Data freshness.
Per-event freshness over the alphakernel parquet store — how many symbols have shards under each event, how many shards total, when the newest and oldest shards were written, and how stale the latest one is. The page polls/v1/data/freshness every 30 seconds. Ingest is run by ak data ingest-polygon(and friends); this page only watches.
For the worst-lag summary on the morning briefing, see/desk. For the data sources backing each event, see/projects/feature-forge.
events · per-partition freshness
—| event | symbols | shards | oldest | latest | lag |
|---|---|---|---|---|---|
| loading… | |||||
gold-layer artifacts · per-file freshness
| kind | name | path | last modified | lag |
|---|---|---|---|---|
| none configured | ||||
how this is wired
The page polls one GET endpoint on a 30-second interval:/v1/data/freshness. The server walks ParquetStore.list_shards()over the hive-partitioned store atALPHAKERNEL_PARQUET_BASE, groups shards by event, and returns symbol/shard counts plus oldest/latest dates per event.
Lag is measured from 00:00 UTCof the latest shard date — same renderable form as dbt-style source-freshness checks. The chip turns green under 24h, amber 24–72h, red beyond 72h.
When the parquet store is unconfigured the API returns503; when configured but empty it returns 200 withevents: [] — distinct empty states for distinct operator actions.
For the operator:point the API at the parquet store viaALPHAKERNEL_PARQUET_BASE (local FS or s3://…); addALPHAKERNEL_PARQUET_STORAGE_OPTIONSfor s3 credentials. Plant the first shards withak data ingest-polygon orak data populate-synth.