Retain batch-change review evidence: originals, consumer version and decision scope
Retain a reviewable relationship, not a new summary pretending to be an original batch
After comparing exports or formats, a team still needs to identify the original materials, consumer version and confirmed scope behind its conclusion. Record these relationships in existing authorized tools. Yihu supplies no complete snapshot version store, migration ledger or immutable audit chain here. Existing cache/reply history has its own limits, not permanent archives of every batch.
Different from comparison, claim review and retention policy
Batch review checks comparability/values; format review checks compatibility; handoff review checks completion claims. Here an existing conclusion must match its archived before/after material, consumer and confirmation evidence. It does not redo the business decision, create approval or prescribe retention periods.
Change-material retention mapping
| Editorial record | Existing material referenced | Missing/confused risk | Boundary |
|---|---|---|---|
| raw_batch_before | Authorized original older batch/scope | Retrospective summary presented as input | Mark unavailable, not recreate a false original |
| raw_batch_after | Authorized newer batch/scope | Overwritten filename or changed source/permission | Verify scope/identity, not matching names alone |
| producer_revision | Version actually supplied by the workflow | v2 text assumed to negotiate a protocol | Record known facts; unknown stays unknown |
| consumer_revision | App/model version actually reviewed | Future/other consumer results substituted | Bind to reviewed version/scope |
| comparison_rule | Collection, fields, time and type rules | Ledger/subset or seconds/milliseconds mixed | State rules, not universal compatibility |
| review_scope | Reviewed/unreviewed fields and content | Few checks labeled full acceptance | Retain actual scope and gaps |
| decision_reference | Original confirmation/approved action | Archive prose treated as approval itself | Notes grant no writes or execution |
| archive_access | Existing location, readers and retention policy | Private disclosure or unnecessary copies | Existing minimum permissions, no new service provisioning |
These are editorial labels, not Radar imports or an app archive API. If an existing tool supplies a digest, it supports byte-integrity checks only. It does not independently establish trusted origin, truth or approval. No hashing/signing feature is implemented here.
A synthetic change-material record
Before: Sample batch A, original scope unreviewed.
After: Sample batch B, original scope unreviewed.
Consumer: Intended review version only; no substitution of another version's results.
Conclusion: Insufficient material for compatibility/add/remove claims.
Retention note: Preserve gaps/references; fabricate no original and perform no migration.
This is neither real data/credentials/private URLs nor an executed migration log. Do not copy a current batch and change its date to reconstruct missing history. That breaks the conclusion-input relationship. Copy/edit time does not replace source-read date_str either.
Review materials against decisions
Check access first, then inspect minimal originals and whether rules still concern the same collection/consumer. If fields/versions changed during review, preserve the change and revised scope; do not overwrite evidence while continuing to cite an earlier conclusion.
Approval supports its explicit scope, not automatic writeback, merging or deployment when a package is complete. Real handoffs retain original request/receipt checks; a new material revision does not erase uncertain prior operations. Actual policies govern documents/data/copies, without recovery guarantees or legal certification here.
FAQ
Does Yihu retain every batch/version change automatically?
No such feature is asserted. Existing cache/history is not a full snapshot store. Authorized tools maintain the suggested manual relationships.
Can a missing old file be recreated by changing the new file's date?
Do not fabricate originals that way. Mark the gap; editing time cannot replace source reading or recreate evidence.
Do matching hashes prove trusted and approved contents?
No. They support corresponding byte identity at most. Origin, business truth, approval and permission need separate evidence.
Can a new app version reuse an old compatibility conclusion?
Not unconditionally. Compare the intended consumer and scope; use the existing process to decide necessary revalidation.
Does archiving all material authorize every change?
No. Retention, review and operation approval differ. Write permission, phone confirmation and receipts remain.
Does this prescribe one retention period or guaranteed recovery?
No. Actual team/destination policies and permissions govern retention/deletion. No backup, restore or compliance service is added.
Reading and evidence
Batch comparison · Format compatibility · Handoff claims · Timezone checks · Start · Privacy
Evidence: source/cache scope, README reply history/authority. Editorial fields/relationships are suggestions, not native features.
