[hassiebp]

What does this PR do?

Removes the available_skills parameter from the observation APIs (start_observation, start_as_current_observation, create_event, update, update_current_span, update_current_generation and the observation constructors). From now on, skills are set only through propagation:

• propagate_attributes(skills_available=[...]) writes the list onto the active observation and every observation created inside the context.
• propagate_attributes(attach_skills=True) attaches the catalog configured with Langfuse(skills={"available": ...}).

The per-observation parameter did the same job as propagation but had to be repeated on every typed overload, which made up most of #2062's API surface. It also used a different name (available_skills) from the propagation key and the span attribute (skills_available). The feature hasn't been released yet, so removing it now doesn't break anyone.

What stays:

• the langfuse.observation.skills_available attribute and its reserved slot in the metadata budget, since propagation still writes it onto observations that have already started;
• the span processor rule that skills already on a span beat propagated skills, which still applies to spans from other OpenTelemetry instrumentation;
• discover_skills, SkillReference and SkillsConfig.

The matching JS change is in langfuse/langfuse-js on the branch of the same name.

Type of change

• Bug fix
• New feature
• Breaking change
• Refactor
• Documentation update
• Tooling, CI, or repo maintenance

Verification

uv run --frozen ruff format .                       # 578 files left unchanged
uv run --frozen ruff check .                        # All checks passed!
uv run --frozen mypy langfuse --no-error-summary    # exit 0
uv run --frozen pytest -n 4 --dist worksteal tests/unit -q   # 1001 passed, 2 skipped

Removed tests that only covered the per-observation parameter. Two tests that pin useful behavior now use propagation instead: propagation replaces skills on the active observation (including with []), and the metadata budget still leaves room for skills added after the span starts.

E2E and live-provider suites were not run; nothing they cover changed.

Checklist

• I self-reviewed the diff using code_review.md.
• I added or updated tests for behavior changes.
• I updated docs, examples, or .env.template if needed.
• I did not hand-edit generated files; if generated files changed, I used the upstream regeneration path.
• I did not commit secrets or credentials.

  

Confidence Score: 5/5

The PR appears safe to merge.

Summary

Removes available_skills from observation creation, constructors, and updates. Skills now come only from propagate_attributes.

• Callers set observation skills through propagation, not individual observation calls.

Acknowledged by hassiebp: removing the per-observation parameter is intentional because the feature has not been released.

_{Reviews (1) · Last reviewed commit: "refactor(skills): set skill availability..." · Reviewed by Greptile}

[github-actions]

@claude review

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-30T15:06:42.263636Z	5eff29b	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[claude]

Claude Code Review

No review was started: this request came from a bot account. Manual reviews can only be requested by someone with write access to this repository. Ask a maintainer to comment @claude review, or have your automation post the comment from a user account with write access.

_{Tip: disable this comment in your organization's Code Review settings.}

[claude]

LGTM. This removes the per-observation available_skills parameter as a straight deletion, and nothing references it anymore.

What was reviewed:

• The diff for client.py, span.py and attributes.py. Every removed line is either an available_skills param, docstring or kwarg, or one of the unused SkillReference/serialize_available_skills imports. None of the remaining logic changed.
• A repo-wide search for leftover available_skills or create_available_skills_attributes callers. The only hits left are the span processor's own _available_skills field and its test fixtures, plus serialize_available_skills in propagation.py, span_processor.py and skills.py.
• The rewritten tests. propagate_attributes is already imported in test_metadata_attribute_limit.py. _propagate_attributes writes skills_available onto the current span, so the new assertions on the parent span still hold.

Extended reasoning...

The change removes the available_skills parameter from the Langfuse and observation-wrapper APIs and the attribute helpers, and drops or rewrites the tests that used it. It adds 19 lines and deletes 239. It touches no security-sensitive surface such as auth, crypto or data exposure. CODEOWNERS covers only .github/, and the conversation has no outstanding reviewer objections. I approved because the change is purely subtractive, the search found no leftover callers, and skills can still be set through the existing propagate_attributes(skills_available=...) path, which the updated tests cover.