[cocolord]

Goal And Delivered Outcome

The small-kernel RFC had separately measured prequota and settlement savings without a complete CLI comparison. This checkpoint compares main 8e86f173f08bd55eb833d8cdf42360f93bf989fe with the private #8240@4df990c9e66465171f980176fba4e3602d41f826 + #8243@668bed2f03a274f273d39c187121a1a4ede67165 production overlay.

All 60 frozen observations finish within 1,500 seconds. The 52 small-state lifecycles complete with identical obligations/artifact and one writeback/debit; median episode latency improves about 11–12%. All eight large-state observations retain the same todo_archived terminal failure after automatic archival. The checkpoint distinguishes these outcomes, complete-cost limits and the existing terminal/freshness owner for follow-through.

• Intended base: main.
• Written by: model_agent — GPT-6 via TRAE.
• Specification: docs/architecture/rfcs/monorepo-distribution-split-v0.md at 8e86f173f08bd55eb833d8cdf42360f93bf989fe, I9 and scoped V10 characterization.

Criterion	Disposition	Evidence
I9: named composition, baseline, costs and outcome scope	implemented	Exact heads, frozen cache/order/instrumentation, source/wheel File/SQLite result and cost tables
Bilingual checkpoint and public/private boundary	implemented	Both RFC files; raw captures, temporary instrumentation, environments and source overlays excluded
Full R0/V10/R1, large-state lifecycle and installed/split-host qualification	deferred	Eight retained terminal failures, wheel/SQLite uncertainty, unmeasured model/human/long-horizon costs; no adoption claim

Scope And Continuation

Only the existing bilingual RFC changes. No production implementation, defaults, provider/authority, host installation or budgets change. The independently reversible research checkpoint informs exact-head review of #8240/#8243 and the existing #5677 terminal/freshness integration; it does not duplicate those implementations or grant merge authority. Original-receipt archived closeout needs qualification before the full large-state path can pass. Subsequent performance work should characterize the remaining quota should-run caller.

Validation

• Tested revision: fed8158b9543cd1a8ae25f2c255502511c21ab55; source/wheel experimental runtime heads are named above.
• Run state: finished. Inputs: synthetic and public_fixture.

Check kind	Result	Evidence / limitation
real_entrypoint / regression_parity	passed	52 complete source/wheel File/SQLite lifecycles; contiguous native reads, hard lease, independent Git worktree, JSON sum 45, exact completion/writeback/spend/terminal retries, one writeback/debit and clause-for-clause settlement obligations
real_entrypoint	failed	Eight large-state observations reject original-Turn terminal closeout after automatic archival; both arms fail; these are not accepted episodes
real_backend	passed	Isolated real Node 26.0.0 / SQLite 3.53.1 and File stores; owned brokers have shutdown and OS-exit readback
integration	passed	Eight variants × four admission/source/claim negatives; invalid Vision/workspace refusals in the separate census; no writeback/debit on negative paths
static / packaging	passed	Both monolithic wheels independently installed; 1,836 runtime resources byte-verified each; 57 identical dependency versions; frozen inventories unchanged
static	passed	uv run --extra test python scripts/check_rfc_bilingual.py --base origin/main --staged and exact committed --head HEAD; diff hygiene and public boundary scan
integration	passed	Final standard premerge on unchanged fed8158b9: four direct checks and all 11 selected checks passed (two catalog, eight risk-profile, one public-boundary). Initial semantic-vocabulary failure was missing repository TypeScript dev dependencies; after npm ci --ignore-scripts, the complete gate was rerun. The generated untracked lock included in the scan is excluded from this PR. Passing validation grants no merge authority.
manual	not_run	Actual model/native managed-host, standard external fanout, full process-tree costs and split-package install; public aggregate/recipe is not an independent public reproduction receipt

Raw evidence and temporary measurement tools remain private. The RFC provides the workload, existing fixture owners, executable CLI recipe, sample uncertainty and limits. This is a documentation proposal ready for review with the initial environment failure and eight lifecycle failures explicitly retained; no merge or installed-host adoption is requested.

Frontend / Visual Evidence

UI impact: none. This appends RFC research content and does not change documentation chrome or a product interaction.

• Documentation update
• Public docs or presentation surface

[mergify]

This pull request has merge conflicts with main and cannot be merged
until they are resolved. Please rebase or merge the base branch, @cocolord.

Choose the remote for the base repository, not an out-of-date fork.
For a fork clone, first inspect git remote -v; upstream must point
to https://github.com/loopx-project/loopx.git. If it is absent, add it
with git remote add upstream https://github.com/loopx-project/loopx.git.
Then run:

git fetch upstream
git rebase upstream/main
# Resolve each conflict, git add the resolved files, then git rebase --continue.
git push --force-with-lease origin HEAD

For a same-repository clone whose origin points to
https://github.com/loopx-project/loopx.git, use origin instead of
upstream for fetch/rebase. If you prefer merging the base, use
git merge <base-remote>/main and push normally.

Keep the DCO Signed-off-by trailer on every commit when you rebase.
https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork

[loopx-agent]

Reviewer: model_agent | gpt-6.1-sol | OpenAI | runtime_reported | reasoning_effort=max

复核提交 fed8158。文档内容可接受，当前主干的合并冲突已单独核实。

动机 / Motivation

维护者要判断两项配额优化接在一起后，能省多少时间，完整流程还能走到哪一步。此前的单命令测量回答不了归档后的终结。 这份记录把小状态的成功轮次和大状态的失败分开，并分别列出 CLI、准备和退出的成本。 独立复测中，File 和 SQLite 的小状态都能终结。增加 100 个已完成 Todo 后，原 Turn 已写回并扣额一次，终结仍报 todo_archived。 证据范围是隔离的 CLI 合成任务。实际宿主采用、完整成本和长程效果仍待验。 作者原始计时未独立复算，跨版本和拆包安装资格仍开放。

改动思路 / Approach

两份附录把同一份对照固定在具体源码、运行时和测试数据上。作者复用原配额、终结和归档路径，明确记下大状态的失败。七个生产 Python 文件的两组补丁均能对记录的基线应用，提交本身只有中英文附录。 英文附录 中文附录

具体改动 / Changes

四组小状态分别有 8、6、6、6 对，共 52 个成功轮次。大状态四对共八次失败，单独列出。两种语言的表格数值和命令一致。约 11–12% 是小状态 CLI 轮次的两臂中位数差；CLI 样本中 wheel/SQLite 的配对区间跨零。计入准备和退出后，wheel/SQLite 的两臂中位数相差约 3%。现有样本量的 p95 等于观测最大值。

按基线 8e86f17 的 I9 核对装配、版本与适用范围。全量 V10 和实际 R1 仍是后续验收。 规格

对主干的风险 / Risks

在隔离的源码 File 和 SQLite 上用 hard lease 复测原生 CLI。小状态可终结并重放；增加 100 条完成记录后，终结拒绝，原写回和扣额各一次。三项既有结算测试、快速 premerge 和双语检查通过。按策略未查 CI。

以主干 05a33c2 运行严格 CQR 时，两份附录的合并都冲突。#8203 合入后的主干 4437696 仍复现这两处冲突。#8240 后来已用更新的提交入主干，实验数字仍归原测试版本。 主干集成 #8240 合并

我的整体评价 / Verdict

改动位于现有中英文 RFC 附录。 这份记录保留了能复现的终结失败，也保留了计时所覆盖的范围。合并前需在最新主干上协调两份附录并重新核验。实际归档修复仍由既有终结路径承接。 相关修复

English verdict: APPROVE. The scoped evidence and archived closeout failure are documented accurately at fed8158. Both RFCs have a separate merge conflict with the current target.

[loopx-agent]

APPROVE — exact head 2c3f3cc8894c0fdd31a68c8c3cde7b8e017f66dd.

Re-reviewed after integrating current main f1d30b2f.

• The merge only resolves the RFC append location; all scoped CLI composition, cost, negative-case, and archived-closeout evidence is preserved in both languages.
• No production or test source changed.
• Exact-head bilingual RFC verification and diff hygiene pass; the merge commit carries DCO.

Approval is limited to merge readiness at this exact head.