Interface BulkIngestionUpsertContextsResponse

Context ids assigned to the staged contexts, in the same order as the request's contexts array. A context rejected as a duplicate still occupies its slot here, so this stays index-aligned with the request; cross-reference duplicates by contextId to tell which ids were actually staged.

Contexts rejected because the knowledge base already holds identical content — either stored under an existing context or accepted earlier in this same call. Absent when nothing was rejected.

Same shape and errorCode: 'DUPLICATE_CONTENT' the synchronous upsertContexts returns, so a caller handles a duplicate identically on both lanes. These contexts are NOT part of the job: they are not extracted, not staged, and never appear in its per-file statuses.

Covers TEXT contexts and MULTIPART file contexts (their bytes are in hand at staging). A file staged by stagedObjectKey is the one exception: its bytes are first read by the deferred extraction pass, so its duplicate surfaces asynchronously as a failed file with errorCode: 'DUPLICATE_CONTENT' in the job's per-file statuses instead.

The id of the staged bulk-ingestion job.