Self-hosted, Self-Improving, Federated Everything App — dev management, agent orchestration, persistent mind, digital twin, second brain, and creative pipeline from a single dashboard over Tailscale to port the user into the OS
JavaScript ai-agentsdashboarddeveloper-toolsdigital-twinquantified-selfsecond-brainself-hostedtailscale
/**
* Tests for the storytelling-craft rubric. The prompt builder and the
* normalizer sit either side of an LLM round trip, so the assertions here are
* about what the rubric GUARANTEES regardless of what the model returns: every
* move and CART stage present in canonical order, scores bounded, and a
* non-object payload rejected rather than persisted as an all-zero score.
*/
import { describe, it, expect } from 'vitest';
import {
STORY_CRAFT_MOVES,
STORY_CRAFT_MOVE_IDS,
STORY_CRAFT_MAX_SCORE,
CART_STAGE_IDS,
buildStoryCraftEvaluationPrompt,
normalizeStoryCraftEvaluation
} from './storytellingCraft.js';
const fullAnswer = (scores) => ({
moves: Object.fromEntries(
STORY_CRAFT_MOVE_IDS.map((id, i) => [id, {
score: scores[i],
evidence: ` evidence ${id} `,
suggestion: `suggestion ${id}`
}])
),
cart: Object.fromEntries(CART_STAGE_IDS.map(id => [id, { present: true, note: `note ${id}` }])),
answersQuestion: true,
revision: ' tighten the opening '
});
describe('buildStoryCraftEvaluationPrompt', () => {
it('names every move id so the model answers in the keys the normalizer reads', () => {
const prompt = buildStoryCraftEvaluationPrompt({ question: 'Why do I like black licorice?', story: 'A story.' });
for (const id of STORY_CRAFT_MOVE_IDS) {
expect(prompt).toContain(id);
}
for (const id of CART_STAGE_IDS) {
expect(prompt).toContain(id);
}
});
it('includes the question and the story text', () => {
const prompt = buildStoryCraftEvaluationPrompt({
question: 'Why do I like black licorice?',
story: 'It started at my grandmother’s kitchen table.'
});
expect(prompt).toContain('Why do I like black licorice?');
expect(prompt).toContain('It started at my grandmother’s kitchen table.');
});
it('tells the model to infer the implied question when none was given', () => {
const prompt = buildStoryCraftEvaluationPrompt({ question: '', story: 'A story.' });
expect(prompt).toMatch(/infer the implied question/i);
});
});
describe('normalizeStoryCraftEvaluation', () => {
it('rejects a non-object payload rather than scoring it zero', () => {
expect(normalizeStoryCraftEvaluation(null)).toBeNull();
expect(normalizeStoryCraftEvaluation('not json')).toBeNull();
expect(normalizeStoryCraftEvaluation([1, 2, 3])).toBeNull();
});
it('returns every move and CART stage in canonical order', () => {
const result = normalizeStoryCraftEvaluation(fullAnswer([5, 4, 3, 2, 1, 0, 5]));
expect(result.moves.map(m => m.id)).toEqual([...STORY_CRAFT_MOVE_IDS]);
expect(result.cart.map(s => s.id)).toEqual([...CART_STAGE_IDS]);
expect(result.moves[0].label).toBe(STORY_CRAFT_MOVES[0].label);
});
it('credits a move the model omitted as zero instead of dropping the row', () => {
const partial = fullAnswer([5, 5, 5, 5, 5, 5, 5]);
delete partial.moves.takeaway;
const result = normalizeStoryCraftEvaluation(partial);
expect(result.moves).toHaveLength(STORY_CRAFT_MOVE_IDS.length);
const takeaway = result.moves.find(m => m.id === 'takeaway');
expect(takeaway.score).toBe(0);
expect(takeaway.evidence).toBe('');
});
it('bounds a score the model returned out of range', () => {
const wild = fullAnswer([99, -4, 3, 3, 3, 3, 3]);
const result = normalizeStoryCraftEvaluation(wild);
expect(result.moves[0].score).toBe(STORY_CRAFT_MAX_SCORE);
expect(result.moves[1].score).toBe(0);
});
it('treats a non-numeric score as zero rather than NaN', () => {
const wild = fullAnswer(['great', 3, 3, 3, 3, 3, 3]);
expect(normalizeStoryCraftEvaluation(wild).moves[0].score).toBe(0);
});
it('averages the move scores to one decimal', () => {
// 5+4+3+2+1+0+5 = 20 over 7 moves = 2.857… → 2.9
const result = normalizeStoryCraftEvaluation(fullAnswer([5, 4, 3, 2, 1, 0, 5]));
expect(result.overallScore).toBe(2.9);
expect(result.maxScore).toBe(STORY_CRAFT_MAX_SCORE);
});
it('persists only what the UI reads — no derived strongest/weakest fields', () => {
const result = normalizeStoryCraftEvaluation(fullAnswer([5, 5, 1, 1, 3, 3, 3]));
// This object is written into every story, so a derived field shipped once
// becomes a compatibility surface. "Your weakest move" is one reduce over
// `moves` at the point of display.
expect(Object.keys(result).sort()).toEqual(
['answersQuestion', 'cart', 'maxScore', 'moves', 'overallScore', 'revision']
);
});
it('trims prose, and reports an omitted answersQuestion as not assessed', () => {
const answer = fullAnswer([3, 3, 3, 3, 3, 3, 3]);
delete answer.answersQuestion;
const result = normalizeStoryCraftEvaluation(answer);
expect(result.revision).toBe('tighten the opening');
expect(result.moves[0].evidence).toBe(`evidence ${STORY_CRAFT_MOVE_IDS[0]}`);
// null, not true: a truncated response must not CLAIM the story answered
// its question, and must not ACCUSE it of failing to either. The UI shows
// its warning only on an explicit false.
expect(result.answersQuestion).toBeNull();
});
it('carries an explicit answersQuestion:true through', () => {
expect(normalizeStoryCraftEvaluation(fullAnswer([3, 3, 3, 3, 3, 3, 3])).answersQuestion).toBe(true);
});
it('carries an explicit answersQuestion:false through', () => {
const answer = { ...fullAnswer([3, 3, 3, 3, 3, 3, 3]), answersQuestion: false };
expect(normalizeStoryCraftEvaluation(answer).answersQuestion).toBe(false);
});
it('treats a CART stage as absent unless the model said true', () => {
const answer = fullAnswer([3, 3, 3, 3, 3, 3, 3]);
answer.cart.takeaway = { present: 'yes', note: 'ambiguous' };
delete answer.cart.result;
const result = normalizeStoryCraftEvaluation(answer);
expect(result.cart.find(s => s.id === 'takeaway').present).toBe(false);
expect(result.cart.find(s => s.id === 'result').present).toBe(false);
expect(result.cart.find(s => s.id === 'context').present).toBe(true);
});
});