# Future plan

The current release uses only workspace-local files and `cache-manifest.json`.

## Database-backed cache

Plan a later release around these components:

1. Store normalized standard identifiers, official metadata, PDF SHA-256, parse-profile SHA-256, output locations, and timestamps in a database.
2. Store PDFs and MinerU artifacts in object storage rather than database binary columns.
3. Deduplicate source PDFs by SHA-256 while retaining the standard-number relationship.
4. Add distributed task locks so multiple agents do not download or parse the same standard concurrently.
5. Version parsing profiles and retain previous results for audit and regression comparison.
6. Add explicit refresh, retention, integrity-check, and cache-eviction policies.

Do not add database access until the storage provider, deployment topology, tenancy boundary, and retention requirements are defined.