o U|%i@s|dZddlmZmZmZddlZddlmZddl m Z ddl m Z m Z mZmZmZmZmZddlmZGdd d eZdS) zEmbeddings integration utilities for MySQL Connector/Python. Provides MyEmbeddings class to generate embeddings via MySQL HeatWave using ML_EMBED_TABLE and ML_EMBED_ROW. )DictListOptionalN) Embeddings) PrivateAttr)atomic_transaction execute_sqlformat_value_sql source_schemasql_table_from_dfsql_table_to_dftemporary_sql_tables)MySQLConnectionAbstractcsveZdZUdZeZeed< ddedee ffdd Z de e d e e e fd d Zd e d e e fd dZZS) MyEmbeddingsa Embedding generator class that uses a MySQL database to compute embeddings for input text. This class batches input text into temporary SQL tables, invokes MySQL's ML_EMBED_TABLE to generate embeddings, and retrieves the results as lists of floats. Attributes: _db_connection (MySQLConnectionAbstract): MySQL connection used for all database operations. schema_name (str): Name of the database schema to use. options_placeholder (str): SQL-ready placeholder string for ML_EMBED_TABLE options. options_params (dict): Dictionary of concrete option values to be passed as SQL parameters. _db_connectionN db_connectionoptionscs6t||_t||_|pi}t|\|_|_dS)a Initialize MyEmbeddings with a database connection and optional embedding parameters. References: https://dev.mysql.com/doc/heatwave/en/mys-hwgenai-ml-embed-row.html A full list of supported options can be found under "options" NOTE: The supported "options" are the intersection of the options provided in https://dev.mysql.com/doc/heatwave/en/mys-hwgenai-ml-embed-row.html https://dev.mysql.com/doc/heatwave/en/mys-hwgenai-ml-embed-table.html Args: db_connection: Active MySQL connector database connection. options: Optional dictionary of options for embedding operations. Raises: ValueError: If the schema name is not valid DatabaseError: If a database connection issue occurs. If an operational error occurs during execution. N)super__init__rr schema_namer options_placeholderoptions_params)selfrr __class__U/home/air/sos_test/back/venv/lib/python3.10/site-packages/mysql/ai/genai/embedding.pyrFs  zMyEmbeddings.__init__textsreturnc Cs.|sgSttt||d}t|jx}t|jb}t||j|\}}| |j|fd|d|d|j d}t |||j dt ||j|}|ds\tdd |dDr`td |d} d d | D} | Wd Wd S1swYWd d S1swYd S)a Generate embeddings for a list of input texts using the MySQL ML embedding procedure. References: https://dev.mysql.com/doc/heatwave/en/mys-hwgenai-ml-embed-table.html Args: texts: List of input strings to embed. Returns: List of lists of floats, with each inner list containing the embedding for a text. Raises: DatabaseError: If provided options are invalid or unsupported. If a database connection issue occurs. If an operational error occurs during execution. ValueError: If one or more text entries were unable to be embedded. Implementation notes: - Creates a temporary table to pass input text to the MySQL embedding service. - Adds a primary key to ensure results preserve input order. - Calls ML_EMBED_TABLE and fetches the resulting embeddings. - Deletes the temporary table after use to avoid polluting the database. - Embedding vectors are extracted from the "embeddings" column of the result table. )idtextzCALL sys.ML_EMBED_TABLE('z .text', 'z.embeddings', )params embeddingscss|]}|duVqdSNr.0errr s z/MyEmbeddings.embed_documents..z:Failure to generate embeddings for one or more text entry.cSsg|]}t|qSr)listr&rrr sz0MyEmbeddings.embed_documents..N)pd DataFramerangelenrrr r rappendrrrr isnullany ValueErrortolist) rrdfcursortemporary_tablesqualified_table_name table_name embed_query df_embeddingsr$rrrembed_documentsdsB RzMyEmbeddings.embed_documentsr cCs`t|j!}t|d|jd|g|jRdt|dWdS1s)wYdS)a Generate an embedding for a single text string. References: https://dev.mysql.com/doc/heatwave/en/mys-hwgenai-ml-embed-row.html Args: text: The input string to embed. Returns: List of floats representing the embedding vector. Raises: DatabaseError: If provided options are invalid or unsupported. If a database connection issue occurs. If an operational error occurs during execution. Example: >>> MyEmbeddings(db_conn).embed_query("Hello world") [0.1, 0.2, ...] zSELECT sys.ML_EMBED_ROW("%s", r!r"rN)rrrrrr*fetchone)rr r6rrrr:s   $zMyEmbeddings.embed_queryr%)__name__ __module__ __qualname____doc__rrr__annotations__rrrrstrfloatr<r: __classcell__rrrrr6s  Dr)rAtypingrrrpandasr,langchain_core.embeddingsrpydanticrmysql.ai.utilsrrr r r r r mysql.connector.abstractsrrrrrrs  $