Skip to content

fix(record): close ENG-NOW-DERIVED honestly (#374) - #386

Merged
localai-bot merged 1 commit into
mainfrom
row/ENG-NOW-DERIVED-DONE
Aug 11, 2026
Merged

fix(record): close ENG-NOW-DERIVED honestly (#374)#386
localai-bot merged 1 commit into
mainfrom
row/ENG-NOW-DERIVED-DONE

Conversation

@localai-bot

Copy link
Copy Markdown
Collaborator

Closes #374.

The implementation merged in #376, but the row remained ACTIVE after its branch was removed, leaving current main red in audit-live-rows. This repair records the shipped W1-W5 outcome as DONE, marks runtime/performance parity VOID for local protocol machinery, releases the claim, and preserves the lifecycle projections in STATUS/BENCHMARKS while moving per-row live position to the owning spec ## Now.

Review evidence:

  • clean-main RED: only ENG-NOW-DERIVED abandoned
  • focused audit/record/doc/gate-command suites green
  • scratch mutations prove the AGENTS rule, runnable departure, and DONE claim guard are load-bearing
  • fresh immutable review PASS at f1cffbe9
  • operator scripts/agent-preflight.sh green immediately before exact-SHA push

The separate PARTIAL parser defect found during review is tracked in #380 and is not changed here.

FOLLOWING_AGENTS_PROTOCOL

Following-Agents-Protocol: true
AI-Assisted: true
Assisted-by: OpenAI Codex GPT-5 [Codex]

W1-W5 landed in dbd0d51, but the row and claim remained ACTIVE and made current main fail audit-live-rows. Close the row as DONE with VOID runtime/performance/parity evidence, preserve progressive spec backfill as the selected compatibility policy, and correct the public lifecycle projection rule.

DONE rows now leave the live runnable-command population. The exact baseline re-pin names all three departures and a paired mutation proves restoring DONE exposes them again.

FOLLOWING_AGENTS_PROTOCOL

Following-Agents-Protocol: true
AI-Assisted: true
Assisted-by: Codex:GPT-5 [Codex]
@localai-bot
localai-bot merged commit b580452 into main Aug 11, 2026
21 of 22 checks passed
localai-bot pushed a commit that referenced this pull request Aug 11, 2026
…ding

`origin/main` advanced a FOURTH time during this landing (2 commits, #374/#386
closing `ENG-NOW-DERIVED`). A plain `git push` was refused as non-fast-forward,
so: fetch, re-merge, reconcile, re-gate. No force, no rebase.

No C++ changed -- the delta is records and checkers only -- so the build is
unchanged (`ninja: no work to do`) and the focused gates were re-run anyway.

KEYED RECORDS.

  .agents/parity-ledger.md  CONFLICT. Append-only dated log, so a UNION: main's
                            `ENG-NOW-DERIVED` entry followed by this landing's
                            `FUSION-DENSE-MIGRATE` entry. Both kept in full.
  docs/BENCHMARKS.md        CONFLICT. Main REWROTE the `Developer agent entry
                            point` row into `Developer/row protocol`; main's
                            rewritten row is taken and PR #282's `LoRA runtime
                            W2` row kept beside it. 44,838 of the hard 45,000.
  .agents/engine-matrix.md  CONFLICT. Main moved a `Serving` row `ACTIVE` ->
                            `DONE`. Main's row kept wholesale, the landing's
                            `LoRA and adapters` row reapplied, **Total**
                            RECOMPUTED from the ten actual area rows:
                            149|35|16|4|9|27|8|10|39. `check-agent-record.py`
                            confirms ENGINE=149.
  docs/STATUS.md, scripts/check-gate-commands.py,
  tests/scripts/test_check_gate_commands.py, .agents/claims/,
  .agents/specs/now-derived.md and AGENTS.md auto-merged and were verified
  against main's own edit set.

GATE, re-run by the operator on this tree, CPU Release, foreground:
  cmake --build build-cpu -j 18   no work to do (no C++ in main's delta)
  test_dense_gate_up_seam_forward  4/4  | 1940 | 0 skipped
  test_linear_method               6/6  |   76 | 0 skipped
  test_lora_layers                16/16 | 4498 | 0 skipped
  test_punica_cpu                  8/8  |  149 | 0 skipped
  test_sampler                    21/21 |  114 | 0 skipped
  test_input_batch                29/29 |  205 | 0 skipped
  test_llm_engine                 24/24 |  491 | 0 skipped
  check-agent-record / check-public-doc-tables / check-now-current /
  check-fusion-consistency / check-gate-commands / check-commit-trailers  all OK
  Checker unit suites: test_check_public_doc_tables 52/52,
  test_check_gate_commands 30/30, test_check_fusion_consistency 20/20,
  test_agent_record 29/29, test_check_now_current 15/15,
  test_record_merge_shape 11/11, test_check_pr_size 38/38,
  test_doc_checkpoint 27/27.

  The full 384-test `ctest -j 6` for this landing is on the merge below; main's
  delta here is records and checkers with no C++, so it cannot move it.

ONE RED, PRE-EXISTING ON MAIN, NOT THIS LANDING'S, AND NOT FIXED HERE:
`tests/scripts/test_now_render.py` is 10/11 --
`test_a_spec_now_line_is_picked_up` asserts the literal `"W2"` appears in
`now.spec_now("now-derived.md")`, and main's own `f1cffbe9` rewrote that spec's
`## Now` to "W1-W5 shipped" without updating the test. `now.spec_now()` reads
exactly one file, and `tests/scripts/test_now_render.py`,
`.agents/specs/now-derived.md` and `scripts/now.py` are all byte-identical to
`b580452d` in this tree, so nothing here can influence the result. Filed as
[#388](#388) and deliberately left to
the `ENG-NOW-DERIVED` owner: the repair is a judgement about what the guard
should assert now, and AGENTS.md forbids turning a red gate green by weakening
an assertion.

FOLLOWING_AGENTS_PROTOCOL

Following-Agents-Protocol: true
AI-Assisted: true
Assisted-by: ClaudeCode:claude-opus-5 [ClaudeCode]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

NOW.md is still a surface every PR must write: the doc-checkpoint gate marches them into it

2 participants