test(web): pin every scenario end-state with an aria golden
Every spec now commits at least one golden and the interactive ones one per distinct end-state (nine new .expected.md): - live-interactions: cancel.expected.md (frozen partial + 已停止 marker), error-auth.expected.md (the prompt bubble alone — the committed artifact of the web-error-surface gap, the diff that flips when error rendering lands), retry.expected.md (indistinguishable from a clean completion — retries are deliberately invisible in the transcript). - question-composer: answered.expected.md (the question resolved into its tool round trip plus the final reply, takeover gone) beside the existing waiting-state golden. - steering: mid-steer.expected.md pins the accepted-but-INVISIBLE state (the loop drains steering only at the step boundary, so no interjection bubble exists while the question still blocks — if the client ever renders pending steers eagerly, this golden flips first) and settled.expected.md the badged bubble plus obeying reply. - navigation-panes: waterfall.expected.md and details-open.expected.md (tool-name header, Input args, Output result) beside the trajectory one. - lifecycle-chrome: reloaded.expected.md — rendering the same settled transcript from persistence alone IS the recovery claim. Fixture inventories extended to the new closed sets; the Agent Note's expected-outputs policy updated in both languages (per-end-state goldens for interactive scenarios), pairing re-recorded.
This commit is contained in:
Tianyi Cui
@@ -2,5 +2,5 @@
|
||||
# side as of the last confirmed-consistent state. Both languages carry equal authority;
|
||||
# after editing either side, bring the other along and re-record with:
|
||||
# pnpm run verify-translation-pairing --write
|
||||
2026-07-24-web-gui-browser-e2e-lane.md: 88730cdecf527ece8033ddab1151afcbc6edd83f
|
||||
2026-07-24-web-gui-browser-e2e-lane.zh.md: 9850023a49a860a8f4bbdacc8c48fc389ec77210
|
||||
2026-07-24-web-gui-browser-e2e-lane.md: cc9b1606a62cfbb2322a4c4647d809dfd809b117
|
||||
2026-07-24-web-gui-browser-e2e-lane.zh.md: 3ab0f3716affef6f1446e50d237d74486161afb1
|
||||
|
||||
@@ -32,23 +32,23 @@ Every scenario fails on any pageerror and on the client's connection-loss/gap-re
|
||||
|
||||
### Expected outputs
|
||||
|
||||
One committed golden per scenario: a normalized `ariaSnapshot()` of the conversation region (`ui.expected.md`) — uuid/cwd/workspace-basename/duration tokens normalized, captured poll-until-equal at the settled milestone — plus a few role/text anchor assertions that stay green under a semantics-preserving component rewrite while the golden churns reviewably. The aria tree is the mechanization of the client rule "assert what the user would see, never class names". World-state assertions ride root-context session events inline (which tool call produced which durable result, whether `turn/end` completed) instead of a second committed log golden: the persisted-log surface is pinned by the ACP/headless/TUI suites through the same loop and persistence, and re-pinning it here would double refresh cost against the tier discipline. `refresh` is the sole golden writer — a missing golden in replay mode fails with the healing command rather than self-bootstrapping.
|
||||
At least one committed golden per scenario, and one per DISTINCT end-state for the interactive scenarios (cancel/error/retry, waiting/answered, mid-steer/settled, panel-open, post-reload): a normalized `ariaSnapshot()` of the scenario's owning region — uuid/cwd/workspace-basename/duration tokens normalized, captured poll-until-equal at the settled milestone — plus a few role/text anchor assertions that stay green under a semantics-preserving component rewrite while the golden churns reviewably. The aria tree is the mechanization of the client rule "assert what the user would see, never class names". World-state assertions ride root-context session events inline (which tool call produced which durable result, whether `turn/end` completed) instead of a second committed log golden: the persisted-log surface is pinned by the ACP/headless/TUI suites through the same loop and persistence, and re-pinning it here would double refresh cost against the tier discipline. `refresh` is the sole golden writer — a missing golden in replay mode fails with the healing command rather than self-bootstrapping.
|
||||
|
||||
The typecheck plane split is structural: the three files that boot the host spine (`scaffold`, `replay-round-trip.e2e`, and `seeded-history.e2e`) are excluded from the client-registered `apps/web` project. Those files and their shared `support.ts` are included file-by-file in `tsconfig.host.json` — one program cannot hold both sides of the cordis `Context` merges.
|
||||
|
||||
### Modes and fixtures
|
||||
|
||||
`DSH_SNAPSHOT` selects replay (default, keyless), record (with key), or refresh (keyless) as inline spec branches — the TUI shape, not a suite factory: at two scenarios the acp-snapshot factory machinery has no owner, and the genuinely shared parts are already exported (`scrubRequestHeaders`, `parseSessionLog`, `installLlmReplay`). Each spec splits into drive steps (type, send, `whenTurnSettled` — run in all modes, never waiting on model-content selectors, so record cannot hang on a live model answering differently) and assertion steps (replay/refresh only). Record = drive live through the real composer + harvest the in-memory `session.header`/`session.events` (the TUI `rawSessionLog` shape — no file decompression) + `scrubRequestHeaders` + `{{sessionId}}`/`{{cwd}}`/`{{rpcId}}` tokenization; a follow-up keyless refresh regenerates `ui.expected.md`. Every prompting scenario's fixture was recorded against this assembly through this flow. A drift guard ties each spec's drive prompt to the fixture's recorded `user/message`. A fixture-inventory guard holds each scenario directory closed (exact file set, every JSONL a scrub fixed-point with no run-local `rpcId`). Web fixtures scrub headers everywhere and pin no header class, following the TUI precedent over the strict [pinned-header](2026-07-06-pin-request-header-content-in-one-scenario.md) reading — see Deferred.
|
||||
`DSH_SNAPSHOT` selects replay (default, keyless), record (with key), or refresh (keyless) as inline spec branches — the TUI shape, not a suite factory: at two scenarios the acp-snapshot factory machinery has no owner, and the genuinely shared parts are already exported (`scrubRequestHeaders`, `parseSessionLog`, `installLlmReplay`). Each spec splits into drive steps (type, send, `whenTurnSettled` — run in all modes, never waiting on model-content selectors, so record cannot hang on a live model answering differently) and assertion steps (replay/refresh only). Record = drive live through the real composer + harvest the in-memory `session.header`/`session.events` (the TUI `rawSessionLog` shape — no file decompression) + `scrubRequestHeaders` + `{{sessionId}}`/`{{cwd}}`/`{{rpcId}}` tokenization; a follow-up keyless refresh regenerates the aria goldens. Every prompting scenario's fixture was recorded against this assembly through this flow. A drift guard ties each spec's drive prompt to the fixture's recorded `user/message`. A fixture-inventory guard holds each scenario directory closed (exact file set, every JSONL a scrub fixed-point with no run-local `rpcId`). Web fixtures scrub headers everywhere and pin no header class, following the TUI precedent over the strict [pinned-header](2026-07-06-pin-request-header-content-in-one-scenario.md) reading — see Deferred.
|
||||
|
||||
### Scenarios
|
||||
|
||||
1. **`replay-round-trip`** — new session, prompt through the real composer, replay streams reasoning + a `bash` tool call that really executes in the temp workspace + final text (paced 15ms). Asserts settled markdown, the aria golden, and inline world state (the bash call's durable result is exactly `WEB_E2E_OK\n`, completed `turn/end`, >10 chunk events).
|
||||
2. **`seeded-history`** — a recorded session seeded cold; the sidebar lists it (group row → session row, collapsed by default), opening renders tool cards and text purely from the log through the implicit cold-resume attach inside `session.history` — zero model calls in replay, so no binding constraints; record mode drives the same turn live (real `read` tool against seeded workspace files) to produce the seed.
|
||||
3. **`live-interactions`** — one tool-free recorded turn serves three replay-only scenarios through override sidecars whose CONTENT is authored in the spec and minted as a per-run file in a spec-owned temp dir (the derived success entry for the retry append is re-derived from the fixture via `deriveReplayScript`, never copied into a committed sidecar). Cancel: a `{ patches }` `hang` with a `readyFile` marker — the marker's existence proves the stream is parked mid-turn before the test clicks Stop, making mid-stream cancellation deterministic by construction (`turn/end` reason `aborted`, composer re-enabled). AUTH error: a pre-chunk `throw` outside llm-retry's retryable set (`turn/end` reason `error`, zero `llm/retry` events, composer recovers). SERVER retry: `throw` at call 0 + the fixture's own success appended at 1, proving llm-retry end-to-end in the browser via the durable `llm/retry` record (`request/header` logs only on change, so attempt count is invisible there).
|
||||
4. **`question-composer`** — the shipped composition's resident `ask_user_question` takeover: a recorded turn blocks mid-step on the real userInteraction seam, the composer (`[data-question-key]`) renders in the browser, the test answers through it (the ONE sanctioned place a drive step reacts to model content: the turn cannot complete without the answer, in record and replay alike), and the tool result carries the chosen label. Golden: the composer's stable waiting state.
|
||||
5. **`steering`** — mid-turn steer while the question composer blocks the step (the deterministic mid-turn window; no timing dependence). The composer locks while running, so the steer POSTs `session.prompt` `mode:'steer'` from the page over the same same-origin `/api` wire the client uses (`TODO(web-steer-composer)`: drive a composer gesture once one exists); everything downstream is product — gateway → `Agent.steer` → step-boundary drain → durable `steering/message` → SSE → badged interjection bubble. Record-mode fixture honesty: the recording is rejected unless the live model's final reply obeys an instruction only the steering message carries.
|
||||
6. **`navigation-panes`** — one rich two-turn seed (turn 1: bash + two parallel reads in one assistant message; turn 2: a markdown-heavy reply) rendered cold through the seeded-history pattern (zero model calls), serving four surfaces: sidebar search (client-side title filter — asserted only after the durable title lands with the attach baseline, because a cold `SessionSummary` carries no title and search matches the `displayTitle` the user sees; negative query empties the tree, positive narrows, clear restores), the Trajectory tab (turn sections + the step group's tool mix plus the view-area aria golden), the Waterfall tab (span stats + one lane per span — the P-I fold counts a turn-0 prologue span because only assistant/steering nodes carry a turn number, pinned as-is), and the details column (the bash toolview row routes click to openDetails; open/closed is asserted on the frame's `data-details-collapsed` attribute because close collapses the grid column to width 0 without unmounting the subtree).
|
||||
7. **`lifecycle-chrome`** — one tiny recorded text turn drives three whole-page concerns. Workspace flow over the real wire: the empty-state hero's first send materializes a real Workspace + Session (the jsdom `workspace-flow.snapshot.ts` suite pins this state machine over the fixture client; this scenario pins it through HTTP RPC + SSE + the gateway), proven durably by the session header's cwd being the create-by-name target `<workspaceRoot>/workspace`, plus the hero waiting-state aria golden. Reload recovery: collapse the sidebar (persisted `dsh.layout.panels`), `page.reload`, and the surface comes back whole from persistence alone — layout collapsed, selection restored (`dsh.sessions.current`), the recorded turn re-rendered from `session.history` with zero model calls (the drained replay cursor makes any stray request fail loud at close). Dark mode: no product control flips the theme yet, so the scenario drives the ThemeService's entire DOM contract — the `body[data-ds-dark-theme]` attribute — and pins the shipped cascade: the alias token flips, a painted surface repaints, and removing the attribute restores the light sample exactly (`TODO(web-theme-gesture)` upgrades to a real settings control); per the scope ruling there is no theme/layout golden (aria is color-blind).
|
||||
3. **`live-interactions`** — one tool-free recorded turn serves three replay-only scenarios through override sidecars whose CONTENT is authored in the spec and minted as a per-run file in a spec-owned temp dir (the derived success entry for the retry append is re-derived from the fixture via `deriveReplayScript`, never copied into a committed sidecar). Cancel: a `{ patches }` `hang` with a `readyFile` marker — the marker's existence proves the stream is parked mid-turn before the test clicks Stop, making mid-stream cancellation deterministic by construction (`turn/end` reason `aborted`, composer re-enabled). AUTH error: a pre-chunk `throw` outside llm-retry's retryable set (`turn/end` reason `error`, zero `llm/retry` events, composer recovers). SERVER retry: `throw` at call 0 + the fixture's own success appended at 1, proving llm-retry end-to-end in the browser via the durable `llm/retry` record (`request/header` logs only on change, so attempt count is invisible there). Each scenario pins its terminal surface as a golden: `cancel.expected.md` (frozen `partial`, 已停止 marker), `error-auth.expected.md` (the prompt bubble alone — the committed artifact of the web-error-surface gap, the diff that flips when error rendering lands), `retry.expected.md` (indistinguishable from a clean completion — retries are deliberately invisible in the transcript).
|
||||
4. **`question-composer`** — the shipped composition's resident `ask_user_question` takeover: a recorded turn blocks mid-step on the real userInteraction seam, the composer (`[data-question-key]`) renders in the browser, the test answers through it (the ONE sanctioned place a drive step reacts to model content: the turn cannot complete without the answer, in record and replay alike), and the tool result carries the chosen label. Goldens: the composer's stable waiting state (`ui.expected.md`) and the answered transcript (`answered.expected.md` — the question resolved into its tool round trip plus the final reply, takeover gone).
|
||||
5. **`steering`** — mid-turn steer while the question composer blocks the step (the deterministic mid-turn window; no timing dependence). The composer locks while running, so the steer POSTs `session.prompt` `mode:'steer'` from the page over the same same-origin `/api` wire the client uses (`TODO(web-steer-composer)`: drive a composer gesture once one exists); everything downstream is product — gateway → `Agent.steer` → step-boundary drain → durable `steering/message` → SSE → badged interjection bubble. Record-mode fixture honesty: the recording is rejected unless the live model's final reply obeys an instruction only the steering message carries. Goldens pin the timing semantics visually: `mid-steer.expected.md` captures the accepted-but-invisible state (the loop drains steering only at the step boundary, so no interjection bubble exists while the question still blocks — if the client ever renders pending steers eagerly, this golden flips first) and `settled.expected.md` the badged bubble plus obeying reply.
|
||||
6. **`navigation-panes`** — one rich two-turn seed (turn 1: bash + two parallel reads in one assistant message; turn 2: a markdown-heavy reply) rendered cold through the seeded-history pattern (zero model calls), serving four surfaces: sidebar search (client-side title filter — asserted only after the durable title lands with the attach baseline, because a cold `SessionSummary` carries no title and search matches the `displayTitle` the user sees; negative query empties the tree, positive narrows, clear restores), the Trajectory tab (turn sections + the step group's tool mix plus the view-area aria golden), the Waterfall tab (span stats + one lane per span — the P-I fold counts a turn-0 prologue span because only assistant/steering nodes carry a turn number, pinned as-is), and the details column (the bash toolview row routes click to openDetails; open/closed is asserted on the frame's `data-details-collapsed` attribute because close collapses the grid column to width 0 without unmounting the subtree). Goldens: `trajectory.expected.md` and `waterfall.expected.md` (each tab's view area) plus `details-open.expected.md` (the open panel: tool-name header, Input args, Output result).
|
||||
7. **`lifecycle-chrome`** — one tiny recorded text turn drives three whole-page concerns. Workspace flow over the real wire: the empty-state hero's first send materializes a real Workspace + Session (the jsdom `workspace-flow.snapshot.ts` suite pins this state machine over the fixture client; this scenario pins it through HTTP RPC + SSE + the gateway), proven durably by the session header's cwd being the create-by-name target `<workspaceRoot>/workspace`, plus the hero waiting-state aria golden. Reload recovery: collapse the sidebar (persisted `dsh.layout.panels`), `page.reload`, and the surface comes back whole from persistence alone — layout collapsed, selection restored (`dsh.sessions.current`), the recorded turn re-rendered from `session.history` with zero model calls (the drained replay cursor makes any stray request fail loud at close), and `reloaded.expected.md` pins the rebuilt conversation region — rendering the same settled transcript from persistence alone IS the recovery claim. Dark mode: no product control flips the theme yet, so the scenario drives the ThemeService's entire DOM contract — the `body[data-ds-dark-theme]` attribute — and pins the shipped cascade: the alias token flips, a painted surface repaints, and removing the attribute restores the light sample exactly (`TODO(web-theme-gesture)` upgrades to a real settings control); per the scope ruling there is no theme/layout golden (aria is color-blind).
|
||||
|
||||
### CI stance
|
||||
|
||||
|
||||
@@ -32,23 +32,23 @@ Web GUI 以一条真实组装链交付——chromium 页面 → client 插件 bu
|
||||
|
||||
### 预期输出
|
||||
|
||||
每场景一份提交的预期输出：会话区规范化 `ariaSnapshot()`（`ui.expected.md`）——uuid/cwd/工作区目录名/时长归一为稳定 token，在安定里程碑处轮询至两次相等再采集——外加几条 role/文本锚断言，让保语义的组件重写在预期输出可评审地变动时仍保持绿色锚点。aria 树是 client 规则「断言用户所见，绝不断言类名」的机械化。世界状态断言内联在根上下文的会话事件上（哪次工具调用产生了哪项已持久化的工具结果、`turn/end` 是否完成）而不是第二份提交的日志预期输出：持久化日志表面已由 ACP/headless/TUI 套件经同一循环和持久化钉住，在此重复钉住会违背分层纪律、翻倍刷新成本。`refresh` 是预期输出的唯一写入者——回放模式下预期输出缺失会连同修复命令一起报错，而不是静默自举。
|
||||
每场景至少一份提交的预期输出，交互类场景则每个不同终态各一份（取消/错误/重试、等待/已作答、steer 中途/安定、面板打开、重新加载后）：该场景所属区域的规范化 `ariaSnapshot()`——uuid/cwd/工作区目录名/时长归一为稳定 token，在安定里程碑处轮询至两次相等再采集——外加几条 role/文本锚断言，让保语义的组件重写在预期输出可评审地变动时仍保持绿色锚点。aria 树是 client 规则「断言用户所见，绝不断言类名」的机械化。世界状态断言内联在根上下文的会话事件上（哪次工具调用产生了哪项已持久化的工具结果、`turn/end` 是否完成）而不是第二份提交的日志预期输出：持久化日志表面已由 ACP/headless/TUI 套件经同一循环和持久化钉住，在此重复钉住会违背分层纪律、翻倍刷新成本。`refresh` 是预期输出的唯一写入者——回放模式下预期输出缺失会连同修复命令一起报错，而不是静默自举。
|
||||
|
||||
类型检查平面切分是结构性的：启动 host 主干的三个文件（`scaffold`、`replay-round-trip.e2e` 和 `seeded-history.e2e`）被排除出注册在 client 侧的 `apps/web` 工程。这三个文件及其共享的 `support.ts` 逐文件纳入 `tsconfig.host.json`——一个程序不能同时持有 cordis `Context` 合并的两侧。
|
||||
|
||||
### 模式与 fixture
|
||||
|
||||
`DSH_SNAPSHOT` 以内联 spec 分支选择 replay（默认，无密钥）、record（带密钥）或 refresh（无密钥）——TUI 的形态，不是套件工厂：两个场景撑不起 acp-snapshot 工厂机制，且真正共享的部分已被导出（`scrubRequestHeaders`、`parseSessionLog`、`installLlmReplay`）。每个 spec 切分为驱动步骤（输入、发送、`whenTurnSettled`——所有模式都执行，绝不等待模型内容选择器，因此 record 不会因真实模型答法不同而挂起）与断言步骤（仅 replay/refresh）。Record = 经真实输入框实时驱动 + 采收内存中的 `session.header`/`session.events`（TUI 的 `rawSessionLog` 形态——无需文件解压）+ `scrubRequestHeaders` + `{{sessionId}}`/`{{cwd}}`/`{{rpcId}}` token 化；随后一次无密钥 refresh 重新生成 `ui.expected.md`。每个发起提示的场景，其 fixture 都经此流程对本组装录制。一条漂移防线把每个 spec 的驱动提示词与 fixture 录制的 `user/message` 绑定。fixture 清单防线保持每个场景目录封闭（精确文件集合，每个 JSONL 都是脱敏不动点，不含当次运行的 `rpcId`）。Web fixture 全部脱敏请求头且不钉任何头类别，沿用 TUI 先例而非[钉住请求头](2026-07-06-pin-request-header-content-in-one-scenario.md)的严格读法——见「暂缓」。
|
||||
`DSH_SNAPSHOT` 以内联 spec 分支选择 replay（默认，无密钥）、record（带密钥）或 refresh（无密钥）——TUI 的形态，不是套件工厂：两个场景撑不起 acp-snapshot 工厂机制，且真正共享的部分已被导出（`scrubRequestHeaders`、`parseSessionLog`、`installLlmReplay`）。每个 spec 切分为驱动步骤（输入、发送、`whenTurnSettled`——所有模式都执行，绝不等待模型内容选择器，因此 record 不会因真实模型答法不同而挂起）与断言步骤（仅 replay/refresh）。Record = 经真实输入框实时驱动 + 采收内存中的 `session.header`/`session.events`（TUI 的 `rawSessionLog` 形态——无需文件解压）+ `scrubRequestHeaders` + `{{sessionId}}`/`{{cwd}}`/`{{rpcId}}` token 化；随后一次无密钥 refresh 重新生成各份 aria 预期输出。每个发起提示的场景，其 fixture 都经此流程对本组装录制。一条漂移防线把每个 spec 的驱动提示词与 fixture 录制的 `user/message` 绑定。fixture 清单防线保持每个场景目录封闭（精确文件集合，每个 JSONL 都是脱敏不动点，不含当次运行的 `rpcId`）。Web fixture 全部脱敏请求头且不钉任何头类别，沿用 TUI 先例而非[钉住请求头](2026-07-06-pin-request-header-content-in-one-scenario.md)的严格读法——见「暂缓」。
|
||||
|
||||
### 场景
|
||||
|
||||
1. **`replay-round-trip`**——新会话，经真实输入框发送提示词，回放流式输出推理（reasoning）+ 一次在临时工作区真实执行的 `bash` 工具调用 + 最终文本（15ms 节奏）。断言安定后的 markdown、aria 预期输出与内联世界状态（这次 bash 调用的已持久化工具结果严格等于 `WEB_E2E_OK\n`、完成的 `turn/end`、>10 个分片事件）。
|
||||
2. **`seeded-history`**——冷播种一份已录会话；侧栏列出它（分组行 → 会话行，默认折叠），打开后纯凭日志经 `session.history` 内的隐式冷恢复挂载渲染工具卡片与文本——replay 下零模型调用，因此没有任何绑定约束；record 模式实时驱动同一轮（真实 `read` 工具读取播种的工作区文件）来产出种子。
|
||||
3. **`live-interactions`**——一段不含工具调用的已录轮次经覆写 sidecar 承载三个仅回放的场景：sidecar 的内容本身写在 spec 里，每次运行时在 spec 自有的临时目录中生成文件（重试追加所用的派生成功条目经 `deriveReplayScript` 从 fixture 重新派生，绝不复制进已提交的 sidecar）。取消：一个带 `readyFile` 标记的 `{ patches }` `hang`——标记文件的存在证明流在测试点击 Stop 之前已停驻在轮次中途，使流中取消按构造即确定（`turn/end` 原因为 `aborted`，输入框重新启用）。AUTH 错误：一次落在 llm-retry 可重试集合之外的分片前 `throw`（`turn/end` 原因为 `error`，零条 `llm/retry` 事件，输入框恢复可用）。SERVER 重试：第 0 次调用 `throw` + 在第 1 次调用处追加 fixture 自身的成功条目，凭持久的 `llm/retry` 记录在浏览器中端到端证明 llm-retry（`request/header` 仅在变化时记录，因此尝试次数在那里不可见）。
|
||||
4. **`question-composer`**——已交付组合中常驻的 `ask_user_question` 接管：一段已录轮次在真实的 userInteraction seam 上阻塞于步骤中途，提问输入框（`[data-question-key]`）在浏览器中渲染，测试经它作答（这是驱动步骤对模型内容作出反应的唯一获准之处：没有这个回答，轮次无法完成，record 与 replay 皆然），工具结果携带所选的 label。预期输出：提问输入框稳定的等待态。
|
||||
5. **`steering`**——在提问输入框阻塞该步骤时做轮次中途 steering（中途引导），此即确定性的轮次中途窗口，不依赖任何时序。输入框在运行期间锁定，因此这一 steer 由页面经客户端所用的同一条同源 `/api` wire POST `session.prompt` `mode:'steer'`（`TODO(web-steer-composer)`：待有输入框手势后改为驱动它）；下游的一切都是产品路径——gateway → `Agent.steer` → 步骤边界排空 → 持久的 `steering/message` → SSE → 带徽标的插话气泡。record 模式的 fixture 诚实性：除非真实模型的最终回复遵循了一条只有 steering 消息才携带的指令，否则该次录制被拒绝。
|
||||
6. **`navigation-panes`**——一份内容丰富的双轮次种子（轮次 1：同一条 assistant 消息内的 bash + 两次并行 read；轮次 2：一段 markdown 密集的回复）经 seeded-history 模式冷渲染（零模型调用），承载四个表面：侧栏搜索（客户端标题过滤；仅在持久的标题随 attach 基线一同到达后才断言，因为冷的 `SessionSummary` 不携带标题，而搜索匹配的是用户所见的 `displayTitle`；反例查询清空整棵树，正例查询收窄，清除后复原）、Trajectory 标签页（轮次分节 + 步骤组的工具构成，外加视图区 aria 预期输出）、Waterfall 标签页（span 统计 + 每个 span 一条泳道；只有 assistant/steering 节点携带轮次编号，因此 P-I 折叠会将一个轮次 0 的序幕 span 计入，按原样钉住）与详情列（bash 工具视图行把点击路由到 openDetails；打开/关闭状态断言在 frame 的 `data-details-collapsed` 属性上，因为关闭把网格列收缩到宽度 0 而不卸载子树）。
|
||||
7. **`lifecycle-chrome`**——一段极小的已录纯文本轮次驱动三个整页关注点。真实 wire 上的 Workspace 动线：空态 hero 的首次发送物化出真实的 Workspace + Session（jsdom 的 `workspace-flow.snapshot.ts` 套件基于 fixture 客户端钉住这一状态机；本场景则经 HTTP RPC + SSE + gateway 钉住它），其持久证据是会话头部的 cwd 恰为按名创建的目标 `<workspaceRoot>/workspace`，外加 hero 等待态的 aria 预期输出。重新加载恢复：折叠侧栏（持久化于 `dsh.layout.panels`），`page.reload`，整个表面纯凭持久化完整归来——布局保持折叠，选中项恢复（`dsh.sessions.current`），已录轮次从 `session.history` 重新渲染且零模型调用（已耗尽的回放游标使任何离群请求都在 close 时大声失败）。暗色模式：产品尚无任何控件能切换主题，因此本场景驱动 ThemeService 的整个 DOM 契约（即 `body[data-ds-dark-theme]` 属性），并钉住已交付的级联：alias token 翻转，某个实际绘制的表面重绘，移除该属性则精确还原亮色采样值（`TODO(web-theme-gesture)`：待有真实设置控件后升级为驱动它）；按范围裁定，主题/布局不设预期输出（aria 感知不到颜色）。
|
||||
3. **`live-interactions`**——一段不含工具调用的已录轮次经覆写 sidecar 承载三个仅回放的场景：sidecar 的内容本身写在 spec 里，每次运行时在 spec 自有的临时目录中生成文件（重试追加所用的派生成功条目经 `deriveReplayScript` 从 fixture 重新派生，绝不复制进已提交的 sidecar）。取消：一个带 `readyFile` 标记的 `{ patches }` `hang`——标记文件的存在证明流在测试点击 Stop 之前已停驻在轮次中途，使流中取消按构造即确定（`turn/end` 原因为 `aborted`，输入框重新启用）。AUTH 错误：一次落在 llm-retry 可重试集合之外的分片前 `throw`（`turn/end` 原因为 `error`，零条 `llm/retry` 事件，输入框恢复可用）。SERVER 重试：第 0 次调用 `throw` + 在第 1 次调用处追加 fixture 自身的成功条目，凭持久的 `llm/retry` 记录在浏览器中端到端证明 llm-retry（`request/header` 仅在变化时记录，因此尝试次数在那里不可见）。每个场景都把各自的终态表面钉为一份预期输出：`cancel.expected.md`（冻结的 `partial`、「已停止」标记）、`error-auth.expected.md`（仅有提示词气泡——web-error-surface 缺口的已提交产物，错误渲染落地时翻转的那份 diff）、`retry.expected.md`（与一次干净完成无从区分——重试在文本记录中刻意不可见）。
|
||||
4. **`question-composer`**——已交付组合中常驻的 `ask_user_question` 接管：一段已录轮次在真实的 userInteraction seam 上阻塞于步骤中途，提问输入框（`[data-question-key]`）在浏览器中渲染，测试经它作答（这是驱动步骤对模型内容作出反应的唯一获准之处：没有这个回答，轮次无法完成，record 与 replay 皆然），工具结果携带所选的 label。预期输出：提问输入框稳定的等待态（`ui.expected.md`）与已作答的文本记录（`answered.expected.md`——提问已落定为其工具往返加最终回复，接管消失）。
|
||||
5. **`steering`**——在提问输入框阻塞该步骤时做轮次中途 steering（中途引导），此即确定性的轮次中途窗口，不依赖任何时序。输入框在运行期间锁定，因此这一 steer 由页面经客户端所用的同一条同源 `/api` wire POST `session.prompt` `mode:'steer'`（`TODO(web-steer-composer)`：待有输入框手势后改为驱动它）；下游的一切都是产品路径——gateway → `Agent.steer` → 步骤边界排空 → 持久的 `steering/message` → SSE → 带徽标的插话气泡。record 模式的 fixture 诚实性：除非真实模型的最终回复遵循了一条只有 steering 消息才携带的指令，否则该次录制被拒绝。预期输出以可视方式钉住这一时序语义：`mid-steer.expected.md` 捕捉「已接受但不可见」的状态（循环仅在步骤边界才排空 steering，因此提问仍在阻塞时不存在插话气泡——若 client 日后提前渲染待处理的 steer，这份预期输出会最先翻转），`settled.expected.md` 则捕捉带徽标的气泡加遵循指令的回复。
|
||||
6. **`navigation-panes`**——一份内容丰富的双轮次种子（轮次 1：同一条 assistant 消息内的 bash + 两次并行 read；轮次 2：一段 markdown 密集的回复）经 seeded-history 模式冷渲染（零模型调用），承载四个表面：侧栏搜索（客户端标题过滤；仅在持久的标题随 attach 基线一同到达后才断言，因为冷的 `SessionSummary` 不携带标题，而搜索匹配的是用户所见的 `displayTitle`；反例查询清空整棵树，正例查询收窄，清除后复原）、Trajectory 标签页（轮次分节 + 步骤组的工具构成，外加视图区 aria 预期输出）、Waterfall 标签页（span 统计 + 每个 span 一条泳道；只有 assistant/steering 节点携带轮次编号，因此 P-I 折叠会将一个轮次 0 的序幕 span 计入，按原样钉住）与详情列（bash 工具视图行把点击路由到 openDetails；打开/关闭状态断言在 frame 的 `data-details-collapsed` 属性上，因为关闭把网格列收缩到宽度 0 而不卸载子树）。预期输出：`trajectory.expected.md` 与 `waterfall.expected.md`（各自标签页的视图区），外加 `details-open.expected.md`（打开的面板：工具名标题、Input 参数、Output 结果）。
|
||||
7. **`lifecycle-chrome`**——一段极小的已录纯文本轮次驱动三个整页关注点。真实 wire 上的 Workspace 动线：空态 hero 的首次发送物化出真实的 Workspace + Session（jsdom 的 `workspace-flow.snapshot.ts` 套件基于 fixture 客户端钉住这一状态机；本场景则经 HTTP RPC + SSE + gateway 钉住它），其持久证据是会话头部的 cwd 恰为按名创建的目标 `<workspaceRoot>/workspace`，外加 hero 等待态的 aria 预期输出。重新加载恢复：折叠侧栏（持久化于 `dsh.layout.panels`），`page.reload`，整个表面纯凭持久化完整归来——布局保持折叠，选中项恢复（`dsh.sessions.current`），已录轮次从 `session.history` 重新渲染且零模型调用（已耗尽的回放游标使任何离群请求都在 close 时大声失败），且 `reloaded.expected.md` 钉住重建后的会话区——纯凭持久化渲染出同一份安定的文本记录，这本身就是恢复主张。暗色模式：产品尚无任何控件能切换主题，因此本场景驱动 ThemeService 的整个 DOM 契约（即 `body[data-ds-dark-theme]` 属性），并钉住已交付的级联：alias token 翻转，某个实际绘制的表面重绘，移除该属性则精确还原亮色采样值（`TODO(web-theme-gesture)`：待有真实设置控件后升级为驱动它）；按范围裁定，主题/布局不设预期输出（aria 感知不到颜色）。
|
||||
|
||||
### CI 立场
|
||||
|
||||
|
||||
Reference in New Issue
Block a user