Files
learnharness/tests/render-cards.test.tsx
T
amosblomqvist 4ba594bab5 fix(arch): parse, don't cast — the UI parse seam for model-authored input
The architectural answer to the validation-crash question: validation IS
first at the model-facing boundary (pi validates arguments before
execute; failures return to the model same-turn — all three live
incidents retried successfully). But the same bytes also exist UNVALIDATED
by design in two other places: the live wire streams raw tool-input
deltas DURING generation, and the transcript keeps the raw input of
REJECTED calls (honesty + the replay invariant). The consumers were
supposed to trust part STATES — but that trust was enforced by
TypeScript casts ('part.input as X'), which are compile-time lies with
no runtime check behind them. That is the hole both crashes came
through, and it should not have been possible.

The structural rule now: model-authored INPUT is consumed only through
parsers running the SAME zod schema the model faced — boundary-valid ⇔
consumer-parseable, mechanically. Server/learner-authored OUTPUT stays
trusted (our code authors it).

- src/core/tool-parts.ts (new): parseQuizInput (display parse over the
  shared quizInputBase — strips the teacher-only nodeId, accepts both
  prober and teacher boundary shapes), parseAskInput, parseProbeDoneInput.
- contracts/probe.ts: probeDoneInputSchema is now core-owned zod — the
  probe_done tool declares it (single source of truth), the seam runs
  the literal same schema.
- question-dock.tsx: the dock parses its input at entry; unparsable
  renders the quiet malformed-question fallback, never a crash.
- probe-done-card.tsx: the input preview goes through the parse; the
  old state reasoning and shape guards collapse into it.

Pinned in tests/tool-parts.test.ts: both exact live crash shapes
(stringified quiz options, stringified assessment), boundary-parity
cases for all three schemas, nodeId stripping, mid-stream partials;
render-cards pins the dock fallback markup.

Gates: tsc, lint, build, 328/328.
2026-09-12 04:51:39 +03:00

392 lines
17 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 { AnsweredCard } from "../components/answered-card.tsx";
import { buildAnswer, coerceCorrectAnswer, normalizeOptions, resolveCorrect, 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(coerceCorrectAnswer(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" | "ask_user_question"; 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("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&#x27;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");
assert.match(html, /Why this answer\? What are you unsure about\?/, "placeholder invites, not apologizes");
assert.ok(!html.includes("optional"), "the field is never called optional");
assert.match(html, /Submit answer/, "send becomes Submit answer");
assert.match(html, /⌘\/Ctrl \+ ↵ submits/, "the ⌘/Ctrl+Enter shortcut is discoverable");
// 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&#x27;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");
});
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/);
assert.match(html, /innermost/, "explanation renders post-answer");
assert.ok(html.includes("✓") && html.includes("✗"));
});
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"],
note: "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 block exists");
assert.match(html, /Your thinking/, "labeled as the learner's voice");
assert.match(html, /brightness isn&#x27;t distance/, "the thinking itself renders");
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("innermost"),
"thinking before the explanation",
);
});
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("✗"), "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)", () => {
// New delimiter contract (W3): the teacher writes \(...\) / \[...\]; a
// bare $ is ALWAYS literal currency (singleDollarTextMath: false kills it
// structurally). transformMathDelimiters rewrites both to the two-dollar
// forms remark-math still parses. Full coverage: render-markdown.test.tsx.
it("renders $...$ as LITERAL text — a $ is never math (the currency fix)", () => {
const html = renderToStaticMarkup(<MessageResponse>{"The map $f(x) = x^2$ is smooth."}</MessageResponse>);
assert.ok(!html.includes("katex"), `$-math must not render, got: ${html.slice(0, 300)}`);
assert.match(html, /\$f\(x\) = x\^2\$/, "old-contract math renders verbatim");
});
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 — belt and suspenders)", () => {
// With singleDollarTextMath: false even a bare $5 can't parse as math;
// an escaped \$ is doubly safe. The prompt contract still asks the
// teacher to prefer plain "$5" (it renders literally now).
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(coerceCorrectAnswer(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 from the seam ack", () => {
const html = renderToStaticMarkup(
<ProbeDoneCard
part={{ state: "output-available", output: { ok: true, assessment: validAssessment } } as never}
/>,
);
assert.ok(html.includes('data-testid="probe-done-card"'));
assert.match(html, /modal logic basics/);
assert.match(html, /indexical &#x27;actual&#x27;|indexical 'actual'/);
assert.match(html, /Bracketed/);
});
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/);
});
});