now.executor.indexer.elastic.es_query_building module#
- now.executor.indexer.elastic.es_query_building.generate_score_calculation(docs_map, encoder_to_fields)[source]#
Generate score calculation from document mappings.
- Parameters
docs_map (
Dict[str,DocumentArray]) – dictionary mapping encoder to DocumentArray.encoder_to_fields (
Dict[str,Union[str,List[str]]]) – dictionary mapping encoder to fields.
- Return type
List[List]- Returns
a list of score calculation, each of which is a tuple of (query_field, document_field, encoder, linear_weight). score calculation would then be for example: [(‘query_text’, ‘title’, ‘clip’, 1.0)]
- now.executor.indexer.elastic.es_query_building.build_es_queries(docs_map, get_score_breakdown, score_calculation, metric='cosine', filter={}, query_to_curated_ids={})[source]#
Build script-score query used in Elasticsearch. To do this, we extract embeddings from the query document and pass them in the script-score query together with the fields to search on in the Elasticsearch index. The query document will be returned with all of its embeddings as tags with their corresponding field+encoder as key.
- Parameters
docs_map – dictionary mapping encoder to DocumentArray.
get_score_breakdown (
bool) – whether to return the score breakdown for matches. For this function, this parameter determines whether to return the embeddings of a query document.score_calculation (
List[Tuple]) – list of nested lists containing (query_field, document_field, matching_method, linear_weight) which define how to calculate the score. Note, that the matching_method is the name of the encoder or bm25.metric (
Optional[str]) – metric to use for vector search.filter (
dict) – dictionary of filters to apply to the search.query_to_curated_ids (
Dict[str,list]) – dictionary mapping query text to list of curated ids.
- Return type
Dict- Returns
a dictionary containing query and filter.
- now.executor.indexer.elastic.es_query_building.get_default_query(doc, score_calculation, filter={})[source]#
- now.executor.indexer.elastic.es_query_building.get_pinned_query(doc, query_to_curated_ids={})[source]#
- Return type
Dict
