Add common gen AI utils into opentelemetry-instrumentation

[aabmass]

Description

Part of #3191

(Copied out of cohere PR #3081)

Type of change

Refactor

How Has This Been Tested?

Please describe the tests that you ran to verify your changes. Provide instructions so we can reproduce. Please also list any relevant details for your test configuration

• Added tests

Does This PR Require a Core Repo Change?

• Yes. - Link to PR:
• No.

Checklist:

See contributing.md for styleguide, changelog guidelines, and more.

• Followed the style guidelines of this project
• Changelogs have been updated
• Unit tests have been added

[aabmass]

Lint and tests for OpenAI are failing because we test against oldest published supported dependencies

opentelemetry-python-contrib/tox.ini

Lines 427 to 429 in 07c97ea

	# packages that are released individually should provide a test-requirements.txt with the lowest version of OTel API
	# and SDK supported to test we are honoring it
	openai-0: -r {toxinidir}/instrumentation-genai/opentelemetry-instrumentation-openai-v2/test-requirements-0.txt

opentelemetry-python-contrib/instrumentation-genai/opentelemetry-instrumentation-openai-v2/test-requirements-0.txt

Lines 12 to 14 in 07c97ea

	opentelemetry-api==1.28 # when updating, also update in pyproject.toml
	opentelemetry-sdk==1.28 # when updating, also update in pyproject.toml
	opentelemetry-semantic-conventions==0.49b0 # when updating, also update in pyproject.toml

To move this PR forward we either need to

1. Disable "oldest supported" tests py3x-test-instrumentation-openai-v2-0
2. Wait until next release of opentelemetry-instrumentation is published to update OpenAI code

[aabmass]

2. Wait until next release of opentelemetry-instrumentation is published to update OpenAI code

I decided to do this, I've split the openai changes for now. I'll make a tracking bug to update the code once this PR is merged.

[lmolkova]

model name is not always available, we should return name in this case

[lmolkova]

maybe

Suggested change

	def handle_span_exception(span: Span, error: Exception) -> None:
	def end_span_with_exception(span: Span, error: Exception) -> None:

?

we're not handling anything, right?

BTW - there is nothing gen-ai specific here, should we move it to common instrumentation helpers?

[aabmass]

Tbh I'm kind of just mechanically copying from openai-v2:

opentelemetry-python-contrib/instrumentation-genai/opentelemetry-instrumentation-openai-v2/src/opentelemetry/instrumentation/openai_v2/utils.py

Lines 232 to 238 in a716949

	def handle_span_exception(span, error):
	span.set_status(Status(StatusCode.ERROR, str(error)))
	if span.is_recording():
	span.set_attribute(
	ErrorAttributes.ERROR_TYPE, type(error).__qualname__
	)
	span.end()

Actually, the API/SDK has a mechanism here when used in start_as_current_span ctx manager. That would also call span.record_exception() which I believe add a span event with the stacktrace.

cc @lzchen @alizenhom do you know why OpenAI doesn't just use the existing mechanism?

[aabmass]

Putting this back in draft for now, see the comments on linked issue #3191 (comment)

[github-actions]

This PR has been automatically marked as stale because it has not had any activity for 14 days. It will be closed if no further activity occurs within 14 days of this comment.
If you're still working on this, please add a comment or push new commits.

[github-actions]

This PR has been closed due to inactivity. Please reopen if you would like to continue working on it.