Skip to content

release: add versioned binary manifest (W5) - #141

Open
localai-bot wants to merge 7 commits into
mainfrom
row/ENG-RELEASE-BINARIES
Open

release: add versioned binary manifest (W5)#141
localai-bot wants to merge 7 commits into
mainfrom
row/ENG-RELEASE-BINARIES

Conversation

@localai-bot

@localai-bot localai-bot commented Aug 8, 2026 •
edited
Loading

Copy link
Copy Markdown
Collaborator

Claim

Implements W5 of ENG-RELEASE-BINARIES from the accepted release matrix in #129. This PR is the active CLAIM-ENG-RELEASE-BINARIES-W5 claim.

Scope

  • versioned manifest schema and deterministic build-time generator/validator
  • independent evidence states with exact per-CPU-tier and per-CUDA-SM metadata
  • canonical CPU/CUDA fixtures, fail-closed mutation tests, and unconditional CI/preflight registration
  • synchronized ACTIVE/pending-only lifecycle and public status records
  • minimal current-main repair for public keyed-document and environment-variable gates introduced by perf(gemma4): dual-GPU FP8 resident without host OOM + peer mix #154
  • explicit PR-size classifications with the governed product total at exactly 900 lines

Excluded from this W5 PR: W1-W4 and W6-W13, install/package/archive/publish targets, runtime artifacts, GPU execution, downloads, services, and release channel advancement.

Evidence

  • exact head: d042533
  • exact base: 6db9ec5
  • focused release/PR-size suites: 70/70
  • release-contract suite: 30/30
  • W5 targeted production mutations: 12/12 killed
  • current-main repair mutations: 4/4 killed
  • PR-size classifier mutations: 5/5 killed
  • actual PR-size gate: GREEN at the exact 900-line product budget
  • public documentation ratchet: GREEN at 276835 chars against the lowered 276860 cap; tightness mutation 1/1 killed
  • committed scripts/agent-preflight.sh: all gates green at the exact head
  • stable W5 patch ID: 9d86ad80468e08a84d216f3483a822dec52d0024

No archive, runtime, correctness, performance, download, or publication evidence is claimed by W5.

mudler added a commit that referenced this pull request Aug 8, 2026
@mudler
feat(release): add versioned binary manifest contract (#141)
c43ed3f 
Implement W5 with a deterministic stdlib-only manifest generator, versioned schema, canonical CPU/CUDA fixtures, and fail-closed mutation coverage. Advance only the release program lifecycle to ACTIVE while preserving every artifact and runtime gate as pending.

FOLLOWING_AGENTS_PROTOCOL
Assisted-by: Codex:GPT-5 [Codex]
@localai-bot

localai-bot commented Aug 8, 2026

Copy link
Copy Markdown
Collaborator Author

CI record-gate repair at 16dc6222853571316f1e0fc29d747fb68c277f54:

  • RED first: a local-only row/LOCAL-ONLY ref made the new branch-scope regression fail; after restricting shared claim evidence to fetched remote refs, the test passes.
  • Clean-ref reproduction (only origin/main + this PR ref): audit suite 42/42; audit-live-rows.py --check rc=0 with 0 abandoned ACTIVE rows.
  • Lifecycle repair: MODEL-EMBED-llama-llama-for-causal-lm is PARTIAL at main 57ed063e (1/8 memberships plus synthetic gate landed; real-checkpoint oracle remains pending), with checklist, coordination, NOW/state, public docs and runnable-gate pin reconciled.
  • Full staged preflight and post-commit push-gated preflight: rc=0. Mutation restoring local-ref parsing fails as intended.

No release artifact/runtime/performance claim changed; reviewed W5 implementation blobs remain untouched.

mudler added a commit that referenced this pull request Aug 8, 2026
@mudler
docs(benchmarks): the Laguna rows measure S-2.1, not "Laguna-XS"
4a6d507 
The at-a-glance row, the section heading and the closed-row table all
called the benchmarked model Laguna-XS-2.1. The measured checkpoint is
poolside/Laguna-S-2.1-NVFP4: 118B total / ~8B active MoE, 48 layers,
256 experts, ~67 GiB.

The label came from the local checkpoint directory being named
laguna-xs-nvfp4. Evidence that the two names are one benchmark: the
same 37.55 -> 44.46 vs vLLM 43.10 pair appears in this file under
"Laguna-XS NVFP4" and in the same document's row for "Laguna-S-2.1 MoE
(LagunaForCausalLM, 118B/8B)", both dated 2026-08-04; and the NVFP4 arm
spec pins the checkpoint at poolside/Laguna-S-2.1-NVFP4, ~67 GiB, with
layers 1..47 MoE.

The section now states the model geometry and says where the XS label
came from, so it cannot drift back. The reproduce row keeps the real
directory name with a note that it holds the S-2.1 checkpoint.

Numbers, ratios and evidence anchors are unchanged; this is a naming
correction only. FEATURES.md and README are untouched: they list
"Laguna-S / Laguna-XS 2.1" as a model family, which is a separate
question from which checkpoint was measured.

No open issue or PR covers this (searched issues and PRs for laguna
naming; open PRs are #127, #128, #140, #141, none related).

FOLLOWING_AGENTS_PROTOCOL
Assisted-by: Claude Code:claude-opus-5 [ClaudeCode]
mudler added a commit that referenced this pull request Aug 8, 2026
@mudler
docs(gemma4): specify #140 regression repairs
e0b185c 
Ground the two direct GeGLU call sites and the runner device-identity leak, bind red-first tests, two independent repair commits, mutation review, and the #141 reconciliation sequence.

FOLLOWING_AGENTS_PROTOCOL
Assisted-by: Codex:GPT-5 [Codex]
@localai-bot localai-bot mentioned this pull request Aug 8, 2026
Draft
mudler added a commit that referenced this pull request Aug 8, 2026
@mudler
docs(gemma4): specify #140 regression repairs
2af390e 
Ground the two direct GeGLU call sites and the runner device-identity leak, bind red-first tests, two independent repair commits, mutation review, and the #141 reconciliation sequence.

FOLLOWING_AGENTS_PROTOCOL

Following-Agents-Protocol: true
AI-Assisted: true
Assisted-by: Codex:GPT-5 [Codex]
mudler added a commit that referenced this pull request Aug 8, 2026
@mudler
docs(gemma4): specify #140 regression repairs
f55ff3b 
Ground the two direct GeGLU call sites and the runner device-identity leak, bind red-first tests, two independent repair commits, mutation review, and the #141 reconciliation sequence.

FOLLOWING_AGENTS_PROTOCOL

Following-Agents-Protocol: true
AI-Assisted: true
Assisted-by: Codex:GPT-5 [Codex]
@localai-bot localai-bot mentioned this pull request Aug 8, 2026
Open
@mudler
mudler force-pushed the row/ENG-RELEASE-BINARIES branch 2 times, most recently from 1f10244 to de124e3 Compare August 9, 2026 00:53
mudler added 6 commits August 9, 2026 01:17
@mudler
feat(release): add versioned binary manifest contract
6e21847 
Rebuild the reviewed W5 schema, deterministic generator/validator, canonical CPU/CUDA fixtures, fail-closed tests, and pending-only lifecycle records as one post-cutover change on current main.

FOLLOWING_AGENTS_PROTOCOL

Following-Agents-Protocol: true
AI-Assisted: true
Assisted-by: Codex:GPT-5 [Codex]
@mudler
fix(ci): repair current-main document and env gates
0e37e7f 
Preserve the rolled Gemma4 evidence in the append-only benchmark record, restore the public keyed-document ratchets, and classify every newly introduced server and ROCm/Gemma environment variable without weakening either checker.

FOLLOWING_AGENTS_PROTOCOL

Following-Agents-Protocol: true
AI-Assisted: true
Assisted-by: Codex:GPT-5 [Codex]
@mudler
fix(ci): classify release schema in PR size gate
4e9bde4 
Classify the versioned release manifest schema explicitly, keep unknown release paths fail-closed, and cover the two existing top-level documentation surfaces so the tracked-tree invariant remains complete.

RED: release/manifest-v1.schema.json raised ValueError as unclassified.
GREEN: test_check_pr_size 21/21.

FOLLOWING_AGENTS_PROTOCOL

Following-Agents-Protocol: true
AI-Assisted: true
Assisted-by: Codex:GPT-5 [Codex]
@mudler
fix(ci): distinguish release contract checker
895258e 
Classify the release binary contract validator as product code rather than a governance-policy checker, while preserving the fail-closed governance classification for every generic check-* entrypoint.

RED: the explicit product-class assertion observed governance_checker.
GREEN: test_check_pr_size 21/21.

FOLLOWING_AGENTS_PROTOCOL

Following-Agents-Protocol: true
AI-Assisted: true
Assisted-by: Codex:GPT-5 [Codex]
@mudler
fix(ci): satisfy explicit release PR size budget
c448041 
Classify release-manifest JSON fixtures as assets, bind the historical agent-record mutation suite, and keep the two W5 product scripts at the exact 900-line review budget without a waiver. Record the green size gate in the keyed public rows.

RED: the actual PR-size gate rejected missing checker evidence and 2222 product lines.
GREEN: focused release/size suites and staged preflight pass; governed product total is 900.

FOLLOWING_AGENTS_PROTOCOL

Following-Agents-Protocol: true
AI-Assisted: true
Assisted-by: Codex:GPT-5 [Codex]
@mudler
fix(ci): classify environment allowlist explicitly
a893c04 
RED: the release PR exceeded its product budget by the allowlist's ten configuration lines. GREEN: the exact-path class and mutation test restore the intended 900-line product total.

FOLLOWING_AGENTS_PROTOCOL

Following-Agents-Protocol: true
AI-Assisted: true
Assisted-by: Codex:GPT-5 [Codex]
@mudler
mudler force-pushed the row/ENG-RELEASE-BINARIES branch from de124e3 to 86b04c2 Compare August 9, 2026 01:34
@mudler
fix(docs): keep release status inside shrink ratchet
d042533 
Compact the OpenAI capability cell into its binding result, lower the shrink-only character ratchet, and record the gate beside the W5 benchmark evidence.

FOLLOWING_AGENTS_PROTOCOL

Following-Agents-Protocol: true
AI-Assisted: true
Assisted-by: Codex:GPT-5 [Codex]
@mudler
mudler force-pushed the row/ENG-RELEASE-BINARIES branch from 86b04c2 to d042533 Compare August 9, 2026 01:39
@localai-bot
localai-bot marked this pull request as ready for review August 9, 2026 02:19
@localai-bot
localai-bot requested a review from mudler August 9, 2026 02:22
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@mudler mudler Awaiting requested review from mudler

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@localai-bot @mudler