docs(evals): one eval page, naming evalkit and aos-eval #305

Merged

coilysiren merged 1 commit from aos/claude/eval-docs-one-page into main

2026-08-18 16:07:59 +00:00

coilyco-ops commented

2026-08-18 15:57:59 +00:00

Member

Five pages covered the eval surface here, three of them reference notes, and two carried headings left over from earlier mechanical merges (### Why one way and what it reads and what it emits and the grader's own notes are withheld and it refuses rather than redacts and commands).

Consolidates the whole surface into docs/evaluation.md: the evalkit and aos-eval seam, the triple, the derived dataset, the pairing rule, annotation, and the one-way export, with the six external references compressed to a closing list.

Deletes eval-pipeline.md, eval-references.md, eval-ref-papers.md, and eval-ref-platforms.md, and repoints the three evalkit/*.py source comments and the two FEATURES entries at the surviving page. Docs count drops 40 to 36.

The deep per-reference notes are in git history at the parent commit if any of them should come back as a single page.

Also corrects a claim the old page carried: aos-eval is shared with sirens-echo in intent, and shares the boundary declaration shape today, but not the dataset schema, so that repo still grades locally.

pre-commit run --all-files passes.

🤖 Generated with Claude Code

Five pages covered the eval surface here, three of them reference notes, and two carried headings left over from earlier mechanical merges (`### Why one way and what it reads and what it emits and the grader's own notes are withheld and it refuses rather than redacts and commands`). Consolidates the whole surface into `docs/evaluation.md`: the `evalkit` and `aos-eval` seam, the triple, the derived dataset, the pairing rule, annotation, and the one-way export, with the six external references compressed to a closing list. Deletes `eval-pipeline.md`, `eval-references.md`, `eval-ref-papers.md`, and `eval-ref-platforms.md`, and repoints the three `evalkit/*.py` source comments and the two FEATURES entries at the surviving page. Docs count drops 40 to 36. The deep per-reference notes are in git history at the parent commit if any of them should come back as a single page. Also corrects a claim the old page carried: `aos-eval` is shared with sirens-echo in intent, and shares the boundary declaration shape today, but not the dataset schema, so that repo still grades locally. `pre-commit run --all-files` passes. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

coilyco-ops added 1 commit

2026-08-18 15:57:59 +00:00

docs(evals): one eval page, naming evalkit and aos-eval

ci / test (pull_request) Successful in 1m11s

Details

bb989eb9ab

Five pages covered the eval surface here, three of them reference notes, and
two carried headings left over from earlier mechanical merges. Consolidate the
whole surface into docs/evaluation.md: the evalkit and aos-eval seam, the
triple, the derived dataset, the pairing rule, annotation, and the one-way
export, with the six external references compressed to a closing list.

Deletes eval-pipeline.md, eval-references.md, eval-ref-papers.md, and
eval-ref-platforms.md, and repoints the evalkit source comments and the two
FEATURES entries at the surviving page.

Also corrects a claim the old page carried: aos-eval is shared with sirens-echo
in intent, and shares the boundary declaration shape today, but not the dataset
schema, so that repo still grades locally.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Co-authored-by: Kai Siren <coilysiren@gmail.com>
Co-authored-by: Claude <noreply@anthropic.com>

coilyco-ops requested review from coilysiren

2026-08-18 15:57:59 +00:00