Complete five-document pilot corpus review and Milestone 1 #66

Merged

JackFrostbyte merged 1 commit from feature/issue-18-pilot-corpus-review into develop

2026-08-30 01:40:53 -04:00

JackFrostbyte commented

2026-08-30 01:39:13 -04:00

Owner

Closes #18

Summary

Completes the human/editorial review and final validation of the approved five-document pilot corpus for Milestone 1 — Reproducible Acquisition & Processing v0.1.

All five approved pilot sources were independently reviewed against their committed originals, provenance and license evidence, generated text, lineage, and chunks.

No source-data, fixture, or substantive processing defects were found, so no corpus or pipeline changes were required.

This PR adds the Milestone 1 completion record at:

docs/roadmap/milestone-1-completion.md

Corpus reviewed

src_attributions-use-tasl
src_introducing-the-shell
src_naming-files-folders-and-other-things
src_summary-licensing-best-practices-for-sharing-scientific-data
src_ten-simple-rules-for-taking-advantage-of-git-and-github

For each applicable source, review covered:

source identity, authorship/publisher metadata, provenance, and acceptance state;
license evidence and redistribution/bundle eligibility;
independently recalculated original byte size and SHA-256;
extracted and normalized text against the authoritative original;
transformation lineage;
every generated chunk's lineage, deterministic ordering, exact text, half-open character offsets, substantive coverage, and available page locations.

The three born-digital PDFs were also manually reviewed across representative beginning, middle, and end passages, physical page transitions, and layout-sensitive material.

Expected plain-PDF reading-order differences were observed on layout-heavy material, but no silent substantive extraction or normalization loss was found. Meaningful visual information remains authoritative in the original PDFs under the existing Arkive contract.

Validation

Complete pilot regeneration produced no committed-output changes.

Canonical validation:

PASS: 6 source(s), 14 document(s), 38 chunk(s), 0 rights profile(s), 0 recipe(s) across 18 record file(s) — all valid.

The sixth source is the existing Arkive-created regression fixture and is not counted among the five Milestone 1 pilot documents.

Final validation results:

complete unit suite: PASS — 304 tests
Ruff check: PASS
Ruff format check: PASS
coverage: PASS — 92% total
canonical schema/cross-record validation: PASS
pip check: PASS
git diff --check: PASS
clean-checkout/reproducibility suite: PASS
complete arkive-process-pilot regeneration: PASS
regenerated corpus diff: clean
final branch based on current origin/develop

Scope

This PR intentionally does not:

add another pilot document;
redesign or modify the processing pipeline;
add OCR;
add search indexes, embeddings, LLM processing, translation, or speech;
expand safety-critical content;
perform production release work.

Milestone 1 completion criteria are now satisfied.

Closes #18 ## Summary Completes the human/editorial review and final validation of the approved five-document pilot corpus for Milestone 1 — Reproducible Acquisition & Processing v0.1. All five approved pilot sources were independently reviewed against their committed originals, provenance and license evidence, generated text, lineage, and chunks. No source-data, fixture, or substantive processing defects were found, so no corpus or pipeline changes were required. This PR adds the Milestone 1 completion record at: `docs/roadmap/milestone-1-completion.md` ## Corpus reviewed * `src_attributions-use-tasl` * `src_introducing-the-shell` * `src_naming-files-folders-and-other-things` * `src_summary-licensing-best-practices-for-sharing-scientific-data` * `src_ten-simple-rules-for-taking-advantage-of-git-and-github` For each applicable source, review covered: * source identity, authorship/publisher metadata, provenance, and acceptance state; * license evidence and redistribution/bundle eligibility; * independently recalculated original byte size and SHA-256; * extracted and normalized text against the authoritative original; * transformation lineage; * every generated chunk's lineage, deterministic ordering, exact text, half-open character offsets, substantive coverage, and available page locations. The three born-digital PDFs were also manually reviewed across representative beginning, middle, and end passages, physical page transitions, and layout-sensitive material. Expected plain-PDF reading-order differences were observed on layout-heavy material, but no silent substantive extraction or normalization loss was found. Meaningful visual information remains authoritative in the original PDFs under the existing Arkive contract. ## Validation Complete pilot regeneration produced no committed-output changes. Canonical validation: ``` PASS: 6 source(s), 14 document(s), 38 chunk(s), 0 rights profile(s), 0 recipe(s) across 18 record file(s) — all valid. ``` The sixth source is the existing Arkive-created regression fixture and is not counted among the five Milestone 1 pilot documents. Final validation results: * complete unit suite: PASS — 304 tests * Ruff check: PASS * Ruff format check: PASS * coverage: PASS — 92% total * canonical schema/cross-record validation: PASS * `pip check`: PASS * `git diff --check`: PASS * clean-checkout/reproducibility suite: PASS * complete `arkive-process-pilot` regeneration: PASS * regenerated corpus diff: clean * final branch based on current `origin/develop` ## Scope This PR intentionally does not: * add another pilot document; * redesign or modify the processing pipeline; * add OCR; * add search indexes, embeddings, LLM processing, translation, or speech; * expand safety-critical content; * perform production release work. Milestone 1 completion criteria are now satisfied.

JackFrostbyte added 1 commit

2026-08-30 01:39:13 -04:00