feat(bigframes): update ai.if_() params to match the SQL version (#16857)

Reference:
https://docs.cloud.google.com/bigquery/docs/reference/standard-sql/bigqueryml-syntax-ai-if

I left out the `embeddings` param to keep things simple. It will be
introduced later.

---------

Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>

Author: sycai

packages/bigframes/bigframes/bigquery/_operations/ai.py

@@ -817,6 +817,9 @@ def if_(
     prompt: PROMPT_TYPE,
     *,
     connection_id: str | None = None,
+    endpoint: str | None = None,
+    optimization_mode: Literal["minimize_cost", "maximize_quality"] = "minimize_cost",
+    max_error_ratio: float | None = None,
 ) -> series.Series:
     """
     Evaluates the prompt to True or False. Compared to `ai.generate_bool()`, this function
@@ -838,20 +841,26 @@ def if_(
         1         Illinois
         dtype: string
 
-    .. note::
-
-        This product or feature is subject to the "Pre-GA Offerings Terms" in the General Service Terms section of the
-        Service Specific Terms(https://cloud.google.com/terms/service-terms#1). Pre-GA products and features are available "as is"
-        and might have limited support. For more information, see the launch stage descriptions
-        (https://cloud.google.com/products#product-launch-stages).
-
     Args:
         prompt (str | Series | List[str|Series] | Tuple[str|Series, ...]):
             A mixture of Series and string literals that specifies the prompt to send to the model. The Series can be BigFrames Series
             or pandas Series.
         connection_id (str, optional):
             Specifies the connection to use to communicate with the model. For example, `myproject.us.myconnection`.
             If not provided, the query uses your end-user credential.
+        endpoint (str, optional):
+            Specifies the Vertex AI endpoint to use for the model. For example `"gemini-2.5-flash"`. You can specify any
+            generally available or preview Gemini model. If you specify the model name, BigQuery ML automatically identifies and
+            uses the full endpoint of the model. If you don't specify an ENDPOINT value, BigQuery ML dynamically chooses a model based on your query to have the
+            best cost to quality tradeoff for the task.
+        optimization_mode (Literal["minimize_cost", "maximize_quality"]):
+            Specifies the optimization strategy to use. Supported values are:
+            * "minimize_cost" (default): uses a local, distilled model to process the majority of rows, reducing latency and cost.
+            * "maximize_quality": always uses the remote LLM for inference.
+        max_error_ratio (float):
+            A float value between 0.0 and 1.0 that contains the maximum acceptable ratio of row-level inference failures to
+            rows processed on this function. If this value is exceeded, then the query fails. The default value is 1.0.
+            This argument isn't supported when `optimization_mode` is set to "minimize_cost".
 
     Returns:
         bigframes.series.Series: A new series of bools.
@@ -863,6 +872,9 @@ def if_(
     operator = ai_ops.AIIf(
         prompt_context=tuple(prompt_context),
         connection_id=connection_id,
+        endpoint=endpoint,
+        optimization_mode=optimization_mode,
+        max_error_ratio=max_error_ratio,
     )
 
     return series_list[0]._apply_nary_op(operator, series_list[1:])

packages/bigframes/bigframes/core/compile/ibis_compiler/scalar_op_registry.py

@@ -1983,6 +1983,9 @@ def ai_if(*values: ibis_types.Value, op: ops.AIIf) -> ibis_types.StructValue:
     return ai_ops.AIIf(
         _construct_prompt(values, op.prompt_context),  # type: ignore
         op.connection_id,  # type: ignore
+        op.endpoint,  # type: ignore
+        op.optimization_mode.upper() if op.optimization_mode is not None else None,  # type: ignore
+        op.max_error_ratio,  # type: ignore
     ).to_expr()
 
 

packages/bigframes/bigframes/core/compile/sqlglot/expressions/ai_ops.py

@@ -139,6 +139,12 @@ def _construct_named_args(op: ops.ScalarOp) -> list[sge.Kwarg]:
                     expression=sge.JSON(this=sge.Literal.string(value)),
                 )
             )
+        elif field == "optimization_mode":
+            args.append(
+                sge.Kwarg(this=field, expression=sge.Literal.string(value.upper()))
+            )
+        elif field == "max_error_ratio":
+            args.append(sge.Kwarg(this=field, expression=sge.Literal.number(value)))
         elif field == "request_type":
             args.append(
                 sge.Kwarg(this=field, expression=sge.Literal.string(value.upper()))

packages/bigframes/bigframes/operations/ai_ops.py

@@ -146,6 +146,9 @@ class AIIf(base_ops.NaryOp):
 
     prompt_context: Tuple[str | None, ...]
     connection_id: str | None
+    endpoint: str | None = None
+    optimization_mode: str | None = None
+    max_error_ratio: float | None = None
 
     def output_type(self, *input_types: dtypes.ExpressionType) -> dtypes.ExpressionType:
         return dtypes.BOOL_DTYPE

packages/bigframes/tests/system/small/bigquery/test_ai.py

@@ -323,7 +323,11 @@ def test_ai_if(session):
     s2 = bpd.Series(["fruit", "tree"], session=session)
     prompt = (s1, " is a ", s2)
 
-    result = bbq.ai.if_(prompt)
+    result = bbq.ai.if_(
+        prompt,
+        optimization_mode="maximize_quality",
+        max_error_ratio=0.5,
+    )
 
     assert _contains_no_nulls(result)
     assert result.dtype == dtypes.BOOL_DTYPE

packages/bigframes/tests/unit/core/compile/sqlglot/expressions/snapshots/test_ai_ops/test_ai_if/None/out.sql

@@ -1,3 +1,7 @@
 SELECT
-  AI.IF(prompt => (`string_col`, ' is the same as ', `string_col`)) AS `result`
+  AI.IF(
+    prompt => (`string_col`, ' is the same as ', `string_col`),
+    optimization_mode => 'MINIMIZE_COST',
+    max_error_ratio => 0.5
+  ) AS `result`
 FROM `bigframes-dev`.`sqlglot_test`.`scalar_types` AS `bft_0`

packages/bigframes/tests/unit/core/compile/sqlglot/expressions/snapshots/test_ai_ops/test_ai_if/bigframes-dev.us.bigframes-default-connection/out.sql

@@ -1,6 +1,8 @@
 SELECT
   AI.IF(
     prompt => (`string_col`, ' is the same as ', `string_col`),
-    connection_id => 'bigframes-dev.us.bigframes-default-connection'
+    connection_id => 'bigframes-dev.us.bigframes-default-connection',
+    optimization_mode => 'MINIMIZE_COST',
+    max_error_ratio => 0.5
   ) AS `result`
 FROM `bigframes-dev`.`sqlglot_test`.`scalar_types` AS `bft_0`

packages/bigframes/tests/unit/core/compile/sqlglot/expressions/snapshots/test_ai_ops/test_ai_if_with_endpoint/out.sql

@@ -0,0 +1,6 @@
+SELECT
+  AI.IF(
+    prompt => (`string_col`, ' is the same as ', `string_col`),
+    endpoint => 'gemini-2.5-flash'
+  ) AS `result`
+FROM `bigframes-dev`.`sqlglot_test`.`scalar_types` AS `bft_0`

packages/bigframes/tests/unit/core/compile/sqlglot/expressions/test_ai_ops.py

@@ -358,6 +358,24 @@ def test_ai_if(scalar_types_df: dataframe.DataFrame, snapshot, connection_id):
     op = ops.AIIf(
         prompt_context=(None, " is the same as ", None),
         connection_id=connection_id,
+        optimization_mode="minimize_cost",
+        max_error_ratio=0.5,
+    )
+
+    sql = utils._apply_ops_to_sql(
+        scalar_types_df, [op.as_expr(col_name, col_name)], ["result"]
+    )
+
+    snapshot.assert_match(sql, "out.sql")
+
+
+def test_ai_if_with_endpoint(scalar_types_df: dataframe.DataFrame, snapshot):
+    col_name = "string_col"
+
+    op = ops.AIIf(
+        prompt_context=(None, " is the same as ", None),
+        connection_id=None,
+        endpoint="gemini-2.5-flash",
     )
 
     sql = utils._apply_ops_to_sql(

packages/bigframes/third_party/bigframes_vendored/ibis/expr/operations/ai_ops.py

@@ -138,6 +138,9 @@ class AIIf(Value):
 
     prompt: Value
     connection_id: Optional[Value[dt.String]]
+    endpoint: Optional[Value[dt.String]] = None
+    optimization_mode: Optional[Value[dt.String]] = None
+    max_error_ratio: Optional[Value[dt.Float64]] = None
 
     shape = rlz.shape_like("prompt")