) (or) embeddings = SambaStudioEmbeddings(batch_size=32) (or) # CoE example embeddings = SambaStudioEmbeddings( batch_size=1, model_kwargs={ 'select_expert':'e5-mistral-7b-instruct' } ) sambastudio_embeddings_base_urlsambastudio_embeddings_base_uri!sambastudio_embeddings_project_id"sambastudio_embeddings_endpoint_idsambastudio_embeddings_api_key model_kwargs batch_size)valuesreturncCsXt|dd|d<t|dddd|d<t|dd|d<t|d d |d <t|d d |d <|S) z?Validate that api key and python package exists in environment.r ZSAMBASTUDIO_EMBEDDINGS_BASE_URLr ZSAMBASTUDIO_EMBEDDINGS_BASE_URIapi/predict/generic)defaultrZ!SAMBASTUDIO_EMBEDDINGS_PROJECT_IDrZ"SAMBASTUDIO_EMBEDDINGS_ENDPOINT_IDrZSAMBASTUDIO_EMBEDDINGS_API_KEY)r)clsrrL/tmp/pip-unpacked-wheel-9gdii04g/langchain_community/embeddings/sambanova.pyvalidate_environmentCs6 z*SambaStudioEmbeddings.validate_environment)rcCs4d|jkr|j}ndd|jD}t|}|S)z Get the tuning parameters to use when calling the model Returns: The tuning parameters as a JSON string. api/v2/predict/genericcSs&i|]\}}|t|jt|dqS))typevalue)r__name__str).0kvrrr hsz.)r ritemsjsondumps)selfZtuning_params_dictZ tuning_paramsrrr_get_tuning_params^s  z(SambaStudioEmbeddings._get_tuning_params)pathrcCs|jd|jd|S)z Return the full API URL for a given path. :param str path: the sub-path :returns: the full API URL for the sub-path :rtype: str /)r r )r(r*rrr _get_full_urlosz#SambaStudioEmbeddings._get_full_url)textsrrccs,tdt||D]}||||VqdS)afGenerator for creating batches in the embed documents method Args: texts (List[str]): list of strings to embed batch_size (int, optional): batch size to be used for the embedding model. Will depend on the RDU endpoint used. Yields: List[str]: list (batch) of strings of size batch size rN)rangelen)r(r-rirrr_iterate_over_batchesys z+SambaStudioEmbeddings._iterate_over_batchesNc Csz|dkr|j}t}||jd|j}t|}g}d|j kr| ||D]}||d}|j |d|j i|d} | j dkrtd| j d | jz| d } || WqVtk rtd | YqVXqVnd |j kr| ||D]}d dt|D} | |d}|j |d|j i|d} | j dkrRtd| j d | jz$dd| dD} || Wqtk rtd| YqXqnd|j krd| ||D]}||d}|j |d|j i|d} | j dkrtd| j d | jz8|dr | dd} n | d} || Wn$tk r\td| YnXqntd|j d|S)a<Returns a list of embeddings for the given sentences. Args: texts (`List[str]`): List of texts to encode batch_size (`int`): Batch size for the encoding Returns: `List[np.ndarray]` or `List[tensor]`: List of embeddings for the given sentences Nr+api/predict/nlpinputsparamskeyheadersr&1Sambanova /complete call failed with status code . Details: data%'data' not found in endpoint responsercSs g|]\}}d||dqS)itemidrr)r!r0r>rrr sz9SambaStudioEmbeddings.embed_documents..r%r5cSsg|] }|dqS)rr)r!r>rrrrAsr%&'items' not found in endpoint responserZ instancesr5 select_expert predictionsr,'predictions' not found in endpoint responsehandling of endpoint uri:  not implemented)rrequestsSessionr,rrr&loadsr)r r1postr status_code RuntimeErrortextextendKeyError enumerateget ValueError) r(r-r http_sessionurlr5Z embeddingsbatchr<response embeddingr%rrrembed_documentss              z%SambaStudioEmbeddings.embed_documents)rPrcCst}||jd|j}t|}d|jkr|g|d}|j |d|j i|d}|j dkr|t d|j d|j z|d d }Wn"tk rtd |YnXnZd |jkrPd |dg|d}|j |d|j i|d}|j dkrt d|j d|j z|dd d}Wn$tk rLtd|YnXnd|jkr|g|d}|j |d|j i|d}|j dkrt d|j d|j z6|dr|dd d }n|dd }Wn$tk rtd|YnXntd|jd|S)aReturns a list of embeddings for the given sentences. Args: sentences (`List[str]`): List of sentences to encode Returns: `List[np.ndarray]` or `List[tensor]`: List of embeddings for the given sentences r+r2r3r6r7r9r:r;r<rr=rZitem0r?rBr%rrCrrDrErFrGrHrI)rJrKr,rrr&rLr)r rMrrNrOrPrRrTrU)r(rPrVrWr5r<rYrZrrr embed_querys             z!SambaStudioEmbeddings.embed_query)N)r __module__ __qualname____doc__r r __annotations__r rrrrdictrintr rrr)r,rrr1rfloatr[r\rrrrr s( #          dr )r&typingrrrrrJZlangchain_core.embeddingsrZlangchain_core.pydantic_v1rZlangchain_core.utilsrr r rrrrs