Files
learnharness/tests/tool-toggles-bridge.test.ts
amosblomqvist 2f39766528 free-recall explain tool: the quizzing-mechanic switch (owner 2026-09-27)
Sycophancy mitigation as a structural commitment: the model writes its
full correctAnswer (plus explanation, in that order) BEFORE the learner
sees the question; the commitment rides the persisted call args, is
revealed on submit, and grades his free text in the model's next reply —
the client never grades free recall.

- tools.explain setting (segmented 'Quizzing — Multiple choice | Free
  recall' in Agent settings' Tools group): free arms explain and drops
  quiz for BOTH agents (prober variant carries no nodeId, same pattern
  as the probe quiz); mechanics keys never pass filterEnabledTools —
  quiz stays unwritable (the 2026-09-19 wedge ruling stands)
- doctrine lives pi-natively in the tool descriptions (owner ruling —
  no conditional system-prompt chunks): explain carries the commitment
  rule; the MC option-building craft moved out of the prompt into the
  quiz description; shared/20-question-tools.md slims to the
  quiz-agnostic core
- display seams: parseExplainInput (LaTeX-debris restoration included),
  dock text-mode with an 'I don't know' row (kbd row 0), AnsweredCard
  reveal (The answer / Why, no verdict chips), [Explain answer] format
- phases/plan seams recognize both quizzing tools' nodeId tags; explain
  answers deliberately feed no ledger evidence (no client grading to
  key on — a future ruling can parse the model's grade)

1049/1049 tests, typecheck/lint/build green.
2026-09-27 08:55:13 +03:00

104 lines
4.3 KiB
TypeScript

/**
* The P2 tool-toggle bridge (2026-09-19): AgentBuildOptions.toolToggles is
* first-class — the agent constructors thread it to the factories' enabled?
* param (an explicit opts.tools registry is never filtered), and the
* settings write side rejects unknown tool keys against TOGGLEABLE_TOOLS
* (the definitive adjudicated list — a persisted `quiz: false` would
* genuinely drop the quiz once the bridge honors the delta).
*
* Parser tests are PURE (no fs — parallel-test discipline); the agent
* construction tests use an in-memory chat fixture like
* question-tools.test.ts.
*/
import assert from "node:assert/strict";
import { z } from "zod";
import { describe, it } from "node:test";
import { Type } from "typebox";
import { questionResult } from "../src/agent/tool-execute.ts";
import { createProberAgent, createTeacherAgent } from "../src/agent/teacher-agent.ts";
import type { ProberToolContext } from "../src/agent/tools.ts";
import { jobStore } from "../src/store/jobs.ts";
import { parseAgentSettingsBody } from "../src/server/settings-body.ts";
import { chatStore } from "../src/store/chats.ts";
import { createFauxCore } from "@earendil-works/pi-ai";
import type { PiEngine } from "../src/agent/models.ts";
/** A scripted pi engine — the agent builds need an injected engine since
* BYOK removed the factory pins (owner, 2026-09-22). */
function fauxEngine(): PiEngine {
const faux = createFauxCore({});
return { model: faux.getModel(), streamFn: faux.streamSimple };
}
import type { AgentSettings } from "../src/core/settings.ts";
describe("AgentBuildOptions.toolToggles (the P2 bridge)", () => {
const toolNames = (agent: { state: { tools: Array<{ name: string }> } }) =>
agent.state.tools.map((t) => t.name).sort();
it("flows into the teacher's registry: a job-tool delta is a no-op on the bare set", () => {
// No ctx → bare agent: quiz/ask only. A job-tool delta names tools
// that don't exist there — the bridge must not crash on it.
const bare = createTeacherAgent(undefined, { engine: fauxEngine(), toolToggles: { research: false } });
assert.deepEqual(toolNames(bare), ["ask_user_question", "quiz"]);
});
it("flows into the prober's registry; probe_done survives everything", () => {
const chatId = chatStore.createChat("Toggle bridge test").id;
const ctx: ProberToolContext = { chatId, chatStore, deliver: () => {}, jobStore, spawn: () => {} };
const prober = createProberAgent(ctx, { engine: fauxEngine(), toolToggles: { ask_user_question: false } });
const names = toolNames(prober);
assert.ok(!names.includes("ask_user_question"), "the delta dropped the tool");
assert.ok(names.includes("probe_done"), "probe_done always-on");
});
it("an explicitly injected opts.tools registry is NEVER toggle-filtered", () => {
const teacher = createTeacherAgent(undefined, {
engine: fauxEngine(),
tools: {
quiz: {
name: "quiz",
label: "quiz",
description: "test",
parameters: Type.Object({}),
execute: async () => questionResult(),
},
},
toolToggles: { quiz: false },
});
assert.deepEqual(toolNames(teacher), ["quiz"]);
});
});
describe("parseAgentSettingsBody: tool keys validate at write time", () => {
const parse = (tools: Record<string, boolean>) =>
parseAgentSettingsBody({ tools }) as
| { ok: true; settings: AgentSettings }
| { ok: false; error: string };
it("a valid job-tool delta parses", () => {
const r = parse({ research: false, pdf: false });
assert.equal(r.ok, true);
assert.deepEqual(r.ok && r.settings.tools, { research: false, pdf: false });
});
it("the quizzing-mechanics key (explain) is writable (owner 2026-09-27)", () => {
const r = parse({ explain: true });
assert.equal(r.ok, true);
assert.deepEqual(r.ok && r.settings.tools, { explain: true });
});
it("unknown keys are rejected with the supported list (the quiz wedge never persists)", () => {
const r = parse({ quiz: false });
assert.equal(r.ok, false);
assert.match(!r.ok && r.error, /unknown tool "quiz" \(supported: research, demo, visual, pdf, explain\)/);
const typo = parse({ reasearch: false });
assert.equal(typo.ok, false);
assert.match(!typo.ok && typo.error, /unknown tool "reasearch"/);
});
it("absent tools field is fine (absent = enabled everywhere)", () => {
const r = parseAgentSettingsBody({}) as { ok: boolean };
assert.equal(r.ok, true);
});
});