🐛 Bug Report: Bedrock profile IDs with a dot in the model name (us.openai.gpt-5.6-sol) are recorded as model 'gpt-5' #4501

Thanks for keeping _get_vendor_model in step with Bedrock's inference profile formats (#2785); this is one more ID shape it now sees.

Which component is this bug for?

Bedrock Instrumentation

📜 Description

_cross_region_check (__init__.py) splits geo-prefixed IDs with value.split(".") and keeps the piece after the vendor, so a dotted model name is cut. us.openai.gpt-5.6-sol (and its profile ARN) resolves to model gpt-5 in the span name and gen_ai.request.model. The prefix-less branch uses split(".", 1) and is correct.

Found while checking GPT-6 on Bedrock: Sol, Luna and Astra model names have no dot; GPT-5.6 and us.xai.grok-4.6 do.

👟 Reproduction steps

from opentelemetry.instrumentation.bedrock import _get_vendor_model
_get_vendor_model("us.openai.gpt-5.6-sol")
_get_vendor_model("arn:aws:bedrock:us-east-1:111122223333:inference-profile/us.openai.gpt-5.6-sol")

👍 Expected behavior

('aws.bedrock', 'openai', 'gpt-5.6-sol'), the same as for the prefix-less openai.gpt-5.6-sol.

👎 Actual Behavior with Screenshots

Main d13e721, with real converse calls in us-east-1:

_get_vendor_model('us.openai.gpt-5.6-sol') -> ('aws.bedrock', 'openai', 'gpt-5')
_get_vendor_model('openai.gpt-5.6-sol') -> ('aws.bedrock', 'openai', 'gpt-5.6-sol')
_get_vendor_model('us.openai.gpt-6-astra') -> ('aws.bedrock', 'openai', 'gpt-6-astra')
[converse us.openai.gpt-5.6-sol] span='chat gpt-5' gen_ai.request.model='gpt-5'
[converse us.openai.gpt-5.6-luna] span='chat gpt-5' gen_ai.request.model='gpt-5'
[converse us.xai.grok-4.6] span='chat grok-4' gen_ai.request.model='grok-4'

🤖 Python Version

3.12

📃 Provide any additional context for the Bug.

Proposed fix: value.split(".", 2) in the prefixed branch; PR with tests to follow. My open #4235 (adds global two lines above) needs it too: alone it turns global.openai.gpt-5.6-sol into gpt-5, merged with this fix (clean, 386 tests pass) it gives gpt-5.6-sol.

_get_vendor_model('us.openai.gpt-5.6-sol') -> ('aws.bedrock', 'openai', 'gpt-5.6-sol')
[converse us.openai.gpt-5.6-sol] span='chat gpt-5.6-sol' gen_ai.request.model='gpt-5.6-sol'
[converse us.openai.gpt-5.6-luna] span='chat gpt-5.6-luna' gen_ai.request.model='gpt-5.6-luna'
[converse us.xai.grok-4.6] span='chat grok-4.6' gen_ai.request.model='grok-4.6'
[converse us.openai.gpt-6-sol] span='chat gpt-6-sol' gen_ai.request.model='gpt-6-sol'
[converse us.openai.gpt-6-luna] span='chat gpt-6-luna' gen_ai.request.model='gpt-6-luna'
[converse us.openai.gpt-6-astra] span='chat gpt-6-astra' gen_ai.request.model='gpt-6-astra'

👀 Have you spent some time to check if this bug has been raised before?

• I checked and didn't find similar issue

Searching open and closed issues/PRs for gpt-5.6, grok, _cross_region_check and inference profile finds nothing on this; the closest are my #4234 / #4235 (global. prefix).

Are you willing to submit PR?

Yes I am willing to submit a PR!