803 lines
36 KiB
TypeScript
803 lines
36 KiB
TypeScript
/**
|
|
* Render smoke tests — the lesson from the v2 bugs: protocol tests don't see
|
|
* rendering. These render the interaction cards to static HTML and assert on
|
|
* the output. No DOM needed (react-dom/server).
|
|
*
|
|
* tsx note: tsx ignores tsconfig's `jsx` field and uses the CLASSIC JSX
|
|
* transform (React.createElement), so `React` must be a resolvable global.
|
|
* The shim below provides it; Vite/Next use the automatic runtime and don't
|
|
* care.
|
|
*/
|
|
|
|
import React from "react";
|
|
(globalThis as Record<string, unknown>).React = React;
|
|
|
|
import assert from "node:assert/strict";
|
|
import { describe, it } from "node:test";
|
|
import { renderToStaticMarkup } from "react-dom/server";
|
|
import { MessageResponse } from "../components/ai-elements/message.tsx";
|
|
import { QuestionDock } from "../components/question-dock.tsx";
|
|
import { ProbeDoneCard } from "../components/artifacts/probe-done-card.tsx";
|
|
import { DispatchCard } from "../components/artifacts/dispatch-card.tsx";
|
|
import { JobsContext } from "../components/jobs-context.ts";
|
|
import { ChatActionsContext } from "../components/chat.tsx";
|
|
import type { JobRecord } from "../src/core/contracts.ts";
|
|
import { AnsweredCard } from "../components/answered-card.tsx";
|
|
import {
|
|
buildAnswer,
|
|
normalizeOptions,
|
|
resolveCorrect,
|
|
questionNeedsReveal,
|
|
type QuestionData,
|
|
} from "../src/core/question-core.ts";
|
|
|
|
const quizInput = {
|
|
question: "Which planet is closest to the sun?",
|
|
options: [{ label: "Venus" }, { label: "Mercury" }, { label: "Mars" }],
|
|
correctAnswer: "Mercury",
|
|
explanation: "Mercury is the innermost planet.",
|
|
};
|
|
|
|
function quizQD(): QuestionData {
|
|
const options = normalizeOptions(quizInput.options);
|
|
return {
|
|
id: "q1",
|
|
kind: "quiz",
|
|
question: quizInput.question,
|
|
mode: "single-select",
|
|
options,
|
|
correctIndices: resolveCorrect(quizInput.correctAnswer, options).indices,
|
|
explanation: quizInput.explanation,
|
|
createdAt: new Date().toISOString(),
|
|
};
|
|
}
|
|
|
|
function askQD(): QuestionData {
|
|
const options = normalizeOptions([{ label: "Very" }, { label: "Somewhat" }]);
|
|
return {
|
|
id: "a1",
|
|
kind: "ask",
|
|
question: "How confident are you?",
|
|
mode: "single-select",
|
|
options,
|
|
createdAt: new Date().toISOString(),
|
|
};
|
|
}
|
|
|
|
function renderDock(props: {
|
|
toolName: "quiz" | "probe" | "ask_user_question" | "explain" | "recall";
|
|
toolCallId: string;
|
|
input: never;
|
|
onSubmit?: () => void;
|
|
}) {
|
|
// expanded=true renders the full form (the collapsed bar is covered by the
|
|
// QuestionDockBar test below). No scroll context needed — the DockOverlay
|
|
// in the chat view owns that.
|
|
return renderToStaticMarkup(
|
|
<QuestionDock {...props} expanded={true} onExpand={() => {}} onSubmit={props.onSubmit ?? (() => {})} />,
|
|
);
|
|
}
|
|
|
|
describe("question submission draft restoration", () => {
|
|
const draft = {
|
|
selection: { selected: new Set([0]), otherSelected: true, dontKnow: false },
|
|
thinking: "preserved reasoning",
|
|
other: "preserved custom answer",
|
|
text: "preserved free-text answer",
|
|
confidenceIndex: 4,
|
|
};
|
|
|
|
it("restores a quiz selection, thinking, and confidence when its form remounts", () => {
|
|
const html = renderToStaticMarkup(<QuestionDock toolName="quiz" toolCallId="q1" input={quizInput} expanded onExpand={() => {}} onSubmit={() => {}} initialDraft={draft} />);
|
|
assert.match(html, /role="radio" aria-checked="true" data-kbdrow="0"/);
|
|
assert.ok(html.includes("preserved reasoning"));
|
|
assert.ok(html.includes("90%"));
|
|
});
|
|
|
|
it("restores an ask's custom answer and a free-text answer", () => {
|
|
const custom = renderToStaticMarkup(<QuestionDock toolName="ask_user_question" toolCallId="ask1" input={{ question: "Fixture?", options: [{ label: "Listed answer" }, { label: "Another answer" }], multiSelect: true }} expanded onExpand={() => {}} onSubmit={() => {}} initialDraft={draft} />);
|
|
assert.ok(custom.includes('value="preserved custom answer"'));
|
|
assert.match(custom, /aria-checked="true" data-testid="other-option"/);
|
|
const text = renderToStaticMarkup(<QuestionDock toolName="ask_user_question" toolCallId="ask2" input={{ question: "Fixture?" }} expanded onExpand={() => {}} onSubmit={() => {}} initialDraft={draft} />);
|
|
assert.ok(text.includes("preserved free-text answer"));
|
|
});
|
|
});
|
|
|
|
describe("DispatchCard — the BYOK refusal tone (owner, 2026-09-22)", () => {
|
|
const job = (over: Partial<JobRecord>): JobRecord => ({
|
|
id: "j1",
|
|
kind: "research",
|
|
chatId: "c1",
|
|
toolCallId: "t1",
|
|
status: "error",
|
|
activity: "",
|
|
createdAt: new Date().toISOString(),
|
|
updatedAt: new Date().toISOString(),
|
|
...over,
|
|
});
|
|
const inJobs = (jobs: JobRecord[]) => (
|
|
<JobsContext.Provider value={jobs}>
|
|
<ChatActionsContext.Provider
|
|
value={{ chatId: "c1", syncFromServer: async () => {}, sendText: () => {}, openSetup: () => {} }}
|
|
>
|
|
<DispatchCard toolCallId="t1" kind="research" />
|
|
</ChatActionsContext.Provider>
|
|
</JobsContext.Provider>
|
|
);
|
|
|
|
it("a models-refused job is INFO: quiet card, 'needs model setup', deep-link to the dialog", () => {
|
|
const html = renderToStaticMarkup(
|
|
inJobs([job({ error: "No model configured — set up your models first (missing: cheap)" })]),
|
|
);
|
|
assert.ok(html.includes('data-testid="dispatch-refused"'));
|
|
assert.ok(html.includes("needs model setup"));
|
|
assert.ok(html.includes('data-testid="dispatch-setup-link"'));
|
|
assert.ok(html.includes("Set up models"));
|
|
assert.ok(!html.includes("failed"));
|
|
assert.ok(!html.includes("text-red-500"));
|
|
});
|
|
|
|
it("a key-refused job is INFO: 'needs a provider key', 'Add a key' (zero-keys wave, 66dbea9)", () => {
|
|
const html = renderToStaticMarkup(
|
|
inJobs([job({ error: "No OpenRouter key is set — add one in Settings → Keys" })]),
|
|
);
|
|
assert.ok(html.includes('data-testid="dispatch-refused"'));
|
|
assert.ok(html.includes("needs a provider key"));
|
|
assert.ok(html.includes("Add a key"));
|
|
assert.ok(html.includes('data-testid="dispatch-setup-link"'));
|
|
assert.ok(!html.includes("failed"));
|
|
});
|
|
|
|
it("a real failure keeps the honest red failure", () => {
|
|
const html = renderToStaticMarkup(inJobs([job({ error: "FAILED: headless crashed" })]));
|
|
assert.ok(html.includes('data-testid="dispatch-settled"'));
|
|
assert.ok(html.includes("failed"));
|
|
assert.ok(html.includes("text-red-500"));
|
|
assert.ok(!html.includes("dispatch-refused"));
|
|
});
|
|
|
|
it("a finished worker reads finished, independently of delivery", () => {
|
|
const html = renderToStaticMarkup(inJobs([job({ status: "done", error: undefined })]));
|
|
assert.ok(html.includes("finished"));
|
|
});
|
|
});
|
|
|
|
describe("render smoke tests", () => {
|
|
it("QuizCard renders question, options, and the automatic 'I don't know'", () => {
|
|
const html = renderDock({ toolName: "quiz", toolCallId: "q1", input: quizInput as never });
|
|
assert.match(html, /Which planet is closest to the sun\?/);
|
|
assert.match(html, /Mercury/);
|
|
assert.match(html, /I don't know|I don't know/);
|
|
// The learner's voice: a labeled "Your thinking" composer-twin lives
|
|
// INSIDE the dock — never called optional, inviting the dump.
|
|
assert.ok(html.includes('data-testid="thinking-field"'), "thinking field present");
|
|
assert.match(html, /YOUR THINKING/, "labeled as the learner's voice (all caps, owner round 3)");
|
|
assert.match(html, /Why this answer\? Whatever comes to mind/, "placeholder invites, not apologizes (owner, placeholder round 2026-10-07)");
|
|
assert.ok(!html.includes("optional"), "the field is never called optional");
|
|
assert.match(html, /Submit answer/, "send becomes Submit answer");
|
|
assert.match(html, /↵ submits · ⇧↵ newline/, "the standard chord is discoverable (owner round 5)");
|
|
// The v1 desync bug: options are buttons with ARIA roles, not native inputs.
|
|
assert.ok(!html.includes('type="radio"'), "no native radio inputs");
|
|
assert.ok(html.includes('role="radio"'), "button-based radio role present");
|
|
});
|
|
|
|
it("QuizCard never leaks the explanation or the correct answer before submit", () => {
|
|
const html = renderDock({ toolName: "quiz", toolCallId: "q1", input: quizInput as never });
|
|
assert.ok(!html.includes("innermost"), "explanation must not render pre-answer");
|
|
});
|
|
|
|
it("ask dock renders the editable Other row + thinking field, no grading affordance", () => {
|
|
const html = renderDock({
|
|
toolName: "ask_user_question",
|
|
toolCallId: "a1",
|
|
input: { question: "How confident are you?", options: [{ label: "Very" }, { label: "Somewhat" }] } as never,
|
|
});
|
|
assert.ok(!html.includes("I don't know"), "ask has no 'I don't know'");
|
|
// Other is an editable OPTION ROW — a sibling of the options, with the
|
|
// radio circle + an inline input; picking it never displaces thinking.
|
|
assert.ok(html.includes('data-testid="other-option"'), "Other is an option row");
|
|
assert.match(html, /Type your own answer/, "type directly into the option");
|
|
assert.ok(html.includes('role="radio"'), "Other row keeps the option role");
|
|
assert.match(html, /aria-checked="false"/, "circle starts unfilled");
|
|
assert.ok(html.includes('data-testid="thinking-field"'), "thinking field coexists with Other");
|
|
assert.match(html, /YOUR THINKING/, "ask select-mode also gets the learner's voice");
|
|
});
|
|
|
|
it("multi-select ask renders checkbox rows and the select-all hint", () => {
|
|
const html = renderDock({
|
|
toolName: "ask_user_question",
|
|
toolCallId: "m1",
|
|
input: {
|
|
question: "Which topics interest you?",
|
|
multiSelect: true,
|
|
options: [{ label: "Physics" }, { label: "Philosophy" }],
|
|
} as never,
|
|
});
|
|
assert.match(html, /select all that apply/);
|
|
assert.ok(html.includes('role="group"'), "checkbox group");
|
|
assert.ok((html.match(/role="checkbox"/g) ?? []).length >= 3, "options AND the Other row are checkboxes");
|
|
assert.ok(html.includes('data-testid="other-option"'), "Other is still an option row in multi-select");
|
|
// The bigger \u2713 glyph (text-base, 2026-09-24 owner round) only
|
|
// mounts on a CHECKED row — unreachable from the unselected static
|
|
// render, so it's live-verified (Playwright: 14px computed in the
|
|
// 16px box) instead of pinned here.
|
|
});
|
|
|
|
it("a malformed question input renders the quiet fallback, never a crash", () => {
|
|
// The seam rule (tool-parts.ts): the dock PARSES the model-authored
|
|
// input instead of casting it — the live crash shape (options as a
|
|
// JSON string) must render the fallback, not die on .map.
|
|
const html = renderDock({
|
|
toolName: "quiz",
|
|
toolCallId: "bad",
|
|
input: { question: "q?", options: JSON.stringify([{ label: "a" }, { label: "b" }]) } as never,
|
|
});
|
|
assert.ok(html.includes('data-testid="malformed-question"'));
|
|
assert.match(html, /malformed/);
|
|
});
|
|
|
|
it("phase scales the thinking field: teaching invites depth, probing stays quick", () => {
|
|
const teach = renderToStaticMarkup(
|
|
<QuestionDock
|
|
toolName="quiz"
|
|
toolCallId="t1"
|
|
input={quizInput as never}
|
|
expanded={true}
|
|
phase="teach"
|
|
onExpand={() => {}}
|
|
onSubmit={() => {}}
|
|
/>,
|
|
);
|
|
assert.ok(teach.includes("min-h-[4.5rem]"), "teach: a generous dump area");
|
|
const probe = renderDock({ toolName: "quiz", toolCallId: "q1", input: quizInput as never });
|
|
assert.ok(!probe.includes("min-h-[4.5rem]"), "probe (default): stays compact");
|
|
assert.ok(probe.includes("min-h-10"), "probe: one quiet row");
|
|
});
|
|
|
|
it("text-mode asks keep Enter-to-submit and the answer label", () => {
|
|
const html = renderDock({
|
|
toolName: "ask_user_question",
|
|
toolCallId: "a2",
|
|
input: { question: "Explain the chain rule in your own words." } as never,
|
|
});
|
|
assert.ok(html.includes('data-testid="answer-field"'), "text mode is the answer field");
|
|
assert.match(html, /Your answer/, "labeled as the answer");
|
|
assert.ok(!html.includes('data-testid="thinking-field"'), "no thinking field in text mode (v1)");
|
|
assert.ok(!html.includes("⌘/Ctrl"), "no dump shortcut hint — Enter already submits");
|
|
});
|
|
|
|
it("AnsweredCard grades quiz answers (✓/✗ + explanation)", () => {
|
|
const answer = buildAnswer(quizQD(), { values: ["Venus"] });
|
|
const html = renderToStaticMarkup(<AnsweredCard answer={answer} />);
|
|
assert.match(html, /incorrect/);
|
|
// Tucked for ALL questions (owner, 2026-10-07): the explanation sits
|
|
// behind the Why toggle — never center stage.
|
|
assert.ok(html.includes("tucked-explanation"), "the Why toggle mounts");
|
|
assert.ok(!html.includes("innermost"), "explanation hidden until opened");
|
|
// Glyph-shaped SVG marks (owner preference) — pinned via data-mark.
|
|
assert.ok(html.includes('data-mark="check"') && html.includes('data-mark="x"'));
|
|
});
|
|
|
|
it("AnsweredCard shows the confidence rail on quiz answers (record, no disc)", () => {
|
|
// The slider's rail, disc removed: same paint, same ticks, the % read-out.
|
|
const answer = buildAnswer(quizQD(), { values: ["Venus"], confidence: 70 });
|
|
const html = renderToStaticMarkup(<AnsweredCard answer={answer} />);
|
|
assert.match(html, /data-testid="answered-confidence"/);
|
|
assert.match(html, /data-testid="confidence-rail"/);
|
|
assert.match(html, /70%/);
|
|
// It is a RECORD, not a control — no range input on the card.
|
|
assert.ok(!html.includes('type="range"'));
|
|
});
|
|
|
|
it("AnsweredCard hides the rail where no confidence exists (IDK, older records)", () => {
|
|
const idk = buildAnswer(quizQD(), { values: [], dontKnow: true });
|
|
const idkHtml = renderToStaticMarkup(<AnsweredCard answer={idk} />);
|
|
assert.ok(!idkHtml.includes("answered-confidence"), "IDK sent no confidence");
|
|
|
|
const ask = buildAnswer(askQD(), { other: "Shaky" });
|
|
const askHtml = renderToStaticMarkup(<AnsweredCard answer={ask} />);
|
|
assert.ok(!askHtml.includes("answered-confidence"), "ask never had one");
|
|
});
|
|
|
|
it("AnsweredCard renders the learner's own Other answer AS an option row", () => {
|
|
const answer = buildAnswer(askQD(), { other: "Fairly confident — but shaky on applications" });
|
|
const html = renderToStaticMarkup(<AnsweredCard answer={answer} />);
|
|
assert.ok(html.includes('data-testid="answered-other"'), "the Other answer is an option row");
|
|
assert.match(html, /border-primary bg-accent/, "styled like its selected siblings");
|
|
assert.match(html, /shaky on applications/, "the typed words render");
|
|
// Inside the options container — same list, not a loose quote outside it.
|
|
assert.ok(
|
|
html.indexOf("Very") < html.indexOf("answered-other"),
|
|
"the Other row sits with the listed options",
|
|
);
|
|
});
|
|
|
|
it("AnsweredCard promotes the learner's thinking to a first-class block", () => {
|
|
const answer = buildAnswer(quizQD(), {
|
|
values: ["Venus"],
|
|
thinking: "I picked Venus because it looks brightest — but now I realize brightness isn't distance.",
|
|
});
|
|
const html = renderToStaticMarkup(<AnsweredCard answer={answer} />);
|
|
assert.ok(html.includes('data-testid="your-thinking"'), "the thinking row exists");
|
|
assert.match(html, /YOUR THINKING/, "labeled as the learner's voice (all caps, owner round 3)");
|
|
// COLLAPSED ARRIVAL (owner round 2, 2026-10-05): the SSR output is the
|
|
// trigger row only — the thinking text is one click away, never loud.
|
|
assert.ok(html.includes('data-testid="your-thinking-trigger"'), "the trigger row");
|
|
assert.ok(html.includes('aria-expanded="false"'), "arrives collapsed");
|
|
// ONE GEOMETRY, NO BAND IN CARDS (owner round five, 2026-10-07): the
|
|
// accent lift read too light — rail only, heights match everywhere.
|
|
const yourRow = html.slice(html.indexOf('data-testid="your-thinking"'));
|
|
assert.ok(!yourRow.slice(0, 400).includes("bg-card") && !yourRow.slice(0, 400).includes("bg-accent"), "no band inside the card");
|
|
// THE RAIL GROUPS (owner correction): the label sits INSIDE the rail —
|
|
// ownership readable from the rail alone, collapsed or expanded.
|
|
const railAt = html.indexOf("border-l-[calc(3ch-14px)]");
|
|
const trigAt = html.indexOf("your-thinking-trigger");
|
|
assert.ok(railAt !== -1 && railAt < trigAt, "the rail opens before the label (same quote)");
|
|
assert.ok(!html.includes("brightness isn't distance"), "the text is hidden until expanded");
|
|
assert.ok(!html.includes("Note:"), "no longer a muted 'Note:' footnote");
|
|
// The thinking sits ABOVE the verdict's explanation — the dialogue of
|
|
// thinking, not an afterthought.
|
|
assert.ok(
|
|
html.indexOf("your-thinking") < html.indexOf("tucked-explanation"),
|
|
"thinking before the explanation toggle",
|
|
);
|
|
});
|
|
|
|
it("AnsweredCard does NOT grade ask answers", () => {
|
|
const qd: QuestionData = {
|
|
id: "a1",
|
|
kind: "ask",
|
|
question: "How confident are you?",
|
|
mode: "single-select",
|
|
options: normalizeOptions([{ label: "Very" }, { label: "Somewhat" }]),
|
|
createdAt: new Date().toISOString(),
|
|
};
|
|
const answer = buildAnswer(qd, { values: ["Somewhat"] });
|
|
const html = renderToStaticMarkup(<AnsweredCard answer={answer} />);
|
|
assert.ok(!html.includes("incorrect") && !html.includes('data-mark="x"'), "ask answers never get quiz grading UI");
|
|
assert.match(html, /Somewhat/);
|
|
});
|
|
|
|
it("AnsweredCard renders dismissal as skipped", () => {
|
|
const answer = { ...buildAnswer(quizQD(), { dontKnow: true }), dismissed: true };
|
|
const html = renderToStaticMarkup(<AnsweredCard answer={answer} />);
|
|
assert.match(html, /skipped/);
|
|
});
|
|
});
|
|
|
|
describe("math rendering (streamdown + KaTeX)", () => {
|
|
// Standard authoring: $...$ inline and $$ on separate lines for display.
|
|
// The legacy backslash reader remains for existing content.
|
|
// Full coverage: render-markdown.test.tsx.
|
|
it("renders standard $...$ as inline math", () => {
|
|
const html = renderToStaticMarkup(<MessageResponse>{"The map $f(x) = x^2$ is smooth."}</MessageResponse>);
|
|
assert.ok(html.includes("katex"), `$-math must render, got: ${html.slice(0, 300)}`);
|
|
assert.ok(!html.includes("katex-display"), "single-dollar math is inline");
|
|
});
|
|
|
|
it("renders display math ($$ fenced on own lines — the transform's target form)", () => {
|
|
const html = renderToStaticMarkup(<MessageResponse>{"$$\n\\int_0^1 x^2\\,dx = \\tfrac{1}{3}\n$$"}</MessageResponse>);
|
|
assert.ok(html.includes("katex-display"), "fenced display math should render as display KaTeX");
|
|
});
|
|
|
|
it("one-line $$...$$ renders inline (micromark behavior — the transform relies on it for \(...\))", () => {
|
|
// micromark treats a one-line $$...$$ inside a paragraph as 2-dollar
|
|
// inline mathText, not flow math. That is exactly how the transformed
|
|
// \(...\) renders inline; pin the behavior so a renderer swap can't
|
|
// silently change it.
|
|
const html = renderToStaticMarkup(<MessageResponse>{"$$\\int_0^1 x^2\\,dx$$"}</MessageResponse>);
|
|
assert.ok(html.includes("katex") && !html.includes("katex-display"), "one-liner should render inline");
|
|
});
|
|
|
|
it("does NOT eat escaped currency (\\$5 stays text)", () => {
|
|
// With dollar math enabled, Markdown-escaped dollars remain literal.
|
|
const html = renderToStaticMarkup(<MessageResponse>{"It costs \\$5 and \\$10 more."}</MessageResponse>);
|
|
assert.ok(!html.includes("katex"), "escaped currency must not parse as math");
|
|
assert.match(html, /\$5/);
|
|
});
|
|
|
|
it("quiz cards render math in question, options, and explanation", () => {
|
|
const mathQuiz = {
|
|
question: "What is $\\frac{dy}{dx}$ of $y = x^2$?",
|
|
options: [{ label: "$2x$" }, { label: "$x^2$" }, { label: "$x$" }],
|
|
correctAnswer: "$2x$",
|
|
explanation: "The power rule gives $\\frac{dy}{dx} = 2x$.",
|
|
};
|
|
const pendingHtml = renderDock({ toolName: "quiz", toolCallId: "m1", input: mathQuiz as never });
|
|
assert.ok(pendingHtml.includes("katex"), "pending dock: question/options math renders");
|
|
|
|
const options = normalizeOptions(mathQuiz.options);
|
|
const qd: QuestionData = {
|
|
id: "m1",
|
|
kind: "quiz",
|
|
question: mathQuiz.question,
|
|
mode: "single-select",
|
|
options,
|
|
correctIndices: resolveCorrect(mathQuiz.correctAnswer, options).indices,
|
|
explanation: mathQuiz.explanation,
|
|
createdAt: new Date().toISOString(),
|
|
};
|
|
const answeredHtml = renderToStaticMarkup(
|
|
<AnsweredCard answer={buildAnswer(qd, { values: ["$2x$"] })} />,
|
|
);
|
|
assert.ok(answeredHtml.includes("katex"), "answered card: option/explanation math renders");
|
|
});
|
|
});
|
|
|
|
describe("teach-check reveal (W5 phase separation)", () => {
|
|
// The collapsed dock renders the bar AND the hidden form (it stays mounted
|
|
// so a half-made selection survives) — assertions target the bar species.
|
|
it("teach-phase check collapses into the invitation bar, not the scroll bar", () => {
|
|
const html = renderToStaticMarkup(
|
|
<QuestionDock
|
|
toolName="quiz"
|
|
toolCallId="t1"
|
|
input={quizInput as never}
|
|
expanded={false}
|
|
reveal
|
|
onExpand={() => {}}
|
|
onSubmit={() => {}}
|
|
/>,
|
|
);
|
|
assert.ok(html.includes('data-testid="teach-check-reveal"'), "invitation affordance present");
|
|
assert.match(html, /Check your understanding/);
|
|
assert.match(html, /Which planet is closest/, "question teaser invites, not minimizes");
|
|
assert.match(html, /Open/, "a clear reveal action");
|
|
});
|
|
|
|
it("probe questions keep the plain collapsed bar (no invitation copy)", () => {
|
|
const html = renderToStaticMarkup(
|
|
<QuestionDock
|
|
toolName="quiz"
|
|
toolCallId="q1"
|
|
input={quizInput as never}
|
|
expanded={false}
|
|
onExpand={() => {}}
|
|
onSubmit={() => {}}
|
|
/>,
|
|
);
|
|
assert.ok(html.includes('data-testid="question-dock-bar"'), "the scroll-collapse strip");
|
|
assert.ok(!html.includes("Check your understanding"), "no invitation copy outside teach phase");
|
|
assert.match(html, /Answer/, "the strip keeps its Answer affordance");
|
|
});
|
|
|
|
it("after the reveal, the teach check renders the full form (probe default, unchanged)", () => {
|
|
// Post-click: expanded=true — exactly today's dock. The probe auto-open
|
|
// path renders the identical form; this is also the regression pin for it.
|
|
const html = renderDock({ toolName: "quiz", toolCallId: "t1", input: quizInput as never });
|
|
assert.ok(html.includes('data-testid="question-dock"'), "the form is the dock's content once expanded");
|
|
assert.match(html, /Submit answer/);
|
|
});
|
|
});
|
|
|
|
// ─── probe-done card (the malformed-assessment crash) ─────────────────────
|
|
// The live bug: the model stringified `assessment` (a rejected input), and
|
|
// the card's input-preview fallback trusted the shape — strands died. The
|
|
// preview now reads only pending inputs; an errored part renders the quiet
|
|
// retry note.
|
|
describe("probe-done card", () => {
|
|
const validAssessment = {
|
|
strands: [
|
|
{ strand: "modal logic basics", assessment: "Floor: solid intuition; ceiling: notation." },
|
|
{ strand: "indexical 'actual'", assessment: "Genuine floor — prior exposure likely." },
|
|
],
|
|
summary: "Bracketed: teach the symbols on the intuition he has.",
|
|
};
|
|
|
|
it("renders the strand-by-strand assessment — the CALL is its home (context diet 2026-10-09)", () => {
|
|
const html = renderToStaticMarkup(
|
|
<ProbeDoneCard
|
|
part={{ state: "output-available", input: { assessment: validAssessment }, output: { ok: true } } as never}
|
|
/>,
|
|
);
|
|
assert.ok(html.includes('data-testid="probe-done-card"'));
|
|
assert.match(html, /modal logic basics/);
|
|
assert.match(html, /indexical 'actual'|indexical 'actual'/);
|
|
assert.match(html, /Bracketed/);
|
|
// An ok output alone (the pre-diet echo shape) renders the quiet
|
|
// fallback, never strands: the assessment no longer rides the result.
|
|
const echoless = renderToStaticMarkup(
|
|
<ProbeDoneCard part={{ state: "output-available", output: { ok: true } } as never} />,
|
|
);
|
|
assert.match(echoless, /Writing the assessment…/);
|
|
assert.ok(!echoless.includes("modal logic basics"), "no assessment echo in the result");
|
|
});
|
|
|
|
it("previews the assessment from a PENDING (validated) input", () => {
|
|
const html = renderToStaticMarkup(
|
|
<ProbeDoneCard part={{ state: "input-available", input: { assessment: validAssessment } } as never} />,
|
|
);
|
|
assert.match(html, /modal logic basics/);
|
|
});
|
|
|
|
it("a REJECTED assessment (the stringified live shape) renders the retry note, never the garbage", () => {
|
|
// The exact live crash: assessment arrived as a string, zod rejected
|
|
// the call, the model retried — this dead part must not explode.
|
|
const html = renderToStaticMarkup(
|
|
<ProbeDoneCard
|
|
part={{
|
|
state: "output-error",
|
|
errorText: "Validation failed",
|
|
input: { assessment: "strands: [{\"strand\": ...stringified garbage" },
|
|
} as never}
|
|
/>,
|
|
);
|
|
assert.match(html, /malformed/);
|
|
assert.match(html, /retried/);
|
|
assert.ok(!html.includes("strands:"), "the rejected input never renders");
|
|
});
|
|
|
|
it("a mid-stream partial without strands renders the writing note, not a crash", () => {
|
|
const html = renderToStaticMarkup(
|
|
<ProbeDoneCard part={{ state: "input-streaming", input: { assessment: { summary: "…" } } } as never} />,
|
|
);
|
|
assert.match(html, /Writing the assessment/);
|
|
});
|
|
});
|
|
|
|
// ─── the explain reveal (free-recall, owner 2026-09-27) ────────────────────
|
|
|
|
describe("AnsweredCard: the explain variant", () => {
|
|
const explainQD = (): QuestionData => ({
|
|
id: "e1",
|
|
kind: "explain",
|
|
question: "Explain why the null result matters.",
|
|
mode: "text",
|
|
options: [],
|
|
committedAnswer: "It ruled out the luminiferous aether — a stationary aether predicts a fringe shift, and none was seen.",
|
|
createdAt: new Date().toISOString(),
|
|
});
|
|
|
|
it("reveals the single committed answer (reasoning inside it) with NO verdict chips", () => {
|
|
const html = renderToStaticMarkup(<AnsweredCard answer={buildAnswer(explainQD(), { text: "Something about relativity?" })} />);
|
|
assert.match(html, /Explain/, "labeled as the explain kind");
|
|
assert.match(html, /The answer:/);
|
|
assert.match(html, /luminiferous aether/);
|
|
assert.match(html, /fringe shift/);
|
|
assert.ok(!html.includes("Why:"), "no separate explanation field — one committed answer (owner 2026-09-27)");
|
|
assert.ok(!html.includes("incorrect") && !html.includes("correct</"), "no client grading");
|
|
assert.ok(!html.includes('data-mark="x"') && !html.includes('data-mark="check"'));
|
|
});
|
|
|
|
it("renders his free text as the answer", () => {
|
|
const html = renderToStaticMarkup(<AnsweredCard answer={buildAnswer(explainQD(), { text: "Something about relativity?" })} />);
|
|
assert.match(html, /Something about relativity\?/);
|
|
});
|
|
|
|
});
|
|
|
|
// ─── probe-phase naming (owner, 2026-10-09): the probe/recall rename —
|
|
// both surfaces carry the tool's OWN name ────────────────────────────
|
|
|
|
describe("probe-phase naming: Probe/Recall labels", () => {
|
|
it("the dock labels a probe quiz as Probe (composer surface)", () => {
|
|
const html = renderDock({ toolName: "probe", toolCallId: "p1", input: quizInput as never });
|
|
assert.match(html, /Probe/, "the chip carries the tool's own name");
|
|
assert.ok(!html.includes(">Quiz<"), "not the teach label");
|
|
assert.match(html, /Which planet is closest to the sun\?/, "still a quiz-shaped card");
|
|
});
|
|
|
|
it("the dock labels a recall question as Recall", () => {
|
|
const html = renderDock({ toolName: "recall", toolCallId: "r1", input: { question: "Explain the chain rule.", correctAnswer: "Outer function times the derivative of the inner." } as never });
|
|
assert.match(html, /Recall/);
|
|
assert.ok(!html.includes(">Explain<"), "not the teach label");
|
|
});
|
|
|
|
it("an answered probe labels Probe — the phase boundary visible in the transcript", () => {
|
|
const qd: QuestionData = {
|
|
id: "p2",
|
|
kind: "quiz",
|
|
question: "Which planet is closest to the sun?",
|
|
mode: "single-select",
|
|
options: [{ label: "Mercury", value: "mercury" }, { label: "Venus", value: "venus" }],
|
|
correctIndices: [1],
|
|
committedAnswer: "Mercury — closest orbit.",
|
|
explanation: "Mercury, always Mercury.",
|
|
createdAt: new Date().toISOString(),
|
|
};
|
|
const html = renderToStaticMarkup(<AnsweredCard answer={buildAnswer(qd, { values: ["mercury"] }, "probe")} />);
|
|
assert.match(html, /Probe/);
|
|
assert.ok(!html.includes(">Quiz<"), "not the teach label");
|
|
});
|
|
|
|
it("an answered recall labels Recall", () => {
|
|
const recallQD: QuestionData = {
|
|
id: "r2",
|
|
kind: "explain",
|
|
question: "Explain why the null result matters.",
|
|
mode: "text",
|
|
options: [],
|
|
committedAnswer: "It ruled out the luminiferous aether.",
|
|
createdAt: new Date().toISOString(),
|
|
};
|
|
const html = renderToStaticMarkup(<AnsweredCard answer={buildAnswer(recallQD, { text: "A hand-wave." }, "recall")} />);
|
|
assert.match(html, /Recall/);
|
|
assert.ok(!html.includes(">Explain<"), "not the teach label");
|
|
});
|
|
|
|
it("old records (no tool) keep the kind labels", () => {
|
|
const oldQD: QuestionData = {
|
|
id: "o2",
|
|
kind: "explain",
|
|
question: "Explain why the null result matters.",
|
|
mode: "text",
|
|
options: [],
|
|
committedAnswer: "It ruled out the luminiferous aether.",
|
|
createdAt: new Date().toISOString(),
|
|
};
|
|
const html = renderToStaticMarkup(<AnsweredCard answer={buildAnswer(oldQD, { text: "A hand-wave." })} />);
|
|
assert.match(html, /Explain/, "no tool field → the standing kind label");
|
|
});
|
|
});
|
|
|
|
// ─── the two roles in the UI (predict vs check, owner 2026-10-03) ─────────
|
|
// Predictions arrive OPEN before the teaching; checks arrive collapsed after
|
|
// it. The answered card tells them apart (eyebrow) and tucks a prediction's
|
|
// explanation (the teaching that follows IS the explanation) while a check's
|
|
// stays inline.
|
|
describe("predict vs check rendering", () => {
|
|
it("a predict quiz renders the 'Prediction' eyebrow in the dock", () => {
|
|
const html = renderToStaticMarkup(
|
|
<QuestionDock
|
|
{...{
|
|
toolName: "quiz" as const,
|
|
toolCallId: "call-p1",
|
|
input: { ...quizInput, type: "predict", nodeId: "planet-order" },
|
|
}}
|
|
expanded={true}
|
|
onExpand={() => {}}
|
|
onSubmit={() => {}}
|
|
/>,
|
|
);
|
|
assert.ok(html.includes("Prediction"), `eyebrow missing: ${html.slice(0, 200)}`);
|
|
});
|
|
|
|
it("a check quiz reads the 'Check' eyebrow — the type name IS the label (owner, 2026-10-09)", () => {
|
|
const html = renderToStaticMarkup(
|
|
<QuestionDock
|
|
{...{
|
|
toolName: "quiz" as const,
|
|
toolCallId: "call-c1",
|
|
input: { ...quizInput, type: "check", nodeId: "planet-order" },
|
|
}}
|
|
expanded={true}
|
|
onExpand={() => {}}
|
|
onSubmit={() => {}}
|
|
/>,
|
|
);
|
|
assert.ok(html.includes("Check"), "the chip carries the question's type");
|
|
assert.ok(!html.includes(">Quiz<"), "the generic label is retired");
|
|
});
|
|
|
|
it("a prediction answer SHOWS the verdict but TUCKS the explanation behind a toggle", () => {
|
|
const qd: QuestionData = { ...quizQD(), questionType: "predict" };
|
|
const answer = buildAnswer(qd, { values: ["Venus"] }); // wrong on purpose
|
|
const html = renderToStaticMarkup(<AnsweredCard answer={answer} />);
|
|
assert.ok(html.includes("Prediction"), "the eyebrow distinguishes the card");
|
|
assert.ok(/incorrect|wrong|×|✗/i.test(html), "the verdict still shows (owner: predictions are graded)");
|
|
assert.ok(!html.includes("Why:"), "the explanation is not center stage");
|
|
assert.ok(html.includes("tucked-explanation"), "it sits behind the toggle");
|
|
});
|
|
|
|
it("a check answer tucks the explanation like every question (owner, 2026-10-07)", () => {
|
|
const answer = buildAnswer(quizQD(), { values: ["Venus"] });
|
|
const html = renderToStaticMarkup(<AnsweredCard answer={answer} />);
|
|
assert.ok(html.includes("tucked-explanation"), "the prediction tuck is now universal");
|
|
assert.ok(!html.includes("Why:"), "no inline explanation — behind the toggle");
|
|
});
|
|
|
|
it("the reading-thinking channel is LIFTED above the card as its own user-side block", () => {
|
|
const withReading = buildAnswer(quizQD(), {
|
|
values: ["Mercury"],
|
|
thinking: "answering thought",
|
|
readingThinking: "was unsure mid-paragraph",
|
|
});
|
|
const html = renderToStaticMarkup(<AnsweredCard answer={withReading} />);
|
|
// The unified ThinkingBlock (owner round, 2026-10-05; round two
|
|
// 2026-10-07): the reading thoughts render BEFORE the card as the
|
|
// banded user-thinking row, arriving EXPANDED (clearly sent).
|
|
const blockAt = html.indexOf("reading-thinking-block");
|
|
const cardAt = html.indexOf("answered-card");
|
|
assert.ok(blockAt !== -1, "the lifted block renders");
|
|
assert.ok(blockAt < cardAt, "it sits ABOVE the answered card");
|
|
assert.ok(html.includes("YOUR THINKING"), "the plain label");
|
|
assert.ok(!html.includes("Your thinking while reading"), "the old verbose label is gone");
|
|
assert.ok(!html.includes("your-reading-thinking"), "the in-card inset is gone");
|
|
// EXPANDED ARRIVAL (owner round two, 2026-10-07): transcript rows
|
|
// render their text; the in-card answer thinking stays collapsed.
|
|
assert.ok(html.includes('data-testid="your-thinking-trigger"'), "the answer thinking stays inside the card");
|
|
assert.ok(html.includes("was unsure mid-paragraph"), "the lifted row arrives expanded");
|
|
assert.ok(!html.includes("answering thought"), "the in-card row stays collapsed");
|
|
const legacy = { ...buildAnswer(quizQD(), { values: ["Mercury"] }), note: "a 2026-09 answer" };
|
|
const legacyHtml = renderToStaticMarkup(<AnsweredCard answer={legacy} />);
|
|
assert.ok(legacyHtml.includes("a 2026-09 answer") || legacyHtml.includes('data-testid="your-thinking-trigger"'), "records from before the rename still render");
|
|
});
|
|
});
|
|
|
|
// ─── details + option descriptions render as markdown (owner, 2026-10-12) ──
|
|
|
|
describe("question context renders as markdown everywhere (owner, 2026-10-12)", () => {
|
|
it("pending dock: ask option descriptions render LaTeX, not plain text", () => {
|
|
const html = renderDock({
|
|
toolName: "ask_user_question",
|
|
toolCallId: "d1",
|
|
input: {
|
|
question: "Which tool?",
|
|
options: [
|
|
{ label: "Pencil", description: "Wood around $\\text{Pb}_x\\text{C}$ graphite — the classic $x^2$ choice." },
|
|
{ label: "Pen" },
|
|
],
|
|
} as never,
|
|
});
|
|
// The description rides the same MessageResponse path as the label:
|
|
// KaTeX spans render inside the option subtext.
|
|
const katexCount = (html.match(/class="katex"/g) ?? []).length;
|
|
assert.ok(katexCount >= 2, `option description math renders (got ${katexCount})`);
|
|
assert.ok(!html.includes("$\\text"), "raw dollar delimiters do not leak as visible text");
|
|
});
|
|
|
|
it("settled ask card keeps the details line and option descriptions — retroactively from the persisted call", () => {
|
|
const input = {
|
|
question: "Which destination?",
|
|
details: "Inline $a^2+b^2=c^2$ and display:\n$$\nE = mc^2\n$$",
|
|
options: [
|
|
{ label: "Classical", description: "Entropy $H(X)$ first." },
|
|
{ label: "ML toolkit" },
|
|
],
|
|
};
|
|
const qd: QuestionData = {
|
|
id: "d2",
|
|
kind: "ask",
|
|
question: input.question,
|
|
mode: "single-select",
|
|
options: normalizeOptions(input.options),
|
|
createdAt: new Date().toISOString(),
|
|
};
|
|
const answer = buildAnswer(qd, { values: ["Classical"] });
|
|
const html = renderToStaticMarkup(<AnsweredCard answer={answer} input={input} />);
|
|
const katexCount = (html.match(/class="katex"/g) ?? []).length;
|
|
assert.ok(katexCount >= 3, `details (inline+display) + description math render (got ${katexCount})`);
|
|
assert.match(html, /Entropy/, "the option description text is present");
|
|
assert.ok(html.includes("katex-display"), "display math in the details line renders");
|
|
});
|
|
|
|
it("settled quiz card keeps the details line; quiz options stay description-less", () => {
|
|
const input = {
|
|
question: "Typical sets?",
|
|
details: "Recall $2^{nH(X)}$ — the asymptotic equipartition bound.",
|
|
options: [
|
|
{ label: "$\\approx 2^{nH(X)}$", value: "a" },
|
|
{ label: "$nH(X)$", value: "b" },
|
|
],
|
|
explanation: "AEP.",
|
|
correctAnswer: "a",
|
|
};
|
|
const options = normalizeOptions(input.options);
|
|
const qd: QuestionData = {
|
|
id: "d3",
|
|
kind: "quiz",
|
|
question: input.question,
|
|
mode: "single-select",
|
|
options,
|
|
correctIndices: resolveCorrect(input.correctAnswer, options).indices,
|
|
explanation: input.explanation,
|
|
createdAt: new Date().toISOString(),
|
|
};
|
|
const answer = buildAnswer(qd, { values: ["a"] });
|
|
const html = renderToStaticMarkup(<AnsweredCard answer={answer} input={input} />);
|
|
const katexCount = (html.match(/class="katex"/g) ?? []).length;
|
|
assert.ok(katexCount >= 3, `details + option-label math renders (got ${katexCount})`);
|
|
// Quiz options carry NO description field at all (owner ruling
|
|
// 2026-09-26) — the desc map is ask-only and stays empty here.
|
|
assert.ok(!html.includes("lh-prose block text-xs"), "no option-subtext slot on quiz rows");
|
|
});
|
|
|
|
it("legacy settled records (no input prop) render exactly as before", () => {
|
|
const answer = buildAnswer(askQD(), { values: ["Very"] });
|
|
const html = renderToStaticMarkup(<AnsweredCard answer={answer} />);
|
|
assert.match(html, /Very/, "the card renders");
|
|
assert.equal((html.match(/class="katex"/g) ?? []).length, 0, "no context slot appears from nothing");
|
|
assert.ok(!html.includes("lh-prose block text-xs"), "no description slot without an ask input");
|
|
});
|
|
});
|