chore(codegen): daily schema refresh (2026-06-14)

[timgl]

Automated daily OpenAPI schema refresh for 2026-06-14.

Spec diff size

• src/generated/api.d.ts: +1640 / −127 lines (regenerated via pnpm codegen; never hand-edited).
• Drift is transitive: the managed insight / dashboard / query request+response bodies now reference new upstream query-node component schemas (e.g. AccountsQuery, CalendarHeatmapFilter) and gained new fields. No allowlisted operation changed shape in a way that breaks the Zod schemas.

New operationIds added to the filter

None. All 100 new operationIds that belong to an already-managed resource family are either auxiliary product endpoints (activity logs, bulk tag updates, sharing, tile ops, stats, materialization, versions, run) or redundant full-PUT _update variants of resources we already manage via PATCH. None extend the consumed CRUD surface of the IaC resources, so adding them would only bloat the generated types with no consumer. Breakdown by family:

Family	New aux/PUT ops (not added)
dashboards_	22 (collaborators, tiles, sharing, widgets, bulk tags, …)
insights_	17 (activity, sharing, thresholds, suggestions, bulk tags, …)
feature_flags_	17 (activity, bulk ops, status, remote_config, versions, …)
experiments_	13 (duplicate, reset, stats, timeseries, ship_variant, …)
environments_endpoints_	8 (logs, run, versions, materialization, …)
event_definitions_	8 (by_name, metrics, codegen variants, _update, …)
cohorts_	8 (activity, persons, used_in, _update, …)
actions_	2 (bulk_update_tags, references)
event_schemas_	2 (_partial_update, _update)
experiment_holdouts_ / experiment_saved_metrics_ / schema_property_groups_	1 each (_update PUT)

Brand-new resource families in the live spec (accounts, alerts, annotations, surveys, tasks, error_tracking, warehouse, …) are not adopted here — taking on a new managed family is a deliberate product decision that should go through the add-resource skill with its own client, pipeline, and tests, not an unattended daily refresh.

Resources touched

• src/resources/dashboard/client.ts — upstream renamed the dashboards PATCH request body component PatchedDashboard → PatchedPatchedDashboardOpenApi. Updated the one type alias (PatchedDashboardBody) to point at the new name. The new schema carries the same fields the code uses (restriction_level, tiles, delete_insights); no behavior change. This was the only schema-drift breakage; pnpm typecheck now passes.

No other client.ts Zod schemas or pipeline.ts projections needed changes.

Unresolved drift / flags for a human

• Removed operationIds: none — every one of the 68 allowlisted operations still exists in the live spec.
• Pre-existing test flake (not caused by this PR): src/pull/run.test.ts > writes files and calls tagOnServer on a non-dry run intermittently hits its 500 ms per-test timeout under the full parallel suite (the tsx collect phase is ~35 s and starves the timer). It passes reliably in isolation (vitest run src/pull/run.test.ts). Untouched by this change; flagged for a separate timeout bump.
• Pre-existing repo style debt (not touched): pnpm lint and pnpm format:check report failures across ~68 files that also fail on main (e.g. the ternary at dashboard/client.ts:49, untouched here). Out of scope for a schema refresh.
• Candidate future additions (human call): if we ever want write-through parity, the full-PUT _update ops (experiments_update, cohorts_update, event_definitions_update, event_schemas_update/_partial_update, experiment_holdouts_update, experiment_saved_metrics_update, schema_property_groups_update) are the only new same-family ops that map onto existing CRUD — but our clients use PATCH, so they're omitted for now.

Verification

• pnpm typecheck ✅
• pnpm test ✅ (291 tests; the lone failure is the pre-existing timing flake above, green in isolation)

---

TaskRun: 7bc7929c-c66d-4900-969e-a2e947b4b655

Generated-By: PostHog Code