tool-result diet: question answers return only learner-added truth (compact JSON for the model, full record in details); dead-code sweep
This commit is contained in:
amosblomqvist
+48
-105
@@ -11,16 +11,12 @@ import {
|
||||
quizInputBase,
|
||||
askOptionProblems,
|
||||
buildAnswer,
|
||||
compactAnswerForModel,
|
||||
coerceCorrectAnswer,
|
||||
CONFIDENCE_STOPS,
|
||||
DEFAULT_CONFIDENCE_INDEX,
|
||||
findPendingQuestion,
|
||||
formatAnswerMessage,
|
||||
isCorrect,
|
||||
makeId,
|
||||
normalizeOptions,
|
||||
QUESTION_ANSWERED_ENTRY,
|
||||
QUESTION_ENTRY,
|
||||
quizInputProblems,
|
||||
explainInputBase,
|
||||
explainInputObject,
|
||||
@@ -415,47 +411,50 @@ describe("buildAnswer", () => {
|
||||
});
|
||||
});
|
||||
|
||||
describe("formatAnswerMessage", () => {
|
||||
it("formats a correct quiz answer for the agent", () => {
|
||||
const ad = buildAnswer(quizData(), { values: ["mercury"] });
|
||||
const text = formatAnswerMessage(ad);
|
||||
assert.ok(text.startsWith("[Quiz answer]"));
|
||||
assert.match(text, /Selected: 1\. Mercury/);
|
||||
assert.match(text, /Result: CORRECT/);
|
||||
|
||||
// ─── compactAnswerForModel (the model-visible diet, owner 2026-10-02) ──────
|
||||
|
||||
describe("compactAnswerForModel", () => {
|
||||
it("quiz: the learner's selection, verdict, confidence, note — NOTHING the model authored", () => {
|
||||
const ad = buildAnswer(quizData(), { values: ["mercury"], confidence: 70, note: "guessed" });
|
||||
const compact = compactAnswerForModel(ad);
|
||||
assert.deepEqual(compact, {
|
||||
kind: "quiz",
|
||||
answers: [{ index: 1, label: "Mercury" }],
|
||||
correct: true,
|
||||
confidence: 70,
|
||||
note: "guessed",
|
||||
});
|
||||
const raw = JSON.stringify(compact);
|
||||
// every one of these lives in the tool CALL's arguments already
|
||||
assert.ok(!raw.includes("Which planet"), "no questionText");
|
||||
assert.ok(!raw.includes("Venus"), "no option-list restatement");
|
||||
assert.ok(!raw.includes("Mercury orbits closest"), "no explanation");
|
||||
assert.ok(!("correctIndices" in compact), "correct is the verdict; indices are the call's");
|
||||
assert.ok(!("questionId" in compact), "questionId was the toolCallId restated");
|
||||
});
|
||||
|
||||
it("formats dontKnow as a gap, not a guess", () => {
|
||||
const ad = buildAnswer(quizData(), { dontKnow: true });
|
||||
const text = formatAnswerMessage(ad);
|
||||
assert.match(text, /did NOT attempt/);
|
||||
assert.match(text, /knowledge gap/);
|
||||
assert.ok(!text.includes("Result:"));
|
||||
it("dontKnow: the gap flag stands alone — no verdict, no confidence", () => {
|
||||
const compact = compactAnswerForModel(buildAnswer(quizData(), { dontKnow: true }));
|
||||
assert.deepEqual(compact, { kind: "quiz", dontKnow: true });
|
||||
});
|
||||
|
||||
it("includes the note when present", () => {
|
||||
const ad = buildAnswer(quizData(), { values: ["venus"], note: "guessed" });
|
||||
assert.match(formatAnswerMessage(ad), /Note: guessed/);
|
||||
it("ask: selection or text; never graded", () => {
|
||||
const askQD: QuestionData = {
|
||||
...quizData(),
|
||||
kind: "ask",
|
||||
mode: "text",
|
||||
correctIndices: undefined,
|
||||
explanation: undefined,
|
||||
};
|
||||
const compact = compactAnswerForModel(buildAnswer(askQD, { text: "lateral thinking" }));
|
||||
assert.deepEqual(compact, { kind: "ask", text: "lateral thinking" });
|
||||
});
|
||||
|
||||
it("states the confidence line on answered quizzes, absent on dontKnow", () => {
|
||||
const ad = buildAnswer(quizData(), { values: ["mercury"], confidence: 70 });
|
||||
assert.match(formatAnswerMessage(ad), /Confidence: 70% — the learner's stated probability that his selected answer is correct\./);
|
||||
const idk = buildAnswer(quizData(), { dontKnow: true, confidence: 70 });
|
||||
assert.ok(!formatAnswerMessage(idk).includes("Confidence:"));
|
||||
});
|
||||
|
||||
it("formats ask answers", () => {
|
||||
const ask = quizData({ kind: "ask", correctIndices: undefined, explanation: undefined });
|
||||
const ad = buildAnswer(ask, { other: "whatever" });
|
||||
const text = formatAnswerMessage(ad);
|
||||
assert.ok(text.startsWith("[Answer to your question]"));
|
||||
assert.match(text, /\(Other\): whatever/);
|
||||
});
|
||||
|
||||
it("formats a dismissed question", () => {
|
||||
const ad = buildAnswer(quizData(), {});
|
||||
ad.dismissed = true;
|
||||
assert.match(formatAnswerMessage(ad), /dismissed/);
|
||||
it("dismissed: a one-field record", () => {
|
||||
const ad = { ...buildAnswer(quizData(), { dontKnow: true }), dismissed: true };
|
||||
const compact = compactAnswerForModel(ad);
|
||||
assert.deepEqual(compact, { kind: "quiz", dismissed: true });
|
||||
});
|
||||
});
|
||||
|
||||
@@ -473,64 +472,6 @@ describe("toClientQuestion", () => {
|
||||
});
|
||||
});
|
||||
|
||||
// ─── findPendingQuestion (resume re-arm) ────────────────────────────────────
|
||||
|
||||
describe("findPendingQuestion", () => {
|
||||
it("finds an unanswered question", () => {
|
||||
const entries = [
|
||||
{ type: "custom", customType: QUESTION_ENTRY, data: quizData() },
|
||||
];
|
||||
const pq = findPendingQuestion(entries);
|
||||
assert.ok(pq);
|
||||
assert.equal(pq.id, "q_test");
|
||||
});
|
||||
|
||||
it("returns null once answered", () => {
|
||||
const entries = [
|
||||
{ type: "custom", customType: QUESTION_ENTRY, data: quizData() },
|
||||
{
|
||||
type: "custom",
|
||||
customType: QUESTION_ANSWERED_ENTRY,
|
||||
data: { questionId: "q_test", dontKnow: true, answers: [] },
|
||||
},
|
||||
];
|
||||
assert.equal(findPendingQuestion(entries), null);
|
||||
});
|
||||
|
||||
it("returns the OLDEST unanswered question", () => {
|
||||
const entries = [
|
||||
{ type: "custom", customType: QUESTION_ENTRY, data: quizData({ id: "q1" }) },
|
||||
{ type: "custom", customType: QUESTION_ENTRY, data: quizData({ id: "q2" }) },
|
||||
];
|
||||
assert.equal(findPendingQuestion(entries)?.id, "q1");
|
||||
});
|
||||
|
||||
it("skips a dismissed question", () => {
|
||||
const entries = [
|
||||
{ type: "custom", customType: QUESTION_ENTRY, data: quizData() },
|
||||
{
|
||||
type: "custom",
|
||||
customType: QUESTION_ANSWERED_ENTRY,
|
||||
data: { questionId: "q_test", dismissed: true, answers: [] },
|
||||
},
|
||||
];
|
||||
assert.equal(findPendingQuestion(entries), null);
|
||||
});
|
||||
|
||||
it("ignores other custom entries", () => {
|
||||
const entries = [{ type: "custom", customType: "md-log", data: { file: null } }];
|
||||
assert.equal(findPendingQuestion(entries), null);
|
||||
});
|
||||
});
|
||||
|
||||
// ─── makeId ─────────────────────────────────────────────────────────────────
|
||||
|
||||
describe("makeId", () => {
|
||||
it("generates unique ids", () => {
|
||||
const ids = new Set(Array.from({ length: 200 }, () => makeId()));
|
||||
assert.equal(ids.size, 200);
|
||||
});
|
||||
});
|
||||
|
||||
// ─── the dock selection state machine (multi-select Other coexistence) ──
|
||||
// The live bug: in a multi-select ask, picking the custom "Other…" row
|
||||
@@ -630,13 +571,15 @@ describe("buildAnswer / formatAnswerMessage: explain", () => {
|
||||
assert.equal(ad.text, undefined);
|
||||
});
|
||||
|
||||
it("formats the model's view: his answer, the immutable commitment, and the grading instruction", () => {
|
||||
const text = formatAnswerMessage(buildAnswer(explainQD(), { text: "No idea about aether." }));
|
||||
assert.ok(text.startsWith("[Explain answer]"));
|
||||
assert.match(text, /The learner answered: No idea about aether\./);
|
||||
assert.match(text, /Your committed correct answer \(immutable/);
|
||||
assert.match(text, /luminiferous aether/);
|
||||
assert.match(text, /Grade his answer strictly/);
|
||||
it("compactAnswerForModel (explain): his words ONLY — never the model's own commitment", () => {
|
||||
const compact = compactAnswerForModel(buildAnswer(explainQD(), { text: "No idea about aether." }));
|
||||
const raw = JSON.stringify(compact);
|
||||
assert.equal(compact.text, "No idea about aether.");
|
||||
assert.equal(compact.kind, "explain");
|
||||
// the commitment is the model's own call argument — NEVER restated
|
||||
assert.ok(!raw.includes("luminiferous"), "the commitment must not ride the result");
|
||||
assert.ok(!("committed" in compact));
|
||||
assert.ok(!("questionText" in compact));
|
||||
});
|
||||
|
||||
});
|
||||
|
||||
Reference in New Issue
Block a user