Files
Jyotisha/frontend/tests/rectification-answer-choice.test.ts
T
Jesse_Chen 7a4360d848
Independent Staging Quality Gate / validate (push) Failing after 11m44s
Independent Staging Quality Gate / publish (push) Has been skipped
fix(web): keep rectification stages aligned and stop jump overlay covering choices
Progress and spoken text shared a 24px consultation-report gap, and the jump chip covered option D. Put D-chart names back on the activity strip from evidence rescore methods.

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-08-25 19:31:14 +08:00

319 lines
12 KiB
TypeScript

import assert from "node:assert/strict";
import { readFileSync } from "node:fs";
import test from "node:test";
import { applyRectificationChoice } from "../src/lib/rectification-agentic/v9/answer-choice.ts";
import {
CHOICE_ACTION,
STOP_ACTION,
composeChoiceNarration,
quoteIsFromAssistantQuestion,
userVisibleChoiceLine,
} from "../src/lib/rectification-agentic/v9/choice-action.ts";
import { isNearBottom, shouldFollowLatest, shouldShowJumpToLatest } from "../src/lib/rectification-sticky-scroll.ts";
import { projectTurnDecision, TURN_DECISION_MAX_BYTES, turnDecisionByteLength } from "../src/lib/rectification-agentic/v9/turn-decision.ts";
import {
mapModelFinishToErrorCode,
userFacingRunFailure,
isIncompleteRunBanner,
} from "../src/lib/rectification-agentic/v9/run-diagnostic.ts";
import { buildInferenceState } from "../src/lib/rectification-agentic/core/build-state.ts";
import { parseV9CaseDossier } from "../src/lib/rectification-agentic/v9/tool-service.ts";
import {
CASE_ID,
FOCUS_ID,
SESSION_ID,
TURN_ID,
USER_ID,
activeFocusFixture,
candidateSnapshotFixture,
conversationSummaryFixture,
dossierFixture,
fakeAccounting,
receiptHandlers,
} from "./rectification-v9-test-support.ts";
const ACTION_ID = "aaaaaaaa-aaaa-4aaa-8aaa-aaaaaaaaaaaa";
const QUESTION_ID = "question-1";
function inferenceState() {
return buildInferenceState({
range_start: "04:50",
range_end: "05:10",
candidates: [
{ id: "05:00", time: "05:00", relative_support: 10 },
{ id: "05:10", time: "05:10", relative_support: 10 },
],
events: [
{ id: "e1", domain: "education", year: 2016, precision: "month" },
{ id: "e2", domain: "career", year: 2018, precision: "year" },
{ id: "e3", domain: "relationship", year: 2021, precision: "year" },
{ id: "e4", domain: "family", year: 2023, precision: "year" },
],
probes: [{
id: "p-cd",
semantic_key: "education.2016",
candidate_split_hash: "05:00|05:10",
domain: "education",
year: 2016,
question: "2016 年前后有没有高考或重要考试发挥失常?",
candidate_ids: ["05:00", "05:10"],
expected_outcomes: [
{ answer_class: "yes", supports: ["05:00"], conflicts: ["05:10"] },
{ answer_class: "no", supports: ["05:10"], conflicts: ["05:00"] },
{ answer_class: "unsure", supports: [], conflicts: [] },
],
information_gain: 0.4,
source: "dasha_boundary",
}],
});
}
function choiceDossier() {
const inference = inferenceState();
const snapshot = candidateSnapshotFixture();
Object.assign(snapshot.decision_receipt, { inference_state: inference });
return dossierFixture({
latestResult: snapshot,
conversationSummary: conversationSummaryFixture({
activeFocus: activeFocusFixture({
expectedAnswerSchema: {
choice: {
prompt: "2016 年前后,有没有明显高考或重要考试发挥失常?",
option_a: "是,大概就在那段时间",
option_b: "有类似,但年份不对或不够重大",
option_c: "没有明显发生",
option_d: "不记得 / 不确定",
},
probe_id: "p-cd",
semantic_key: "education.2016",
},
}),
}),
});
}
function choiceAccounting(overrides: Parameters<typeof fakeAccounting>[0] = {}) {
const receipts = new Map<string, Record<string, unknown>>();
return fakeAccounting({
...receiptHandlers,
get_agentic_rectification_case_dossier: () => choiceDossier(),
apply_agentic_rectification_choice_action: (_fn, args) => {
const existing = receipts.get(String(args.p_action_id));
if (existing) return { ...existing, idempotent: true };
const receipt = {
action_id: args.p_action_id,
status: "applied",
idempotent: false,
question_id: args.p_question_id,
option_id: args.p_option_id,
probe_id: args.p_inference && typeof args.p_inference === "object"
? (args.p_inference as { probe_id?: string }).probe_id ?? "p-cd"
: "p-cd",
revision: Number(args.p_expected_revision) + 1,
source_quote: args.p_source_quote,
derived_context: args.p_derived_context,
narration: args.p_narration,
focus_status: args.p_focus_status,
};
receipts.set(String(args.p_action_id), receipt);
return receipt;
},
append_agentic_rectification_turn: () => ({ turn_id: TURN_ID, idempotent: false }),
...overrides,
});
}
test("does not force-follow when the reader has scrolled upward", () => {
assert.equal(isNearBottom(1_000, 100, 400), false);
assert.equal(shouldFollowLatest(false), false);
assert.equal(isNearBottom(1_000, 520, 400), true);
assert.equal(shouldFollowLatest(true), true);
});
test("jump-to-latest stays hidden while a choice card still sits in the overlay band", () => {
assert.equal(shouldShowJumpToLatest(1_000, 100, 400), true);
assert.equal(shouldShowJumpToLatest(1_000, 520, 400), false);
assert.equal(shouldShowJumpToLatest(1_000, 500, 400, true), false);
assert.equal(shouldShowJumpToLatest(1_000, 200, 400, true), true);
});
test("choice quotes come from the option label, not the assistant question year", () => {
const question = "2016 年前后,有没有明显高考或重要考试发挥失常?";
assert.equal(quoteIsFromAssistantQuestion("2016年", question), true);
assert.equal(quoteIsFromAssistantQuestion("是,大概就在那段时间", question), false);
const card = {
question_id: QUESTION_ID,
method_id: "d5",
prompt: question,
why: "",
varga: null,
choice_mode: "A/B/C/D" as const,
options: [
{ key: "A" as const, label: "是,大概就在那段时间", role: "primary" as const },
{ key: "B" as const, label: "有类似,但年份不对或不够重大", role: "primary" as const },
{ key: "C" as const, label: "没有明显发生", role: "primary" as const },
{ key: "D" as const, label: "不记得 / 不确定", role: "secondary" as const },
],
stop_label: "先这样,先看当前范围",
stop_message: "先这样",
scoring: true,
probe_id: "p-cd",
case_revision: 1,
focus_id: FOCUS_ID,
};
assert.equal(userVisibleChoiceLine(card, "A"), "A. 是,大概就在那段时间");
});
test("clicking A applies the choice without invoking a language model", async () => {
const accounting = choiceAccounting();
const applied = await applyRectificationChoice(accounting.client, {
userId: USER_ID,
caseId: CASE_ID,
sessionId: SESSION_ID,
actionId: ACTION_ID,
action: CHOICE_ACTION,
questionId: QUESTION_ID,
probeId: "p-cd",
optionId: "A",
expectedRevision: inferenceState().revision,
});
assert.equal(applied.applied, true);
assert.equal(applied.optionId, "A");
assert.equal(applied.sourceQuote, "是,大概就在那段时间");
assert.equal(applied.derivedContext.sourceType, "structured_probe_answer");
assert.deepEqual(applied.derivedContext.referencedDateRange, {
start: "2016-01-01",
end: "2016-12-31",
});
assert.equal(applied.narration, composeChoiceNarration({
optionId: "A",
scoring: true,
appliedInference: true,
}));
const fns = accounting.calls.map((call) => call.fn);
assert.ok(fns.includes("apply_agentic_rectification_choice_action"));
assert.ok(fns.includes("append_agentic_rectification_turn"));
assert.equal(fns.includes("record_agentic_rectification_evidence_batch"), false);
assert.equal(fns.some((fn) => fn === "create_agentic_rectification_run_attempt" || fn.includes("stream")), false);
const persist = accounting.calls.find((call) => call.fn === "apply_agentic_rectification_choice_action");
assert.equal(persist?.args.p_focus_id, FOCUS_ID);
assert.equal(persist?.args.p_focus_status, "resolved");
assert.equal(persist?.args.p_source_quote, "是,大概就在那段时间");
assert.notEqual(persist?.args.p_source_quote, "2016年");
assert.ok(persist?.args.p_inference);
});
test("replaying the same actionId does not duplicate the applied receipt", async () => {
const accounting = choiceAccounting();
const command = {
userId: USER_ID,
caseId: CASE_ID,
sessionId: SESSION_ID,
actionId: ACTION_ID,
action: CHOICE_ACTION,
questionId: QUESTION_ID,
probeId: "p-cd",
optionId: "A" as const,
expectedRevision: inferenceState().revision,
};
const first = await applyRectificationChoice(accounting.client, command);
const second = await applyRectificationChoice(accounting.client, command);
assert.equal(first.idempotent, false);
assert.equal(second.idempotent, true);
assert.equal(
accounting.calls.filter((call) => call.fn === "apply_agentic_rectification_choice_action").length,
2,
);
const secondCall = accounting.calls.filter((call) => call.fn === "apply_agentic_rectification_choice_action").at(-1);
assert.equal(secondCall?.args.p_action_id, ACTION_ID);
});
test("keeps the applied answer when narration persistence fails", async () => {
const accounting = choiceAccounting({
append_agentic_rectification_turn: () => {
throw new Error("turn_write_failed");
},
});
const applied = await applyRectificationChoice(accounting.client, {
userId: USER_ID,
caseId: CASE_ID,
sessionId: SESSION_ID,
actionId: ACTION_ID,
action: CHOICE_ACTION,
questionId: QUESTION_ID,
probeId: "p-cd",
optionId: "A",
expectedRevision: inferenceState().revision,
});
assert.equal(applied.applied, true);
assert.equal(applied.status, "applied");
assert.equal(applied.narrationPersisted, false);
assert.match(applied.narration, /已记录你的选择/);
});
test("stop_and_review does not write an inference transition", async () => {
const accounting = choiceAccounting();
const applied = await applyRectificationChoice(accounting.client, {
userId: USER_ID,
caseId: CASE_ID,
sessionId: SESSION_ID,
actionId: ACTION_ID,
action: STOP_ACTION,
questionId: QUESTION_ID,
optionId: "stop",
expectedRevision: inferenceState().revision,
});
assert.equal(applied.optionId, "stop");
const persist = accounting.calls.find((call) => call.fn === "apply_agentic_rectification_choice_action");
assert.equal(persist?.args.p_inference, null);
assert.equal(persist?.args.p_focus_status, "skipped");
});
test("turn_decision stays inside the configured byte budget", () => {
const dossier = parseV9CaseDossier(choiceDossier());
assert.ok(dossier);
const projection = projectTurnDecision(dossier);
assert.equal(projection.projection, "turn_decision");
assert.ok(turnDecisionByteLength(projection) <= TURN_DECISION_MAX_BYTES);
assert.ok(!("birth_context" in projection));
assert.ok(!("baseline_birth_snapshot" in projection));
});
test("truncated and timed-out runs return concrete finish reasons", () => {
assert.equal(mapModelFinishToErrorCode({
finishReason: "length",
aborted: false,
timedOut: false,
answerText: "部分回答",
stepCount: 1,
maxSteps: 8,
}), "answer_truncated");
assert.equal(mapModelFinishToErrorCode({
finishReason: "stop",
aborted: true,
timedOut: true,
answerText: "",
stepCount: 1,
maxSteps: 8,
}), "run_timeout");
assert.equal(userFacingRunFailure("answer_truncated"), "模型输出达到上限,状态已记录。");
assert.equal(userFacingRunFailure("run_timeout"), "服务端运行超时,状态已记录。");
assert.equal(isIncompleteRunBanner("本轮处理未完成,请稍后再试"), true);
assert.equal(isIncompleteRunBanner("已记录你的选择,并更新了候选比较。"), false);
});
test("the public agent route treats structured choice as a non-model command", () => {
const route = readFileSync(new URL("../src/app/api/rectification/agent/route.ts", import.meta.url), "utf8");
const start = route.indexOf("if (isStructuredChoice)");
const end = route.indexOf("const selectedModel", start);
assert.ok(start >= 0 && end > start);
const block = route.slice(start, end);
assert.match(block, /applyRectificationChoice\(accounting/);
assert.doesNotMatch(block, /runV9AgentTurn/);
assert.doesNotMatch(block, /getRectificationV9Agent/);
assert.doesNotMatch(block, /authorizeUsage/);
assert.match(route, /"answer_choice"/);
assert.match(route, /"stop_and_review"/);
});