Source data lifecycle
Sentium separates provider authorization, immutable source capture, derived processing, regulated review, and deletion. Each stage keeps its own identity and audit history.
1. Authorize and select
An authorized user grants bounded provider access or uploads a file. Sentium records the tenant, destination company or device, selected provider scope, and responsible actor.
2. Capture a snapshot
Sentium copies eligible source bytes into an immutable Source Inbox version. The provider remains the source of incoming content; the Sentium version becomes the stable input for processing and citation.
3. Scan and process
Source bytes are scanned before parsing. Accepted versions are extracted and divided into retrieval units that retain source provenance. Processing failures are visible and the failed version is not presented as ready evidence.
4. Retrieve and review
Only processed, tenant-authorized versions can support search, Agent responses, or drafting. Citations point to an exact source version. Inferred facts and relationships remain labeled until reviewed; generated content does not become controlled evidence automatically.
5. Reconcile change
Continuing connections detect provider differences. Changed content creates a new immutable version; unchanged content reuses compatible work; deleted or inaccessible content receives a tombstone and leaves current retrieval.
6. Retain or delete
Disconnect stops future provider reads but does not silently erase governed history. Retention follows the tenant's policy and applicable record obligations. Authorized deletion removes eligible stored content and derived data while preserving the minimum audit evidence required to document the action.
Read versions and deletions for source-history behavior and source permissions and security for the access boundary.