Run a ten-source user-overlay end-to-end pilot for v0.1 #73

New issue

Open

opened 2026-09-27 01:20:42 -04:00 by JackFrostbyte · 0 comments

JackFrostbyte commented

2026-09-27 01:20:42 -04:00

Owner

Goal

Run the final ten-source end-to-end pilot for the User-Acquired Content Overlay v0.1 milestone using the completed reviewed acquisition, recovery, processing, indexing and core-separation tools. Keep this pilot separate from the existing five-document redistributable-core pilot; it validates optional user-local external content only.

Corpus and review gate

Select ten real external source items/recipes spanning useful categories, document types and source failure modes. Aim for both manual_import and official_direct_download; include at least two of each if reviewed and legally eligible, and document an exception if that distribution is impossible.
Each included source needs its own documented rights, allowed acquisition mode/host, jurisdiction and review dates, expected format, source identifiers, version/known checksums (if published), review evidence and owner. Do not invent permissive legal conclusions or use material that cannot pass the approved eligibility review.
Do not commit or redistribute externally acquired originals, their derived text, personal data or user-local indexes as part of the public core or test fixtures.

Test matrix

Review and validate all ten acquisition records before live work; run public-service checks only where the publisher's reviewed terms and adapter policy permit. CI uses local fixtures/fakes, never live publishers.
On a clean user-local overlay, acquire/import each approved item and exercise local-only processing and separate index generation. Record per-source provenance and actual results.
Inject deterministic local-fixture failures: unavailable resource/redirect, checksum mismatch, interrupted transfer, and one bad input. Demonstrate secure quarantine plus a usable partial optional overlay and visible missing-source report.
After correcting controlled fixtures or legitimately updating a reviewed recipe, rerun incrementally: unchanged valid items are revalidated/reused; only missing, invalid, appropriately changed or incomplete items are reacquired/reprocessed/reindexed as required.
Exercise maintainer catalogue audit output and compare findings to controlled known states; confirm no checksum or source version is automatically trusted.
Verify the canonical validator, core/overlay separation, index manifests and #49 (federated search), citation provenance where available. Specifically prove no user-local original/derived data, caches or diagnostics enter core release inputs or support bundles.
Reproduce the pilot using exact recorded commands/environment and publish only redacted results and approved metadata, not restricted source content.

Acceptance criteria

Ten independently reviewed external source items have a recorded acquisition decision and actual result; all gaps and eligibility exceptions are documented.
Validated acquisitions can be processed and indexed locally while an intentionally bad optional source is quarantined and reported without corrupting the valid remainder.
Rerunning demonstrates verifiable per-source reuse and targeted recovery after controlled errors or a reviewed version change.
Maintainer audit diagnoses intentionally unavailable/changed content and never silently approves altered bytes.
User-facing completion clearly distinguishes complete from partial, listing missing/invalid items and recommended next actions.
No unapproved external content or user-private diagnostics contaminate the redistributable core; existing five-document core pilot still passes unchanged.
Offline-only automated regressions, canonical validator, full unit tests, lint/format checks and reproducibility tests pass; publish a redacted pilot report and measured resource needs.

Dependencies and exclusions

Final validation after #46, #47, #48, #49 (federated search), #50, #71 and #72 are complete. This is an integration and evidence issue, not permission to implement new adapters, weaken source policy or redistribute restricted files. If full packaging is outside this milestone, demonstrate the local overlay with its supported development CLI rather than silently broadening scope.

## Goal Run the final **ten-source end-to-end pilot** for the **User-Acquired Content Overlay v0.1** milestone using the completed reviewed acquisition, recovery, processing, indexing and core-separation tools. Keep this pilot **separate from the existing five-document redistributable-core pilot**; it validates optional user-local external content only. ## Corpus and review gate - Select **ten real external source items/recipes** spanning useful categories, document types and source failure modes. Aim for both `manual_import` and `official_direct_download`; include at least two of each *if reviewed and legally eligible*, and document an exception if that distribution is impossible. - Each included source needs its own documented rights, allowed acquisition mode/host, jurisdiction and review dates, expected format, source identifiers, version/known checksums (if published), review evidence and owner. Do not invent permissive legal conclusions or use material that cannot pass the approved eligibility review. - Do not commit or redistribute externally acquired originals, their derived text, personal data or user-local indexes as part of the public core or test fixtures. ## Test matrix 1. Review and validate all ten acquisition records before live work; run public-service checks only where the publisher's reviewed terms and adapter policy permit. CI uses local fixtures/fakes, never live publishers. 2. On a clean user-local overlay, acquire/import each approved item and exercise local-only processing and separate index generation. Record per-source provenance and actual results. 3. Inject deterministic local-fixture failures: unavailable resource/redirect, checksum mismatch, interrupted transfer, and one bad input. Demonstrate secure quarantine plus a usable **partial optional overlay** and visible missing-source report. 4. After correcting controlled fixtures or legitimately updating a reviewed recipe, rerun incrementally: unchanged valid items are revalidated/reused; only missing, invalid, appropriately changed or incomplete items are reacquired/reprocessed/reindexed as required. 5. Exercise maintainer catalogue audit output and compare findings to controlled known states; confirm no checksum or source version is automatically trusted. 6. Verify the canonical validator, core/overlay separation, index manifests and #49 (federated search), citation provenance where available. Specifically prove no user-local original/derived data, caches or diagnostics enter core release inputs or support bundles. 7. Reproduce the pilot using exact recorded commands/environment and publish **only redacted results and approved metadata**, not restricted source content. ## Acceptance criteria - [ ] Ten independently reviewed external source items have a recorded acquisition decision and actual result; all gaps and eligibility exceptions are documented. - [ ] Validated acquisitions can be processed and indexed locally while an intentionally bad optional source is quarantined and reported without corrupting the valid remainder. - [ ] Rerunning demonstrates verifiable per-source reuse and targeted recovery after controlled errors or a reviewed version change. - [ ] Maintainer audit diagnoses intentionally unavailable/changed content and never silently approves altered bytes. - [ ] User-facing completion clearly distinguishes **complete** from **partial**, listing missing/invalid items and recommended next actions. - [ ] No unapproved external content or user-private diagnostics contaminate the redistributable core; existing five-document core pilot still passes unchanged. - [ ] Offline-only automated regressions, canonical validator, full unit tests, lint/format checks and reproducibility tests pass; publish a redacted pilot report and measured resource needs. ## Dependencies and exclusions Final validation after #46, #47, #48, #49 (federated search), #50, #71 and #72 are complete. This is an integration and evidence issue, not permission to implement new adapters, weaken source policy or redistribute restricted files. If full packaging is outside this milestone, demonstrate the local overlay with its supported development CLI rather than silently broadening scope.

JackFrostbyte added this to the User-Acquired Content Overlay v0.1 milestone

2026-09-27 01:20:42 -04:00

JackFrostbyte referenced this issue

2026-09-27 01:20:42 -04:00

Implement resilient batch acquisition and incremental retry #72

JackFrostbyte referenced this issue

2026-09-27 01:20:42 -04:00

Implement safe user-side acquisition and manual import #46

JackFrostbyte referenced this issue

2026-09-27 01:20:43 -04:00

Process user-acquired content as local-only derived artifacts #47

JackFrostbyte referenced this issue