Tracking issue for the v0.10.0 release, mirroring the #437 pattern.
main is at ad8362db. Provenance pins bc3 dffa7e11b3 (2026-07-28); API_VERSION is 2026-07-28; SDK version is still 0.9.0.
Why this one is heavy
v0.10.0 is a substantial breaking release — eight breaking PRs, five of them landed in this cycle:
PR
Break
#393 Message types become bucket-scoped
#461 Total-attempt retry semantics (observed 4 → 3)
#469 Node floor raised to >=22.12.0
#481 Go *WithBody request bodies replayed across retries
#482 Go maps HTTP 400 to validation, not api_error
#483 Per-operation retry.max applied as a ceiling in Go and Python
#486 Declared retry_on honored — 500/502/504 are no longer retried
#487 SearchResult.content/description become required + nullable (Go *string, Swift init position moves)
#488 bubble_up_url required on Todolist/TodolistGroup
#489 cards.update → updateVerbatim; the plain update is now merge-safe and costs a GET
The three most likely to bite a consumer silently:
Checklist
SearchResult.content/description are always null on the wire; plain_text_content/plain_text_description are unmodeled #465 Cards.update silently erases due_on on every sparse update (BC3 forced-replace) #467 ScheduleEntry update omitting participant_ids silently removes all participants #477 breaking label audited across all release PRs — Absorb bubble_up_url across the four shapes that carry it #488 Cards: give the plain update name to a merge-safe composite #489 documentation label on Registry/doc nits: durable smithy_refs, count-agnostic idempotency comments, qualified schedule rule #457 Correct the SearchResult generic-attachments carve-out (closes #428) #464 run 30407217650 )make bump VERSION=0.10.0, 10 version files + 4 lockfiles, never inside a feature PR ruby/Gemfile.lock and typescript/package-lock.json inspected by hand (make release validates 7 of 10 version files and neither lockfile)make release VERSION=0.10.0 from a fresh, clean, current mainv0.10.0 and go/v0.10.0 both present (the command pushes the first; the Go workflow creates the second)release-github.yml hard-fails past 30 minutes)
Not in this release
Tracking issue for the v0.10.0 release, mirroring the #437 pattern.
mainis atad8362db. Provenance pins bc3dffa7e11b3(2026-07-28);API_VERSIONis2026-07-28; SDK version is still0.9.0.Why this one is heavy
v0.10.0 is a substantial breaking release — eight
breakingPRs, five of them landed in this cycle:>=22.12.0*WithBodyrequest bodies replayed across retriesapi_errorretry.maxapplied as a ceiling in Go and Pythonretry_onhonored — 500/502/504 are no longer retriedSearchResult.content/descriptionbecome required + nullable (Go*string, Swift init position moves)bubble_up_urlrequired onTodolist/TodolistGroupcards.update→updateVerbatim; the plainupdateis now merge-safe and costs a GETThe three most likely to bite a consumer silently:
cards.updatekeeps working and gets safer, but now makes two requests; callers wanting the single PUT must move toupdateVerbatim.Content string→*stringbreaks every dereference; Swift's memberwise init parameters move into the required position.Checklist
breakinglabel audited across all release PRs — Absorb bubble_up_url across the four shapes that carry it #488 and Cards: give the plain update name to a merge-safe composite #489 were missing it, which would have filed them under Features and Bug Fixes respectivelydocumentationlabel on Registry/doc nits: durable smithy_refs, count-agnostic idempotency comments, qualified schedule rule #457 / Correct the SearchResult generic-attachments carve-out (closes #428) #464make bump VERSION=0.10.0, 10 version files + 4 lockfiles, never inside a feature PRruby/Gemfile.lockandtypescript/package-lock.jsoninspected by hand (make releasevalidates 7 of 10 version files and neither lockfile)make release VERSION=0.10.0from a fresh, clean, currentmainv0.10.0andgo/v0.10.0both present (the command pushes the first; the Go workflow creates the second)release-github.ymlhard-fails past 30 minutes)Not in this release
SubscribeandEnableCardColumnOnHoldunder-classified as non-idempotent. Follow-up.b238a074is post-sync drift, outside thedffa7e11run-boundary pin; belongs to the next sync.