Skip to content

Skip legacy models when selecting a Bedrock inference profile #72521

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
vincbeck wants to merge 1 commit into
base: main
Choose a base branch
from
+23 −2
Open

Skip legacy models when selecting a Bedrock inference profile #72521

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 23 additions & 2 deletions providers/amazon/tests/system/amazon/aws/utils/bedrock.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,6 +26,11 @@
from airflow.decorators import task # type: ignore[attr-defined, no-redef]


def _foundation_model_id(model_arn: str) -> str:
"""Return the model ID part of a foundation model ARN, which is identical in every region."""
return model_arn.rpartition("/")[2]


@task
def get_text_inference_profile_arn() -> str:
"""
Expand All @@ -37,11 +42,27 @@ def get_text_inference_profile_arn() -> str:
from airflow.providers.amazon.aws.hooks.bedrock import BedrockHook

client = BedrockHook().conn

# Bedrock rejects a model its provider marked as legacy, so a legacy model can not be relied on here.
# The inference profile summaries do not carry the lifecycle status, only the foundation models a
# profile resolves to do.
legacy_model_ids = {
model["modelId"]
for model in client.list_foundation_models()["modelSummaries"]
if model.get("modelLifecycle", {}).get("status") == "LEGACY"

@ramitkataria ramitkataria Sep 4, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should we filter for just ACTIVE so that if there's a new state like DEPRECATED or something else, this would still keep working?

}
log.info("Legacy model IDs: %s", sorted(legacy_model_ids))

profiles = client.list_inference_profiles(typeEquals="SYSTEM_DEFINED")["inferenceProfileSummaries"]
arns = [

@ramitkataria ramitkataria Sep 4, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Since the pick among non-legacy candidates still follows list_inference_profiles order, the model under test can change between runs. Would it make sense to add a deterministic tiebreak, e.g. sort the candidates and pick the oldest non-legacy Sonnet, so runs are reproducible and we don't land on a brand-new release before batch inference or RAG support it?

profile["inferenceProfileArn"]
for profile in profiles
if profile.get("status") == "ACTIVE" and profile["inferenceProfileId"].startswith("global.anthropic.")
if profile.get("status") == "ACTIVE"
and profile["inferenceProfileId"].startswith("global.anthropic.")
and not any(
_foundation_model_id(model["modelArn"]) in legacy_model_ids
for model in profile.get("models", [])
)
]
log.info("Valid text inference profile ARNs: %s", arns)

Expand All @@ -50,4 +71,4 @@ def get_text_inference_profile_arn() -> str:
if "sonnet" in arn:
log.info("Selected inference profile ARN: %s", arn)
return arn
raise RuntimeError("No valid inference profiles found")
raise RuntimeError(f"No valid inference profiles found. Non legacy candidates were: {arns}")
Loading