googleapis / googleapis/python-aiplatform
find_neighbors with REST transport fails unless project and location are set in aiplatform.init
- 主要语言
- Python
- 星标
- 905
- 派生
- 465
- 平均合并
- 1 天 13 小时
- 30 天内合并 PR
- 44
描述
`MatchingEngineIndexEndpoint.find_neighbors` fails when using `api_transport="rest"` if `project` and `location` are not explicitly passed to `aiplatform.init()`. The gRPC transport works correctly without these explicit parameters.
The REST call results in a `400 BadRequest` error: `Could not find field "%24alt" in the type "google.cloud.aiplatform.v1beta1.FindNeighborsRequest"`.
Is it intended that the REST transport requires explicit `project` and `location` in `init`, while gRPC can infer them from the endpoint's resource name? The behavior seems inconsistent.
**Code Example:**
```python
from google.cloud import aiplatform
# Fails with the below configuration
aiplatform.init(api_transport="rest")
# This configuration works
# aiplatform.init(api_transport="grpc")
# This configuration also works
# from google.cloud.aiplatform.initializer import initializer
# aiplatform.init(
# project=GCP.project,
# location=GCP.location,
# api_transport="rest"
# )
def get_docs_search_endpoint():
return aiplatform.MatchingEngineIndexEndpoint(Artifacts.docs_search_endpoint_id)
docs_search_endpoint = get_docs_search_endpoint()
try:
match_resp = docs_search_endpoint.find_neighbors(
deployed_index_id=Artifacts.docs_search_deployment_id,
queries=[[1] * 768],
num_neighbors=500,
)
print(match_resp)
except Exception as e:
print(e)
```
**Stack Trace:**
```
google.api_core.exceptions.BadRequest: 400 POST https://209XXXX46.us-central1-901XXXX18323.vdb.vertexai.goog/v1beta1/projects/901XXXX8323/locations/us-central1/indexEndpoints/793471XXXX752470528:findNeighbors?%24alt=json%3Benum-encoding%3Dint: Could not find field "%24alt" in the type "google.cloud.aiplatform.v1beta1.FindNeighborsRequest".
```
贡献指南
调研方向
从 MatchingEngineIndexEndpoint.find_neighbors 和 stack trace 中显示的 REST 请求开始。比较 aiplatform.init(api_transport="rest") 和 gRPC 传输在省略 project 和 location 时如何获取它们。完成标准是 REST 能够一致地推断这些值,或者预期要求已得到明确记录并由回归测试覆盖。
由索引模型根据 Issue 内容生成。
评估
- 技术栈
- python
- 领域
- api, machine-learning
- Issue 类型
- 缺陷
- 难度
- 4/5
- 预计耗时
- 3-5 天
- 活跃度
- 停滞
- 描述清晰度
- 基本清楚
- 新手友好度
- 35/100