refactor(client): remove two unused OpenLLMetry span helpers

[apucacao]

Summary

set_openllmetry_prompt and set_openllmetry_completion have had no callers since the span work moved every handler onto the content layer in content.py. TELEMETRY-CONTRACT.md said to keep them exported for one release and then remove them, and that release has shipped. This deletes them, along with the parity-test code that only existed to skip them, and updates the contract doc to match.

Testing

• uv run pytest tests/test_cross_handler_parity.py: 87 passed, 11 skipped.
• ruff check passes on the changed files.
• Leaving the full suite and mypy to CI.

Matching TypeScript cleanup: launchdarkly/js-ai-sdk#88

Original branch: ag/remove-openllmetry-helpers

---

Note

Overview
Removes legacy set_openllmetry_prompt and set_openllmetry_completion from the client package after handlers already write prompts, completions, and usage through content.py and set_usage_span_attributes.

The helpers are dropped from utils.py, unexported from launchdarkly_ai_server, and TELEMETRY-CONTRACT.md section 11 now states they must not return. The cross-handler parity test no longer strips “superseded” function bodies before the vocabulary lock scan—that workaround existed only while dead code still lived in utils.py.

^{Reviewed by Cursor Bugbot for commit 0be21dd. Bugbot is set up for automated code reviews on this repo. Configure here.}

[apucacao]

bugbot run

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 0be21dd. Configure here.}

[jeffdupont]

Reviewed with the 1.0 freeze in mind. Nothing blocking. This is item D in the 1.0 API-surface plan, and it's the right call: both helpers go and nothing breaks. On 0be21dd, make test passes (1316 passed, 11 skipped), and make typecheck and make lint are clean. It also merges cleanly onto current main (fee904a).

What I checked:

• Nothing calls them. On current main, outside utils.py and __init__.py, the only mentions are the parity test's skip list and the contract doc, and this PR removes both. There's nothing in examples or docs. The gen_ai.prompt.N.* / gen_ai.completion.N.* keys Monitoring reads are still written by content.py (_set_openllmetry_messages, line 388). The prompt_tokens / completion_tokens aliases still come from set_usage_span_attributes, which test_span_usage.py:301 pins. So the span data contract doesn't change, only the public names do.
• JS parity is done. launchdarkly/js-ai-sdk#88 merged on 1 Oct, and js-ai-sdk main has no setOpenLLMetryPrompt / setOpenLLMetryCompletion left. JS never exported them from index.ts, so its public surface doesn't move.
• No spec change needed. ai-sdks-monorepo main doesn't name either function, and the attribute keys are unchanged.

Smaller notes:

• These were public. Both names are in __all__ in every released launchdarkly-ai-server from 0.2.0 through 0.2.4, so anyone importing them will break on the next release. Removing them before 1.0 is fine. But with refactor(client): and no ! or BREAKING CHANGE: footer, release-please won't mention the removal in the changelog. I'd add one so the release notes say what went. Details inline.
• The old contract text asked for direct tests of the replacements. Dropping that ask is fine now: tests/test_ld_span_attributes.py and the usage-alias test in test_span_usage.py cover them.

[jeffdupont]

These two names were in __all__ from 0.2.0 through 0.2.4, so this is a breaking removal for anyone who imported them, even though nothing in the repo does. Pre-1.0 that's fine, and it matches the API-surface plan. Could the squash commit be refactor(client)!: or carry a BREAKING CHANGE: footer? Then release-please lists the removal in the changelog instead of hiding it under a refactor. With bump-minor-pre-major set, that makes the next release 0.3.0 rather than 0.2.5. I think that's the honest version number, but your call.