HANAMLinVectorEngine

class hana_ai.vectorstore.hana_vector_engine.HANAMLinVectorEngine(connection_context, table_name, schema=None, model_version='SAP_NEB.20240715')

HANA vector engine.

Parameters:
connection_context: ConnectionContext

Connection context.

table_name: str

Table name. Table/schema names cannot contain single or double quotes, NUL bytes or semicolons; ordinary case-sensitive names are supported.

schema: str, optional

Schema name. Default to None.

model_version: str, optional

Model version. Default to 'SAP_NEB.20240715'. NUL bytes and semicolons are rejected; apostrophes in values are safely encoded.

get_knowledge()

Get knowledge dataframe.

create_knowledge(option='python')

Create knowledge base.

Parameters:
option: {'python', 'sql'}, optional

The option of language. Default to 'python'.

upsert_knowledge(knowledge)

Upsert knowledge.

Parameters:
knowledge: dict

Knowledge data. {'id': '1', 'description': 'description', 'example': 'example'}

query(input, top_n=1, distance='cosine_similarity')

Query.

Parameters:
input: str

Query text. Apostrophes are supported; NUL bytes and semicolons are rejected because hana-ml splits SQL text before parsing quotes.

top_n: int, optional

Query limit and 1-based result rank, from 1 to 1000. Default to 1.

distance: str, optional

Case-insensitive function name: COSINE_SIMILARITY, L2DISTANCE, EUCLIDEAN_DISTANCE, DOT_PRODUCT, MANHATTAN_DISTANCE or INNER_PRODUCT. Default to 'cosine_similarity'. The selected function must also be supported by the target HANA database.

Returns:
str

Template text at the requested rank, or the last available rank when fewer rows match. This method does not return a result list.