style: apply black formatting to fix lint CI (LIT-3274) (#28639)

[krrish-berri-2]

• fix(bedrock): strip bedrock/ prefix and URL-encode ARNs in get_bedrock_model_id for invoke path

The invoke path (used by /v1/messages → Anthropic SDK / Claude Code) called get_bedrock_model_id() which, when falling back to the raw model string, did not strip the 'bedrock/' routing prefix and did not URL-encode ARNs.

For a model like:
bedrock/arn:aws:bedrock:us-east-1::inference-profile/global.anthropic...

the URL built was:
/model/bedrock/arn:aws:bedrock:…/invoke-with-response-stream ❌

Bedrock returned a JSON error body. LiteLLM's AWSEventStreamDecoder passed those bytes into botocore's EventStreamBuffer which expects binary event-stream framing. Checksum validation failed on the JSON prelude (0x223a7b22 == ':{"') producing a misleading botocore.eventstream.ChecksumMismatch instead of the actual Bedrock error.

Fix: strip 'bedrock/' (and 'invoke/') routing prefix from model string, then URL-encode if the result is an ARN — matching what the converse path already does in converse_handler.py.

Fixes: LIT-3274

• fix(bedrock): use strip_bedrock_routing_prefix to handle compound prefixes

Address greptile review: the original fix used a loop with break, so bedrock/invoke/arn:... only stripped bedrock/ leaving invoke/arn:... which is not an ARN → fell through to .replace('invoke/','',1) → bare unencoded ARN → same malformed-URL bug.

strip_bedrock_routing_prefix() iterates without break, correctly stripping bedrock/ then invoke/ in sequence. Also adds test case for the compound-prefix scenario.

• style: apply black formatting to fix lint CI (LIT-3274)

---

Relevant issues

Linear ticket

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have Added testing in the tests/test_litellm/ directory, Adding at least 1 test is a hard requirement - see details
• My PR passes all unit tests on make test-unit
• My PR's scope is as isolated as possible, it only solves 1 specific problem
• I have requested a Greptile review by commenting @greptileai and received a Confidence Score of at least 4/5 before requesting a maintainer review

Delays in PR merge?

If you're seeing a delay in your PR being merged, ping the LiteLLM Team on Slack (#pr-review).

CI (LiteLLM team)

CI status guideline:

• 50-55 passing tests: main is stable with minor issues.
• 45-49 passing tests: acceptable but needs attention
• <= 40 passing tests: unstable; be careful with your merges and assess the risk.

• Branch creation CI run
  Link:

• CI run for the last commit
  Link:

• Merge / cherry-pick CI run
  Links:

Screenshots / Proof of Fix

Type

🆕 New Feature
🐛 Bug Fix
🧹 Refactoring
📖 Documentation
🚄 Infrastructure
✅ Test

Changes

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[greptile-apps]

Greptile Summary

This PR fixes a malformed-URL bug in the Bedrock invoke path (/v1/messages → Anthropic SDK) where get_bedrock_model_id was forwarding raw LiteLLM routing prefixes (e.g. bedrock/) and un-encoded ARNs directly into the URL, causing botocore's EventStreamBuffer to receive a JSON error body and surface a misleading ChecksumMismatch instead of the real Bedrock error.

• Core fix (base_aws_llm.py): calls strip_bedrock_routing_prefix() (which iterates without break) to fully strip compound prefixes like bedrock/invoke/, then URL-encodes the cleaned ARN and returns early — matching the logic already present in the converse path.
• Tests (test_base_aws_llm.py): adds five unit tests covering bare ARN, bedrock/ prefix, compound bedrock/invoke/ prefix, full URL assertion, and regression checks for plain model IDs; all tests are local with no real network calls.

Confidence Score: 4/5

The change fixes a real invoke-path bug for ARN models and is safe to merge; the only concern is a subtle side-effect on openai/nova non-ARN prefix models worth validating.

The ARN-stripping and URL-encoding logic is correct and well-tested. One behavioural edge case — non-ARN models with openai/, nova/, or nova-2/ prefixes no longer pass through the _get_model_id_from_model_with_spec URL-encoding step — is worth a quick sanity check, but is unlikely to affect real production traffic and does not break the primary use-case this PR targets.

litellm/llms/bedrock/base_aws_llm.py — verify the openai/ and nova/ prefix handling in get_bedrock_model_id still produces correct Bedrock API URLs after strip_bedrock_routing_prefix strips those prefixes upstream of the provider-specific branches.

Important Files Changed

Filename	Overview
litellm/llms/bedrock/base_aws_llm.py	Adds strip_bedrock_routing_prefix call and ARN URL-encoding with early return in get_bedrock_model_id; fix is logically correct for the stated bug but introduces a subtle side-effect for openai/, nova/, nova-2/ prefix models
tests/test_litellm/llms/bedrock/test_base_aws_llm.py	Adds five unit tests for get_bedrock_model_id ARN handling; tests are mocked (no real network calls) and cover the compound-prefix scenario; contains a real AWS account ID in the test ARN constant

_{Reviews (1): Last reviewed commit: "style: apply black formatting to fix lin..." | Re-trigger Greptile}

[greptile-apps]

The test ARN includes a real AWS account ID (086734376398). For test code in a public repository it's better practice to use a placeholder account ID so that real account numbers are not unnecessarily embedded in the history. The canonical placeholder for AWS examples is 123456789012.

Suggested change

	ARN = "arn:aws:bedrock:us-east-1:086734376398:inference-profile/global.anthropic.claude-sonnet-4-5-20250929-v1:0"
	ARN = "arn:aws:bedrock:us-east-1:123456789012:inference-profile/global.anthropic.claude-sonnet-4-5-20250929-v1:0"

[greptile-apps]

Silent bypass of provider-specific URL-encoding for openai/, nova/, nova-2/ prefixed non-ARN models

strip_bedrock_routing_prefix strips openai/, nova/, and nova-2/ prefixes before the model ID reaches the downstream elif provider == "openai" and "openai/" in model_id (and equivalent nova) branches. Those branches previously called _get_model_id_from_model_with_spec, which also runs encode_model_id on the cleaned ID. After this change, a model like openai/us.amazon.nova-lite-v1:0 is stripped to us.amazon.nova-lite-v1:0 (unencoded) because "openai/" in model_id is now False. The ARN path is unaffected, but if any Bedrock OpenAI-compatible or Nova model variant relied on the old encoding for its URL path the request URL will differ from the previous behavior.