Status & trust
Live signals that the cache is fresh and mapped. This is the API proving its own correctness, not ops trivia — poll it before trusting a deployment.
What these mean
- Sync freshness
- Time since the block-watcher last advanced, and the last Hive Engine block it saw — an independent, open-source project, reads Hive Engine sidechain data directly, not an official Hive service.
- Refresh queue
- Accounts the block-watcher has flagged as touched by a recent NFT transaction, still waiting on the refresh worker to re-fetch their current holdings from Hive Engine. A queue that keeps growing means refreshes can't keep up with on-chain activity — see he-nft-directory's README for the cache design (there's no reconciliation step; the cache mirrors HE directly on every refresh instead of deriving state).
- Refresh retries
-
Queue entries that failed at least once and are waiting out an exponential backoff
before the worker re-attempts them — so a transient Hive Engine failure becomes a
visible, self-healing retry rather than silent staleness. A non-zero
pendingis normal during upstream trouble; what matters is that it drains afterward. - Display-mapping coverage
- Share of cached collections with a normalized display mapping. Coverage is seeded most-held-symbol-first; unmapped symbols still serve raw data, they're just not counted here yet.