chore(codegen): daily schema refresh (2026-05-18)

[timgl]

Summary

Daily OpenAPI schema refresh against https://us.posthog.com/api/schema/?format=json.

Spec diff size: 15 lines added, 1 removed in src/generated/api.d.ts. Purely docstring additions on TrendsFilter:

• aggregationAxisFormat — full description of the supported formats (numeric, duration, duration_ms, percentage, percentage_scaled, currency, short).
• aggregationAxisPostfix / aggregationAxisPrefix — guidance on when to use literal suffixes/prefixes vs picking the matching format.
• decimalPlaces — note on typical values for percentages and currency.

No structural schema changes. openapi-filter.yaml unchanged.

Filter changes

None. No new operationIds added.

Resources touched

None. No client/pipeline edits required.

New operationIds in live spec (not added to filter)

The live spec exposes a very large set of operationIds outside the managed resource families (action, cohort, dashboard, endpoint, event-definition, experiment, experiment-holdout, experiment-saved-metric, feature-flag, insight, project-settings, property-group). None of them are missing core CRUD ops for a managed family, so nothing was added.

Notable categories that remain intentionally out of scope:

• Adjacent resource families (not managed by posthog-definitions): persons_*, groups_*, cohorts_persons_*, external_data_*, warehouse_*, llm_analytics_*, hog_flows_*, hog_functions_*, session_recordings_*, notebooks_*, surveys_*, error_tracking_*, logs_*, subscriptions_*, roles_*, members_*, organizations_*, tasks_*, tracing_spans_*, visual_review_*, signals_*, etc.
• environments_* mirrors of project-scoped routes (e.g. environments_insights_*, environments_dashboards_*) — duplicates of the routes we already use under projects/*.
• Extras on managed families that are not core CRUD and not currently needed:
  • actions_bulk_update_tags_create, actions_references_list
  • cohorts_activity_retrieve, cohorts_all_activity_retrieve, cohorts_calculation_history_retrieve, cohorts_persons_retrieve, cohorts_add_persons_to_static_cohort_partial_update, cohorts_remove_person_from_static_cohort_partial_update, cohorts_update (PUT — we use PATCH)
  • dashboards_* activity/collaborator/sharing/snapshot/copy/move/reorder helpers, dashboards_bulk_update_tags_create
  • endpoints_* materialization/run/versions/last_execution_times helpers
  • event_definitions_bulk_update_tags_create, event_definitions_by_name_retrieve, event_definitions_metrics_retrieve, event_definitions_primary_properties_retrieve, event_definitions_{golang,python,typescript}_retrieve, event_definitions_update (PUT)
  • event_schemas_partial_update, event_schemas_update (we only need list/create/destroy server-side)
  • experiment_holdouts_update (PUT)
  • experiment_saved_metrics_update (PUT)
  • experiments_copy_to_project_create, experiments_create_exposure_cohort_for_experiment_create, experiments_duplicate_create, experiments_eligible_feature_flags_retrieve, experiments_recalculate_timeseries_create, experiments_requires_flag_implementation_retrieve, experiments_reset_create, experiments_ship_variant_create, experiments_stats_retrieve, experiments_timeseries_results_retrieve, experiments_update (PUT)
  • feature_flags_* activity/dashboard/blast-radius/local-evaluation/remote-config/status/versions helpers, feature_flags_bulk_*
  • insights_* activity/sharing/suggestions/thresholds/trending/viewed helpers, insights_bulk_update_tags_create, insights_analyze_retrieve, insights_my_last_viewed_retrieve, insights_cancel_create, insights_generate_metadata_create
  • schema_property_groups_update (PUT)

Convention: managed families intentionally use PATCH (*_partial_update) and skip PUT (*_update) where both exist, except where the existing filter already includes _update (actions, insights, dashboards, feature_flags, endpoints) — left as-is in this PR.

Unresolved drift

• One pre-existing flaky test: src/pull/run.test.ts > runPull orchestrator > writes files and calls tagOnServer on a non-dry run times out at 500ms under full-suite load (~318ms in isolation). Not caused by this refresh — schema change is doc-only. Suggest bumping the per-test timeout or relaxing this case.
• No operationIds removed from the live spec relative to the current filter.

Test plan

• pnpm install --frozen-lockfile
• pnpm codegen
• pnpm typecheck
• pnpm test (1 flaky timeout unrelated to schema; passes in isolation)

TaskRun

Task-Id: dc9b071d-e37e-442f-aec0-cc3315a3c350