Corpus Backfill

Closed tickets without an embedding. Bulk-add up to 20 per batch into aug_ticket_embeddings + aug_resolution_enrichment.

Corpus totals (all time)

Source tickets

—

in ticket (Zendesk mirror)

Embeddings

—

in aug_ticket_embeddings

Enriched tickets

—

DISTINCT in aug_resolution_enrichment

Enrichment rows

—

raw row count

Memory coverage

Resolved tickets since 2024-01-01 — the AI does not learn from older resolutions because they describe a product that has since changed.

Eligible

—

resolved tickets in range

In memory

—

Still missing

—

not yet ingested

Fully enriched

—

of what is in memory

Top up the AI's memory

Ingest resolved tickets from a date range. Tickets already in the corpus are skipped, so the total below counts only what is still missing — not every ticket in the range. Stop any time and resume later; nothing already ingested is redone.

From To (optional) Admin access is required to start a backfill.

Lookback window· Next in 5:00