feat(vector-stores): forward per-request params to Vertex AI Search

The vertex_ai/search_api search transform hardcoded the request body to
query plus pageSize 10, dropping max_num_results and extra_body. Map
max_num_results to pageSize and merge extra_body through with precedence,
so callers can send native Discovery Engine fields such as dataStoreSpecs.

Resolves LIT-3506
This commit is contained in:
ryan-crabbe-berri 2026-06-01 17:08:56 -07:00
parent 1cce49b9d0
commit 26a79f4524
2 changed files with 80 additions and 9 deletions

View file

@ -133,22 +133,25 @@ class VertexSearchAPIVectorStoreConfig(BaseVectorStoreConfig, VertexBase):
extra_body: Optional[Dict[str, Any]] = None,
) -> Tuple[str, Dict[str, Any]]:
"""
Transform search request for Vertex AI RAG API
Transform a search request for the Vertex AI Search (Discovery Engine) API.
Per-request params pass through to the engine: max_num_results maps to
pageSize, and extra_body is merged in raw and takes precedence, so callers
can send native Discovery Engine fields such as dataStoreSpecs, filter,
boostSpec, or contentSearchSpec.
"""
# Convert query to string if it's a list
if isinstance(query, list):
query = " ".join(query)
# Vertex AI RAG API endpoint for retrieving contexts
url = f"{api_base}:search"
# Construct full rag corpus path
# Build the request body for Vertex AI Search API
request_body = {"query": query, "pageSize": 10}
request_body: Dict[str, Any] = {"query": query, "pageSize": 10}
max_num_results = vector_store_search_optional_params.get("max_num_results")
if max_num_results is not None:
request_body["pageSize"] = max_num_results
if isinstance(extra_body, dict):
request_body.update(extra_body)
#########################################################
# Update logging object with details of the request
#########################################################
litellm_logging_obj.model_call_details["query"] = query
return url, request_body

View file

@ -1,3 +1,5 @@
from types import SimpleNamespace
import pytest
from litellm.llms.vertex_ai.vector_stores.search_api.transformation import (
@ -126,3 +128,69 @@ def test_should_raise_when_neither_engine_id_nor_vector_store_id_provided():
"vertex_location": "global",
},
)
_ENGINE_BASE = (
"https://discoveryengine.googleapis.com/v1/projects/p/locations/global/"
"collections/default_collection/engines/app-2/servingConfigs/default_serving_config"
)
def _search_request(**overrides):
kwargs = dict(
vector_store_id="vs",
query="hello",
vector_store_search_optional_params={},
api_base=_ENGINE_BASE,
litellm_logging_obj=SimpleNamespace(model_call_details={}),
litellm_params={},
)
kwargs.update(overrides)
return VertexSearchAPIVectorStoreConfig().transform_search_vector_store_request(
**kwargs
)
def test_search_request_defaults_to_query_and_pagesize_10():
url, body = _search_request()
assert url == _ENGINE_BASE + ":search"
assert body == {"query": "hello", "pageSize": 10}
def test_search_request_maps_max_num_results_to_pagesize():
_, body = _search_request(
vector_store_search_optional_params={"max_num_results": 25}
)
assert body["pageSize"] == 25
def test_search_request_passes_datastorespecs_through_extra_body():
specs = [
{
"dataStore": "projects/p/locations/global/collections/default_collection/dataStores/ds-beta"
}
]
_, body = _search_request(extra_body={"dataStoreSpecs": specs})
assert body["dataStoreSpecs"] == specs
assert body["query"] == "hello"
assert body["pageSize"] == 10
def test_search_request_extra_body_takes_precedence_over_defaults():
_, body = _search_request(
vector_store_search_optional_params={"max_num_results": 5},
extra_body={"pageSize": 50, "filter": 'category: ANY("docs")'},
)
assert body["pageSize"] == 50
assert body["filter"] == 'category: ANY("docs")'
def test_search_request_joins_list_query():
_, body = _search_request(query=["foo", "bar"])
assert body["query"] == "foo bar"