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
/**
* Model Personality validation schemas (issue #2610).
*
* Zod schemas for the LLM personality self-profile test: the structured
* self-evaluation response, the twin-alignment scorer response, and the
* route input/settings shapes. Kept a LEAF module (zod only — no imports
* from other PortOS modules) so mocked suites never drag transitive deps.
*
* The trait taxonomy is versioned: `PERSONALITY_TAXONOMY_VERSION` is stamped
* on every persisted result so a future taxonomy change can distinguish old
* records instead of silently comparing scores across different trait sets.
*/
import { z } from 'zod';
// Trait taxonomy v1 — the single source of truth for the self-profile JSON
// keys. The prompt template receives this list via `{{traitKeys}}` so the
// schema and the prompt can never drift apart.
export const PERSONALITY_TRAIT_KEYS = [
'agreeableness',
'humor',
'errorAversion',
'selfCensorship',
'conciseness',
'dogmatism',
'sycophancy',
'creativity',
'formality',
'empathy'
];
export const PERSONALITY_TAXONOMY_VERSION = 1;
const score01 = z.number().min(0).max(1);
// One self-reported trait: a 0–1 score plus a short self-observation.
export const personalityTraitEntrySchema = z.object({
score: score01,
rationale: z.string().max(2000).optional().default('')
});
// Call 1 (self-profile) response — every taxonomy key must be present so a
// partial answer retries through the runner's schema machinery instead of
// persisting a half-scored profile.
export const personalityProfileResponseSchema = z.object({
traits: z.object(
Object.fromEntries(PERSONALITY_TRAIT_KEYS.map((k) => [k, personalityTraitEntrySchema]))
),
summary: z.string().max(4000).optional().default('')
});
// Call 2 (alignment scorer) response — dimension keys are free-form (the
// scorer names whichever twin dimensions it found evidence for).
export const personalityAlignmentResponseSchema = z.object({
alignmentScore: score01,
dimensions: z
.record(
z.string(),
z.object({
score: score01,
note: z.string().max(1000).optional().default('')
})
)
.optional()
.default({})
});
// POST /api/model-personality/run
export const runPersonalityTestInputSchema = z.object({
providerId: z.string().min(1),
model: z.string().min(1).optional(),
includeAlignment: z.boolean().optional(),
personaId: z.string().min(1).nullable().optional()
});
// GET /api/model-personality/history
export const personalityHistoryQuerySchema = z.object({
limit: z.coerce.number().int().min(1).max(1000).optional()
});
// UI selects submit '' for "no scorer configured" — normalize to null (clear).
const emptyToNull = (v) => (v === '' ? null : v);
// PUT /api/model-personality/settings — partial update, no defaults so an
// omitted key never overwrites the stored value.
export const personalitySettingsUpdateSchema = z.object({
scorerProviderId: z.preprocess(emptyToNull, z.string().min(1).nullable()).optional(),
scorerModel: z.preprocess(emptyToNull, z.string().min(1).nullable()).optional(),
historyCap: z.number().int().min(1).max(1000).optional(),
defaultIncludeAlignment: z.boolean().optional()
});