Skip to main content

Registry Health

Read-only health snapshot of your field registry: tier and admission counts, singleton and weak-name rates, occurrence and resolution stats, transfer hit rate, and coherence/atomicity metrics.

The registry health endpoint returns a read-only snapshot of your workspace's field registry — the same view the in-app registry dashboard reads. It is how you monitor the health of the deduplicated knowledge graph that extraction builds over time: how many canonical concepts exist, how many fields are still provisional, how often deterministic transfer fills cells without an AI call, and where naming or coherence is degrading. No lifecycle or cleanup actions are exposed; this is a metrics read only.

The response is grouped into sections. registry carries tier counts (tier 1 canonical, tier 2 provisional, tier 3 candidate), admission counts, and the singleton rate. occurrences and resolution_last_24h track how values bind into the graph. transfer reports the deterministic transfer hit rate (the share of cells filled with no AI call). name_quality, coherence, and atomicity surface quality signals — weak names, incoherent canonicals, non-atomic values — and junk_classes flags clusters that look like noise.

Use this to track registry health as your corpus grows. A rising transfer hit rate means more cells resolve for free; a rising singleton or weak-name rate is an early signal that naming or clustering needs attention.
GET/v1/registry/health

Response

Response sections

registryobjecttotal_fields, tier_1/2/3, singleton_count, singleton_rate, admission_canonical, admission_provisional, admission_provisional_rate.
embeddingsobjectprovisional_backlog and oldest_provisional_minutes.
occurrencesobjecttotal, unresolved, needs_confirmation.
resolution_last_24hobjecttotal resolutions in the last 24h and their confidence-band breakdown.
convergenceobjectconvergence_rate and confirm_rate (nullable).
name_qualityobjectweak_name_count and weak_name_rate.
coherenceobjectmulti_occ_fields, incoherent_canonical_count, incoherent_rate.
transferobjecttransfer_cells, extraction_cells, transfer_hit_rate.
atomicityobjecttotal_values, non_atomic_count, non_atomic_rate.
junk_classesobjectCounts of clusters flagged as likely noise.

curl

Response (truncated)

{
  "registry": {
    "total_fields": 8174,
    "tier_1": 162,
    "tier_2": 724,
    "tier_3": 7288,
    "singleton_count": 6099,
    "singleton_rate": 0.746,
    "admission_canonical": 6449,
    "admission_provisional": 1725,
    "admission_provisional_rate": 0.211
  },
  "transfer": { "transfer_cells": 12044, "extraction_cells": 5310, "transfer_hit_rate": 0.69 },
  "name_quality": { "weak_name_count": 88, "weak_name_rate": 0.011 },
  "coherence": { "multi_occ_fields": 2075, "incoherent_canonical_count": 12, "incoherent_rate": 0.006 }
}