Search the top K semantically similar records
Search the top K semantically similar records
If you have defined indices, you can query Epsilla with natural language questions. The question will be embedded using the same model as the index definition:
Response example:
If there is only one embedding field, the query field parameter can be omitted. Below is an example with minimum required parameters:
If you brought in your own embedding vectors via vector fields, you can query by providing the target vector:
Response example:
Filter Expression
Epsilla supports filter expression in search that is compatible with SQL. The following operators are supported:
+
Add
-
Subtract
*
Multiply
/
Divide
%
Modulo
=
Equal to
>
Greater than
<
Less than
>=
Greater than or equal to
<=
Less than or equal to
<>
Not equal to
AND
TRUE if all the conditions separated by AND is TRUE
NOT
TRUE if the condition afterward is NOT TRUE
OR
TRUE if any of the conditions separated by OR is TRUE
A string literal needs to be surrounded by single quotes.
Refer to a field of the record directly by the field name.
Providing an empty filter is same as not providing the filter at all (all records will pass the evaluation).
Filter on @distance
Epsilla supports filter the result by the distance. For example, use the following filter condition to only return the matching results that have a semantic similarity distance less than 0.5:
@distance can be combined and nested with all the other filter condition operators.
Last updated