Skip to main content

AI Personalization

AI Personalization

Configure AI personalization behavior, embedding sync, and profile calculation at site scope.

Audience: Personalization teams tuning AI relevance controls in production storefronts.

Critical: Personalization config endpoints are under /api/v1/ai/sites/:siteId/* and require AI permission scopes.

Who This Page Is For

Use this page when you need to turn AI personalization on/off, tune weighting strategy, and run embedding/profile pipelines.

Quick Start (2-5 Minutes)

1

Read current AI config

Inspect active personalization configuration for a site.

GET /api/v1/ai/sites/:siteId/personalization-config
2

Update weighting strategy

Tune embedding/popularity/recency weights and fallback behavior.

PUT /api/v1/ai/sites/:siteId/personalization-config
3

Sync embeddings

Trigger embedding generation synchronization job.

POST /api/v1/ai/sites/:siteId/sync-embeddings
4

Calculate user profiles

Recompute user profiles used by AI personalization ranking.

POST /api/v1/ai/sites/:siteId/calculate-user-profiles
5

Monitor progress and coverage

Use sync-progress and embedding-stats endpoints for operational tracking.

GET /api/v1/ai/sites/:siteId/sync-progress
GET /api/v1/ai/sites/:siteId/embedding-stats

Personalization endpoints

GET  /api/v1/ai/sites/:siteId/personalization-config
PUT  /api/v1/ai/sites/:siteId/personalization-config
POST /api/v1/ai/sites/:siteId/sync-embeddings
GET  /api/v1/ai/sites/:siteId/embedding-stats
POST /api/v1/ai/sites/:siteId/calculate-user-profiles
GET  /api/v1/ai/sites/:siteId/sync-progress

Required Fields / Minimum Payload

FieldRequiredTypeUsed by eventsDescription
isEnabledOptionalbooleanUpdate personalization configMaster toggle for AI personalization usage.
embeddingWeightOptional0..1 numberUpdate personalization configWeight of embedding similarity signal.
popularityWeightOptional0..1 numberUpdate personalization configWeight of popularity signal.
recencyWeightOptional0..1 numberUpdate personalization configWeight of recency signal.
coldStartStrategyOptionalpopular | trending | new_arrivalsUpdate personalization configFallback strategy for sparse profiles.
minProductsForAiOptional5..10000 numberUpdate personalization configMinimum product count required before AI mode engages.

Update config example

{
"isEnabled": true,
"embeddingWeight": 0.7,
"popularityWeight": 0.2,
"recencyWeight": 0.1,
"useFallback": true,
"coldStartStrategy": "trending",
"minProductsForAi": 20
}

Event or Endpoint Decision Matrix

ScenarioUse ThisWhy
Need current AI config baselineGET personalization-configRead before changes.
Need to tune ranking behaviorPUT personalization-configPrimary control plane for AI weighting.
Need fresh embedding corpusPOST sync-embeddingsRegenerates embedding state after catalog changes.
Need profile refresh after identity changesPOST calculate-user-profilesRebuilds user profile vectors/signals.
Need operational sync visibilityGET sync-progress and embedding-statsTracks job state and coverage quality.

Common Errors and Fixes

Config update rejected

Cause: Weights out of 0..1 bounds or invalid coldStartStrategy value.

Fix: Use allowed ranges and enum values only.

Personalization quality drops after update

Cause: Overweighted popularity/recency and underweighted embeddings.

Fix: Rebalance weights and monitor performance endpoint trends.

Sync remains in progress too long

Cause: Large corpus or processing bottleneck.

Fix: Monitor sync-progress and plan updates during lower-load windows.

Coverage metrics low

Cause: Insufficient product/user data for profile generation.

Fix: Increase data completeness and rerun sync/profile calculation.

Production Checklist

  • AI config changes are versioned and approved before production rollout.Required
  • Weight values remain within controlled policy ranges.Required
  • Embedding sync and profile jobs are monitored after catalog updates.Required
  • Fallback strategy is explicitly set for cold-start cases.Required
  • Performance is reviewed after each major config change.Required

Next Steps