chris/OSVauco
1
0
Fork 0
Code Issues 1 Pull Requests Actions 2 Packages Projects Releases Wiki Activity

fix: drop backend_config from create_corpus — rely on pre-flight PATCH alone

Browse Source 
SDK 1.153.1 has a bug in set_backend_config where it calls .CopyFrom()
on RagManagedDb which is not a real protobuf message.
The project-level ragEngineConfig PATCH already sets basic tier,
so passing backend_config is unnecessary and breaks corpus creation.
This commit is contained in:
chrischristiansen-glitch 2026-05-24 19:04:31 +02:00
parent 46b3c52607
commit 2beb6a359d
1 changed files with 11 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

28
agents/rag/setup_corpus.py
View File

@ -3,12 +3,9 @@
setup_corpus.py — Create a Vertex AI RAG Engine corpus and import documents.
Project: propane-will-491900-m5
Explicitly uses Serverless (RagManagedDb) to avoid Spanner allowlist
restrictions on new projects in us-central1.
Tested against vertexai SDK 1.153.1:
RagVectorDbConfig(vector_db=RagManagedDb())
RagManagedDb is a no-arg marker class — selecting it is sufficient for serverless.
Serverless mode is controlled via the project-level ragEngineConfig PATCH
(ensure_serverless_engine_config). Passing backend_config to create_corpus
triggers a SDK 1.153.1 bug and must be omitted.
"""
import os
@ -31,7 +28,8 @@ GCS_SOURCE = os.environ.get(
def ensure_serverless_engine_config() -> None:
"""
Pre-flight: set project-level RAG Engine Config to basic (serverless) tier.
Idempotent — safe to re-run.
This is what actually controls the backend — backend_config in create_corpus
is intentionally omitted due to SDK bug in 1.153.1.
"""
print("Ensuring RAG Engine Config is set to serverless (basic) tier...")
endpoint = (
@ -71,21 +69,17 @@ def ensure_serverless_engine_config() -> None:
def get_or_create_corpus() -> RagCorpus:
"""Return existing corpus by display name, or create a new serverless one."""
"""Return existing corpus by display name, or create a new one."""
for c in rag.list_corpora():
if c.display_name == CORPUS_DISPLAY_NAME:
print(f"Corpus '{CORPUS_DISPLAY_NAME}' already exists: {c.name}")
return c
print(f"Creating corpus '{CORPUS_DISPLAY_NAME}' in {LOCATION} (serverless)...")
corpus = rag.create_corpus(
display_name=CORPUS_DISPLAY_NAME,
# RagManagedDb() is a no-arg marker class in SDK 1.153.1
# Selecting it routes to serverless managed DB (basic tier)
backend_config=rag.RagVectorDbConfig(
vector_db=rag.RagManagedDb()
),
)
print(f"Creating corpus '{CORPUS_DISPLAY_NAME}' in {LOCATION}...")
# NOTE: backend_config is intentionally omitted.
# SDK 1.153.1 has a bug where passing any backend_config crashes.
# Serverless mode is already guaranteed by ensure_serverless_engine_config().
corpus = rag.create_corpus(display_name=CORPUS_DISPLAY_NAME)
print(f"✓ Corpus created: {corpus.name}")
return corpus