ADR-0006 leftovers in `@objectstack/client`: `client.project()` / `ScopedProjectClient` / two test titles still speak the pre-v5.0 noun

[hotlong]

Found while implementing #12866 (ADR-0006 D2, SDK half). Filed unassigned. Related: #12865 (the D2 epic — this is NOT in its declared scope, which is the projects METHOD namespace, its unwrap keys and its JSDoc).

What is left

#12866 renames client.projects.* to client.environments.* and moves the project/projects response keys. It deliberately does not touch the SECOND project surface on the same class — the environment-scoped sub-client — because that is a different shape (a factory method plus an exported class) and a different decision. After #12866 lands, packages/client/src/index.ts still carries:

site	what it says today
ObjectStackClient.project(environmentId)	the scoped-client factory; its own parameter is already called environmentId
export class ScopedProjectClient	published class name
its docblock: "Project-scoped sub-client … a single client instance can talk to multiple projects"	the entity noun
the environmentId client-option docblock: "Active project id (UUID of sys_environment)"	the entity noun, on an option already named environmentId
projectResolution (named in that same docblock, pointing at packages/rest/src/rest-server.ts)	needs its own check — this card does not assert it is still the live spelling

AGENTS.md's standing line is "Project now only means the npm/monorepo sense", and ADR-0006's v5.0 rename is aliasless, so every row above is a rename the ADR already decided; what it does not have is a card.

A measurable half that is cheap and independent

packages/client/src/client.test.ts lines 1989 and 1999 are two ScopedProjectClient test TITLES that read

it('prefixes meta.getTypes with /projects/:id', …)
it('prefixes data.find with /projects/:id', …)

while the assertions immediately below them pin http://localhost:3000/api/v1/environments/proj-123/…. The titles describe a URL the code has not built since the v5.0 path rename; the tests pass, so nothing says so. That half is a two-line fix with no API consequence and could be split out.

Scope note

Renaming the exported ScopedProjectClient class is a published-API break on @objectstack/client and should ride a coordinated window the same way #12866 does, not a drive-by. The docblock prose and the two test titles are not breaks and could land at any time.

Generated by Claude Code

[hotlong]

决策请求:是否并入 ADR-0006 D2 同一个 coordinated major?(epic #12865)

D2 执行中测得本卡描述的第四面(client.project(id) / ScopedProjectClient / 相关散文)仍是 project 拼写 —— D1 addendum 的三面普查没有点名它,属漏网而非裁定保留。两个 rename PR(#12885 / cloud#1692)已 ACCEPT 并 draft 停放等协调窗口,现在决定它是否搭同一班 major。

四维分析

• 实际业务需求:SDK 调用方在同一个 major 里被 break 一次,好过相邻两个 major 各 break 一次 —— 分两次意味着同一批用户写两轮迁移;现在 caller base 最小(裁决原话:窗口随采用率贬值)。
• 项目长远合理性:ADR-0006 的方向是 environment 一个词汇;留一个 project 拼写的导出公开类是永久双词汇缝,和 D3 拒绝映射层的理由同构。
• 防 AI 写代码犯错:两套拼写并存正是 AI 消费端最易犯错的形状 —— 猜错命名空间能通过类型检查;一次收敛消除整个猜测面。
• 创业阶段不扩散需求:零新增能力,纯收敛;增量 diff 小(一个类 + 一个 getter + 散文),搭已开的窗成本最低,窗口外单独再破一次反而最贵。

推荐:并入 —— 维护者一句同意,本卡即挂入 epic #12865 子树、按 SDK 侧同流程派发(draft 停放同一窗口)。不同意则本卡按独立 backlog 留给下一次 major。

[hotlong]

Maintainer ruling (2026-08-28, live PM chat) — FOLD APPROVED

Verbatim, from the batched decision reply: 「1 同意 2 不挂 major 3 可以 4 推」 — item 3 (「可以」) approves this card's recommendation: the fourth project-spelled surface folds into the ADR-0006 D2 coordinated major (epic #12865).

Execution: family-fold into the SDK-half PR #12885 (same package, same rename family, now same ruling). This card gets its own commits and its own Fixes #12882 line on that PR; the shared branch is claude/issue-12866-adr0006-d2-sdk-environments. Item 2 of the same ruling (「不挂 major」) applies to the shared changeset: it STAYS minor with BREAKING described in prose — do not flip it.

[hotlong]

Claim: PM loop round 4 (epic #12865; family-fold member, maintainer ruling of 2026-08-28 recorded above)
Session: session_65d2faee-1ff4-4be8-be28-72b972c539d9
Branch: claude/issue-12866-adr0006-d2-sdk-environments (fold chain head — this member card carries no branch of its own by design)
Worktree: continues the #12866 dev's worktree
Domain: domain:cli family (epic territory)
File surface: packages/client/src/index.ts (ScopedProjectClient class, the client.project getter, environmentId option prose) + client tests (stop on breach; explain in the report)
Container & model: M, mode:subagent, model: opus (fold rider on the accepted #12866 implementation)
Clause-②: no
(Same rationale as the chain head: implements the maintainer-ruled rename family; spec untouched.)
Serial constraints cleared: trivially — the fold IS the serial answer for this file (one dev owns packages/client/src/index.ts inside the epic; five fold gates checked: same fix shape, same package, member ruled 2026-08-28, independently verifiable via its own commits + Fixes line, exclusion list = the three D1 surfaces already renamed on the chain head).

[hotlong]

Claim: dev seat implementing this card as a fold rider on the ADR-0006 D2 SDK half.

• Session: session_65d2faee-1ff4-4be8-be28-72b972c539d9 / subagent agent-a1d45b69a14c2302b (os-dev; same seat that filed this card while implementing ADR-0006 D2 (SDK half): rename client.projects.* to client.environments.*; unwrap keys follow the wire; JSDoc tells the truth #12866)
• Branch: claude/issue-12866-adr0006-d2-sdk-environments — the fold-chain head, shared with ADR-0006 D2 (SDK half): rename client.projects.* to client.environments.*; unwrap keys follow the wire; JSDoc tells the truth #12866 by the 2026-08-28 ruling. This card carries no branch of its own by design; its work is the single commit e1c8d20ea.
• Worktree: objectstack-issue-12882, recreated from the pushed chain head 92db06fa
• PR: ADR-0006 D2 (SDK half): rename client.projects.* to client.environments.*, unwrap keys follow the wire, JSDoc names the endpoint (#12866) #12885 (DRAFT), which now carries Fixes #12882 on its own line
• File surface: packages/client/src/index.ts (the exported scoped-client class, the scoping getter, the environmentId option prose), packages/client tests, scripts/check-exported-any-returns.mts (one prose reference to the renamed class), and two hand-written docs that asserted the old getter survives as a compatibility name

Governance check before writing code, since this card renames something an accepted ADR discusses: ADR-0006's D1 addendum (merged to main as ac35e2a2a after the chain head was cut) names THREE surfaces as deliberately retained — the SDK method namespace, the control-plane response fields, and the provisioning JSDoc sentence. All three are D2's to rename and are already done on the chain head. This card's surface is not among them; the ruling comment above records it as an oversight in that census rather than a retention, so Prime Directive 13 is not engaged.

Landing constraint unchanged and inherited from the chain head: DRAFT only, no un-drafting, no auto-merge. Lands with the producer half in the maintainer-coordinated window.

Generated by Claude Code