[Tracker] Python 3.15 integration parity

[vlad-scherbich]

Python 3.15 integration parity tracker

This issue tracks bringing dd-trace-py integration coverage on Python 3.15
to parity with the 3.14 release. The list below is the inverse of the
dd-trace-py v3.16.0 release notes "still does not work with Python 3.14"
section — for 3.15 we expect the same set of items to be skipped until each
row's blocker is resolved.

How this works

• Each checkbox below should result in one independent PR.
  The 3.14 migration followed the same pattern (see @emmettbutler's PR series linked at the bottom).
• Each unchecked checkbox below is also a real GitHub sub-issue (created via the task-list "convert to issue" option on each bullet). The parent checkbox auto-flips when the sub-issue closes. Rows already pre-checked (- [x]) are already merged on main and don't need a sub-issue.
• To claim a row: click into the row's sub-issue (each - [ ] bullet links to one), then either self-assign on that sub-issue page (top-right "Assignees" → "assign yourself") or leave a Taking this comment on the sub-issue itself and a maintainer will assign you. Don't claim on this parent tracker — comments here are ambiguous about which row and add noise.
• PRs that close a row should use the
  python_315_bump
  PR template (added in #17791; open via ?template=python_315_bump.md) and include Closes #<sub-issue-number> in the description.
  That auto-closes the sub-issue on merge, which auto-flips the parent checkbox.
• Most rows are a one-line riotfile.py max_version= bump + riot generate <suite> regen + lockfile commit. A few are real integration work and are called out below.

Tier A: wheel-build prerequisites

These must land before we can flip the cp315 wheel job from allow-fail to required. Scope is intentionally narrow — anything that can be skipped at runtime on 3.15 the way 3.14 skipped Profiling / IAST / ci_visibility (see dd-trace-py v3.16.0 release notes) lives in Tier B, not here.

Owned

• [3.15] Tracer wrapping 3.15 #17810
• [3.15] Lib-injection max-Python bump in lib-injection/sources/requirements.csv (analogue of #12283) #17813
• [3.15] Regen supported_versions_table.csv, supported_versions_output.json, requirements.csv for 3.15 #17814
• [3.15] Add Programming Language :: Python :: 3.15 classifier to pyproject.toml #17815
• [3.15] cp315 smoke tests (tests/smoke_test.py, tests/lib-injection) green → flip cp315 wheel job from allow-fail to required #17816
• [3.15] Crashtracker/native: pyo3 0.28 + PyFrame_GetBack removal on 3.14/3.15 #18473

Completed

• [3.15] Test-runner / pyenv 3.15-dev #17946
• [3.15] CI Visibility coverage instrumentation for Python 3.13+ (partial) #17945
• [3.15] Crashtracker 3.15 verification #17811
• [3.15] Native build / requires-python / cp315 wheel allow-fail #17812

Tier B: integration parity

Notes:

• None of these rows block the cp315 wheel cutover. They are integration-parity work that ships as each row lands, same model as the dd-trace-py v3.16.0 release notes' "still does not work with Python 3.14" list (those wheels still shipped). Each row failing only means that one integration / product is silently skipped on 3.15.
• All caps in riotfile.py are at max_version="3.13" (not "3.14"). A volunteer PR must therefore do a 3.13 → 3.15 jump and validate the integration on both 3.14 and 3.15, then remove any # X doesn't yet work with Python 3.14 comments left over from previous iterations.
• A few rows are more nuanced than the dd-trace-py v3.16.0 release notes flatten them; see notes inline below.

Owned

Profiling — @DataDog/profiling-python

• [3.15] Profiling 3.15 #17817

Frameworks / IDM — @DataDog/apm-idm-python

• [3.15] django / django:celery / django:django_hosts / django:djangorestframework — enable on Python 3.15 #17829
• [3.15] dramatiq — enable on Python 3.15 #17830
• [3.15] kafka — enable on Python 3.15 #17831
• [3.15] sqlite3 — enable on Python 3.15 #17832
• [3.15] grpc_aio — enable on Python 3.15 #17833
• [3.15] rq — enable on Python 3.15 #17834

Needs an owner

LLM Observability — @DataDog/ml-observability

Most are blocked on tiktoken or protobuf cutting a 3.15 release; bump the upstream pin and regen the lockfile.

• [3.15] openai — enable on Python 3.15 #17818
• [3.15] langchain — enable on Python 3.15 #17819
• [3.15] langgraph — enable on Python 3.15 #17820
• [3.15] litellm — enable on Python 3.15 #17821
• [3.15] crewai — enable on Python 3.15 #17822
• [3.15] ai_guard_langchain — enable on Python 3.15 #17823
• [3.15] google_generativeai — enable on Python 3.15 #17824
• [3.15] vertexai — enable on Python 3.15 #17825
• [3.15] openai_agents — enable on Python 3.15 #17826
• [3.15] llmobs (ragas test path) — re-pin ragas #17827
• [3.15] ray — enable on Python 3.15 #17828

API SDK — @DataDog/apm-sdk-capabilities-python

• [3.15] opentelemetry — enable on Python 3.15 #17835

Serverless — @DataDog/apm-serverless

• [3.15] aws_lambda — enable on Python 3.15 #17836
• [3.15] azure_functions — enable on Python 3.15 #17837

CI Visibility — @DataDog/ci-app-libraries

• [3.15] CI Visibility coverage broader 3.13-3.15 #17838
• [3.15] pytest integration — real integration work #17839
• [3.15] pytest_plugin_v2 — enable on Python 3.15 #17840
• [3.15] pytest_bdd / pytest_benchmark / unittest / freezegun / selenium — enable on Python 3.15 #17841
• [3.15] dd_coverage — enable on Python 3.15 #17842

ASM — @DataDog/asm-python

• [3.15] IAST 3.15 #17843
• [3.15] appsec_integrations_fastapi — enable on Python 3.15 #17844

Data Streams — @DataDog/data-streams-monitoring

• [3.15] datastreams — enable on Python 3.15 #17845

Completed

Non-goals (no action required for the py-315 migration itself)

The integrations below are explicitly out of scope for the 3.15 effort. They were already capped below 3.14
long before 3.15 came up, so flipping on cp315 wheels does not change anything for them.
They are listed here only as an opportunity for a future cleanup / deprecation initiative to happen independently
of the 3.15 cycle (all are likely owned by @DataDog/apm-idm-python / @DataDog/apm-core-python).

• vertica (capped at 3.9) — already deprecated in dd-trace-py v3.16.0; on its way out.
• bottle (capped at 3.10) — Bottle the framework is dormant upstream.
• aredis (capped at 3.9) — library abandoned upstream; superseded by redis>=4.2.0.
• yaaredis (capped at 3.10) — library abandoned upstream; superseded by redis>=4.2.0.
• rediscluster (capped at 3.11) — redis-py-cluster superseded by native cluster support in redis>=4.2.0.
• asynctest (capped at 3.9) — library obsoleted by stdlib unittest.IsolatedAsyncioTestCase in Python 3.8.
• pynamodb (capped at 3.11) — test-harness incompat at 3.12+; orthogonal to 3.15 (fixing it would unblock 3.12-3.15 in one shot). Belongs in a general test-hygiene backlog.

If any of these turn out to need attention, file a separate issue —
don't reopen this tracker for them.

References

• dd-trace-py v3.16.0 release notes — source list of "doesn't work on 3.14"
• PR #14264 — initial 3.14 support — riotfile bump precedent
• emmettbutler's 3.14 follow-up PR series — pattern for per-integration enablement
• PR #17624 — profiling 3.15 support — Tier B reference (in flight)
• PR #17791 — adds python_315_bump PR template — referenced from "How this works" above; the ?template=python_315_bump.md URL only resolves once this merges