Files
learning-platform/src/lib/llmSchemas.js
RaymondVerhoef 66e0c275da
All checks were successful
On Pull Request to Main / test (pull_request) Successful in 31s
On Pull Request to Main / publish (pull_request) Successful in 1m1s
On Pull Request to Main / deploy-dev (pull_request) Successful in 1m31s
feat: phase 4 of AI pipeline hardening — quiz & content quality
- src/lib/random.js: Fisher–Yates shuffle/sample/pickInt; replace every
  biased .sort(() => 0.5 - Math.random()) site in testService.
- testService: debias correctIndex via prompt + runtime re-roll (up to 2x
  when one position holds >50%); quality gate rejecting <4 distinct
  options, banned filler ("all of the above" etc) and explanations
  shorter than 20 chars; dedup new questions against the existing bank
  via normalised question text.
- Quiz schema/tool/prompt require difficulty ('easy'|'medium'|'hard');
  db.getQuizBank defaults legacy records to 'medium' on read.
- learningService.generateCustomTopic: kebab-case slug ID from the
  polished label with collision suffixes; default learning_relevance
  'standard' when the model omits it.
- Tests for random helpers, dedup/quality-gate behaviour and the
  extended quiz schema.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-20 19:22:10 +02:00

236 lines
6.6 KiB
JavaScript

/**
* Zod schemas for every structured LLM output the platform consumes.
*
* Field names mirror what callers already produce — do not rename them
* without migrating the corresponding service module.
*/
import { z } from 'zod';
const topicTypeEnum = z.enum(['concept', 'role', 'process']);
const relationTypeStrict = z.enum(['related_to', 'depends_on', 'part_of', 'executed_by']);
const relationTypeLoose = z.enum(['related_to', 'depends_on', 'part_of', 'executed_by', 'executes']);
const learningRelevanceEnum = z.enum(['core', 'standard', 'peripheral', 'exclude']);
const extractionTopicSchema = z.object({
id: z.string().min(1),
label: z.string().min(1),
type: topicTypeEnum,
description: z.string().min(1),
learning_relevance: learningRelevanceEnum,
});
const extractionRelationSchema = z.object({
source: z.string().min(1),
target: z.string().min(1),
type: relationTypeStrict,
});
export const extractionResultSchema = z.object({
topics: z.array(extractionTopicSchema),
relations: z.array(extractionRelationSchema),
});
const handbookTopicSchema = extractionTopicSchema.extend({
metadata: z.object({ source: z.string() }).optional(),
});
const handbookRelationSchema = z.object({
source: z.string().min(1),
target: z.string().min(1),
type: relationTypeLoose,
description: z.string().optional(),
});
export const handbookResultSchema = z.object({
topics: z.array(handbookTopicSchema),
relations: z.array(handbookRelationSchema),
});
/**
* Normalise legacy `executes` relations into the canonical `executed_by`
* vocabulary by swapping source and target. The handbook prompt previously
* emitted `role → executes → process`; the canonical form is
* `process → executed_by → role`.
*/
export function normalizeHandbookResult(parsed) {
return {
...parsed,
relations: parsed.relations.map((r) =>
r.type === 'executes'
? { ...r, type: 'executed_by', source: r.target, target: r.source }
: r,
),
};
}
const articleSectionSchema = z.object({
heading: z.string().min(1),
body: z.string().min(1),
});
const articleBodySchema = z.object({
title: z.string().min(1),
intro: z.string().min(1),
sections: z.array(articleSectionSchema).min(1),
keyTakeaways: z.array(z.string().min(1)).min(1),
});
export const learningArticleSchema = z.object({
article: articleBodySchema,
});
const slideSchema = z.object({
title: z.string().min(1),
bullets: z.array(z.string().min(1)).min(1),
speakerNote: z.string().min(1),
});
export const learningSlidesSchema = z.object({
slides: z.array(slideSchema).min(1),
});
const infographicStatSchema = z.object({
value: z.string().min(1),
label: z.string().min(1),
icon: z.string().min(1),
});
const infographicStepSchema = z.object({
number: z.number().int().min(1),
title: z.string().min(1),
description: z.string().min(1),
icon: z.string().min(1),
});
const infographicBodySchema = z.object({
headline: z.string().min(1),
tagline: z.string().min(1),
stats: z.array(infographicStatSchema).min(1),
steps: z.array(infographicStepSchema).min(1),
quote: z.string().min(1),
colorTheme: z.string().min(1),
});
export const learningInfographicSchema = z.object({
infographic: infographicBodySchema,
});
export const learningAllSchema = z.object({
article: articleBodySchema,
slides: z.array(slideSchema).min(1),
infographic: infographicBodySchema,
});
const quizDifficultyEnum = z.enum(['easy', 'medium', 'hard']);
const quizQuestionSchema = z.object({
id: z.string().min(1),
question: z.string().min(1),
topicLabel: z.string().min(1),
options: z.array(z.string().min(1)).length(4),
correctIndex: z.number().int().min(0).max(3),
explanation: z.string().min(1),
difficulty: quizDifficultyEnum,
});
export const quizQuestionsSchema = z.object({
questions: z.array(quizQuestionSchema).min(1),
});
export const customTopicSchema = z.object({
label: z.string().min(1),
type: topicTypeEnum,
description: z.string().min(1),
});
const mergeActionSchema = z.object({
keepId: z.string().min(1),
deleteId: z.string().min(1),
});
const newRelationSchema = z.object({
source: z.string().min(1),
target: z.string().min(1),
type: relationTypeStrict,
});
const relevanceUpdateSchema = z.object({
id: z.string().min(1),
learning_relevance: learningRelevanceEnum,
});
export const graphActionsSchema = z.object({
merges: z.array(mergeActionSchema).optional().default([]),
deletions: z.array(z.string().min(1)).optional().default([]),
newRelations: z.array(newRelationSchema).optional().default([]),
relevanceUpdates: z.array(relevanceUpdateSchema).optional().default([]),
});
const deltaTopicSchema = z.object({
id: z.string().min(1),
label: z.string().min(1),
type: topicTypeEnum,
description: z.string().min(1),
});
const deltaRelationSchema = z.object({
source: z.string().min(1),
target: z.string().min(1),
type: relationTypeStrict,
});
export const proposeGraphDeltaSchema = z.object({
reason: z.string().min(1),
topics: z.array(deltaTopicSchema).max(3).optional(),
relations: z.array(deltaRelationSchema).max(5).optional(),
});
// ── Article patch operation schemas (Phase 2.4) ──────────────────────────────
export const setIntroPatchSchema = z.object({
intro: z.string().min(1),
});
export const setSectionPatchSchema = z.object({
heading: z.string().min(1),
body: z.string().min(1),
});
export const addSectionPatchSchema = z.object({
heading: z.string().min(1),
body: z.string().min(1),
position: z.enum(['start', 'end']),
});
export const removeSectionPatchSchema = z.object({
heading: z.string().min(1),
});
export const replaceTakeawaysPatchSchema = z.object({
items: z.array(z.string().min(1)).min(1),
});
/**
* Registry mapping known tool names to their input schemas. `callLLM`
* consults this when the caller does not pass an explicit `toolSchemas`
* override.
*/
export const toolSchemaRegistry = {
emit_knowledge_graph: extractionResultSchema,
emit_handbook_delta: handbookResultSchema,
emit_learning_article: learningArticleSchema,
emit_learning_slides: learningSlidesSchema,
emit_learning_infographic: learningInfographicSchema,
emit_learning_all: learningAllSchema,
emit_quiz_questions: quizQuestionsSchema,
emit_custom_topic: customTopicSchema,
emit_graph_actions: graphActionsSchema,
propose_graph_delta: proposeGraphDeltaSchema,
set_intro: setIntroPatchSchema,
set_section: setSectionPatchSchema,
add_section: addSectionPatchSchema,
remove_section: removeSectionPatchSchema,
replace_takeaways: replaceTakeawaysPatchSchema,
};