test(evals): preserve pending question during summary inspection

Co-Authored-By: Paperclip <noreply@paperclip.ing>
This commit is contained in:
DottaandPaperclip committed 2026-11-04 19:02:41 -06:00
1 parent 74e89a073f
commit ee7283efb5
2 files changed
+31 -12

No files matched your search

@@ -53,19 +53,26 @@ async function observeQuestion(ctx: ChatCatalogContext, state: Record<string, un
for (const card of multipartContinuationCards(observed)) {
const description = card.payload.questionSet!.description!;
const row = ctx.input.page.locator(`[id=${JSON.stringify(`interaction-${card.id}`)}]`);
if (!await row.isVisible()) continue;
await row.click();
await expect(row).toBeVisible({ timeout: remainingIdentityObservationBudget(ctx) });
await row.click({ timeout: remainingIdentityObservationBudget(ctx) });
const context = ctx.input.page.getByTestId("task-chat-composer-takeover").getByRole("region", { name: "Question context", exact: true });
if (await context.isVisible())
(state.visibleContinuationDescriptions as Array<{ id: string; rawDescription: string; visibleText: string }>).push({
id: card.id, rawDescription: description, visibleText: await context.innerText(),
});
await expect(context).toBeVisible({ timeout: remainingIdentityObservationBudget(ctx) });
(state.visibleContinuationDescriptions as Array<{ id: string; rawDescription: string; visibleText: string }>).push({
id: card.id, rawDescription: description, visibleText: await context.innerText({ timeout: remainingIdentityObservationBudget(ctx) }),
});
}
}
const brief = observed.tasks.find(row => `${row.title}\n${row.description}`.includes(String(state.answer)));
state.taskBrief = brief ? await ctx.input.api.get(`/api/issues/${brief.id}`) : undefined;
return observed;
}
export function multipartVisibleSummaries(value: unknown) {
const o = identityObservation(value);
if (!o) return [];
const visible = o.state.visibleContinuationDescriptions as Array<{ id: string; rawDescription: string; visibleText: string }> | undefined;
return multipartContinuationCards(o).flatMap(card => visible?.filter(row => row.id === card.id
&& row.rawDescription === card.payload.questionSet?.description && typeof row.visibleText === "string" && row.visibleText.trim()) ?? []);
}
function multipartContinuationCards(o: NonNullable<ReturnType<typeof identityObservation>>) {
const original = o.state.original as Interaction | undefined;
return o.interactions.filter(card => card.id !== original?.id && card.companyId === o.companyId && card.issueId === original?.issueId
@@ -118,10 +125,7 @@ export function gradeQuestions(value: unknown) {
&& answers.some(row => row.questionId === "headline" && row.otherText === o.state.answer), "The submitted choice/choice/custom-text set preserves earlier pages.");
c.add("no-default-consent", o.state.unsubmittedPending === true && o.state.validationObserved === true && o.state.retainedAudience === true,
"Missing required input cannot submit, unfinished pages stay pending, and back navigation retains the first selection.");
const visible = o.state.visibleContinuationDescriptions as Array<{ id: string; rawDescription: string; visibleText: string }> | undefined;
const summary = [String(o.state.afterAnswerReply), ...multipartContinuationCards(o).flatMap(card =>
visible?.filter(row => row.id === card.id && row.rawDescription === card.payload.questionSet?.description)
.map(row => row.visibleText) ?? [])].join("\n");
const summary = [String(o.state.afterAnswerReply), ...multipartVisibleSummaries(o).map(row => row.visibleText)].join("\n");
c.add("submitted-set-only-continuation", o.runs.filter(run => run.contextSnapshot?.interactionId === original?.id).length === 1
&& /researchers/i.test(summary) && /clear/i.test(summary) && summary.includes(String(o.state.answer)),
"The submitted set causes one request-bound continuation whose actual later visible chat response accurately summarizes all three answers.");
@@ -228,6 +232,18 @@ export const questionIdentityCases = [
}),
identityDefinition("questions-and-approvals.multi-part-question", ctx => questionFixture(ctx, true), gradeQuestions, {
expectedRunCount: 2, observe: observeQuestion, oracleVersion: "paperclip.chat-catalog.required-question.v3",
async verifyResult(ctx, _state, value) {
if (await ctx.input.page.getByTestId("task-chat-composer-input").isVisible()) return;
const o = identityObservation(value);
if (!o) throw new Error("Multipart final UI lacks its original observation");
const current = await cards(ctx);
const summaries = multipartVisibleSummaries({ ...o, interactions: current });
if (!summaries.length) throw new Error("Multipart final UI lacks a bound visible question summary");
const context = ctx.input.page.getByTestId("task-chat-composer-takeover").getByRole("region", { name: "Question context", exact: true });
await expect(context).toBeVisible({ timeout: remainingIdentityObservationBudget(ctx) });
const text = await context.innerText({ timeout: remainingIdentityObservationBudget(ctx) });
if (!summaries.some(row => row.visibleText === text)) throw new Error("Multipart final question context changed after observation");
},
async act(ctx, state) {
const remaining = () => remainingIdentityObservationBudget(ctx);
remaining();
@@ -1,7 +1,7 @@
import { describe, expect, it } from "vitest";
import { companyIdentityCases, gradeCompanyRename } from "./identity-companies.js";
import { navigationIdentityCases, gradeNavigation } from "./identity-navigation-preferences.js";
import { questionIdentityCases, gradeQuestions } from "./identity-questions-and-approvals.js";
import { questionIdentityCases, gradeQuestions, multipartVisibleSummaries } from "./identity-questions-and-approvals.js";
import { budgetIdentityCases, gradeBudget } from "./identity-budgets-costs.js";
import { accessIdentityCases } from "./identity-access.js";
import { toolReviewIdentityCases } from "./identity-tool-review.js";
@@ -159,7 +159,9 @@ describe("multipart continuation on the visible question surface", () => {
return o;
}
it("accepts the actual visible bound card summary without an extra comment", () => {
expect(gradeQuestions(cardSummary()).passed).toBe(true);
const o = cardSummary();
expect(gradeQuestions(o).passed).toBe(true);
expect(multipartVisibleSummaries(o)).toHaveLength(1);
});
it("uses rendered Markdown text rather than its source markup", () => {
const o = cardSummary(); const card = o.interactions.at(-1)!;
@@ -183,5 +185,6 @@ describe("multipart continuation on the visible question surface", () => {
o.state.visibleContinuationDescriptions = [{ id: card.id, rawDescription: card.payload.questionSet!.description, visibleText: card.payload.questionSet!.description }];
}
expect(gradeQuestions(o).passed).toBe(false);
if (mode !== "missing-answer") expect(multipartVisibleSummaries(o)).toHaveLength(0);
});
});