huangruiteng · Oct 27, 2026 · Oct 26, 2026 · Oct 26, 2026 · Oct 27, 2026 · Oct 27, 2026
@@ -0,0 +1,59 @@
+# Todo summary-display projection retirement (2026-10-26)
+
+| Boundary | Delivered evidence |
+| --- | --- |
+| Goal/source | Existing TypeScript migration T3/T4 transaction payoff; baseline `c6462f902`. One read-only summary-display slice, with no new capability, RPC operation, provider or runtime. |
+| Observable gap | Python still selected summary/planning sources, normalized display declarations and assembled handoff displays before asking the typed quota owner to select lanes. |
+| Single owner | `todos/context_projection.ts` now owns standalone summary compaction, source order/deduplication, ready-successor overlays and terminal-inclusive frontier preservation through the existing `todo.context.page`. `todos/read_codec.ts` reuses `handoffSource`; quota planning composes the display in process before lane decisions. Both complete planning decisions remain independent. |
+| Retired symbols/callers | Python `_compact_todo_summary_item_record`, `_refs`, `_TEXT_FIELDS` and `handoff_context_source` are deleted. The bodies of `compact_todo_summary_items`, `todo_summary_source_items` and `todo_planning_source_items` become transport facades. `todo_index.py` sends raw record facts for both handoff stages. Existing `todo_summary.py` compact/open callers reach the same raw-fact adapter through `attach_todo_handoff_notes`; they require no new decision branch. `quota_selection.py` removes its display-source join and sends raw display facts. |
+| Retained Python placement | `summary_item.py` retains public import signatures, source-key compatibility exports, empty/bounded transport, response validation and `todo_text_content_revision`. `handoff_note.py` retains generic JSON conversion with Python string/repr facts for tuples, sets, floats and other non-JSON scalars, batching/private snapshots, rejection propagation and mutable-row attachment identity. Existing handoff-gate source IO/compatibility remains. No Python decision fallback or per-field RPC is added. |
+| Facade deletion condition | Delete the summary/handoff transport facades and their raw-fact request carriers after the last Python display, index, quota, management and planning callers consume the existing typed owners in process. Delete `todo_text_content_revision` only after its last compatibility caller exits. This does not retire the other context-page callers, delegation facts, Turn recovery or monitor-poll facades. |
+| Formats/rollback | No persisted format, authority, CAS, writer fence, lease, receipt, provider selection or default changes. Rollback is a code-only revert; no data migration. |
+
+## Cost and real-path evidence
+
+Production source: `summary_item.py` 329 → 100 lines, `handoff_note.py` 150 →
+112 lines; Python −273 net, TypeScript +157 net, production −116 net. The seven
+production files total 242 added / 358 deleted lines. The related refactoring
+pass reuses the existing handoff/read codecs rather than introducing another
+owner or RPC leaf. Larger caller retirement is deferred to the existing exits.
+
+One real-CLI trace per command and fixture size (N=1/N=3) gives:
+
+| Command | Crossings base → candidate | Stdout bytes N=1 | Stdout bytes N=3 |
+| --- | --- | --- | --- |
+| `agent-context --phase before_plan` | 1 → 1 | 1885 → 1885 | 2247 → 2247 |
+| `agent-context --phase after_delegate_result` | 3 → 3 | 1612 → 1612 | 1612 → 1612 |
+| Same result phase with a Turn identity | 3 → 3 | 2594 → 2594 | 2594 → 2594 |
+| `quota should-run` | 75 → 74 | 25520 → 25520 | 25904 → 25904 |
+
+The estimated 73-crossing target is **not met**: the first (empty user) planning
+call has no display rows and therefore no preceding display RPC to remove.
+Only the nonempty Agent planning display crossing is retired. Do not remove a
+second planning decision or absorb an unrelated caller to manufacture the
+estimate. Fewer crossings do not certify latency improvement; the matched
+25-run performance comparison remains external qualification.
+
+The pre-cutover synthetic quota Todo rows and their Python display results are
+pinned in `tests/control_plane_ts/fixtures/todo_summary_display.json`. An external isolated-runtime replay with Python/Node wall clocks pinned matches
+all eight complete CLI outputs byte for byte, including `context_id`, against
+the base Python owners. Independent TS cases cover nested-field precedence,
+legacy identity, Unicode and credential
+exclusion before truncation, bounded/public-safe blocked causes, open-advancement
+hint visibility, normalized declarations, last-outcome precedence, ordered source
+deduplication, explicit false readiness and terminal frontier fields. Python
+retains real File/SQLite CLI recovery, mutation-identity, large/oversize transport
+and fail-closed adapter tests, including tuple/set/float facts.
+
+Focused validation: 89 TS tests pass; typecheck, changed-Python ruff, configured
+mypy (17 files), diff whitespace and the semantic advisory pass. The requested
+Python set reports 467 passed and one temporary-path depth skip, with four
+existing legacy-path failures (three quota legacy-provider
+cases and `test_absent_fence_preserves_legacy_path_without_starting_typescript`),
+reproduced with the base Python owners in the same environment. These remain
+failures, not passing skips. The configured mypy list does not cover the changed
+Python modules. Real installed-wheel qualification, the minimum Node lane and
+isolated PostgreSQL CI are unverified here; no PostgreSQL code is changed. This
+source checkpoint does not certify full CI, installed delivery or a latency gain.
+
+Review correction: summary display now preserves the canonical `decision_scope_outcomes[*].schema_version` as `todo_decision_scope_outcome_v0` by reusing the completion writer’s shared constant, replacing the unregistered `decision_scope_outcome_v0`; TS regression and File/SQLite `quota should-run` coverage were added.
@@ -0,0 +1,52 @@
+# Todo 摘要展示投影退役（2026-10-26）
+
+| 边界 | 交付证据 |
+| --- | --- |
+| 目标／来源 | 既有 TypeScript 迁移 T3/T4 事务收益；基线 `c6462f902`。仅一个只读摘要展示切片，不新增 capability、RPC 操作、provider 或 runtime。 |
+| 可观察缺口 | Python 仍在选择摘要／规划来源、规范化展示声明，并在请求 typed quota owner 选择 lane 前组装 handoff 展示。 |
+| 唯一 owner | `todos/context_projection.ts` 通过既有 `todo.context.page` 统一负责独立摘要压缩、来源顺序／去重、后继就绪覆盖以及含终态的 frontier 字段保留。`todos/read_codec.ts` 复用 `handoffSource`；quota planning 在进程内先组合展示再决定 lane。两次完整规划决策保持独立。 |
+| 删除的符号／调用方 | 删除 Python `_compact_todo_summary_item_record`、`_refs`、`_TEXT_FIELDS`、`handoff_context_source`。`compact_todo_summary_items`、`todo_summary_source_items`、`todo_planning_source_items` 的实现缩为传输 facade。`todo_index.py` 的两个 handoff 阶段改传原始记录事实。`todo_summary.py` 既有 compact/open 调用经 `attach_todo_handoff_notes` 使用同一原始事实适配器，无需新增决策分支。`quota_selection.py` 删除展示来源 join，改传原始展示事实。 |
+| 保留 Python 的放置依据 | `summary_item.py` 保留公开导入签名、来源 key 兼容导出、空批次／有界传输、响应校验及 `todo_text_content_revision`。`handoff_note.py` 保留通用 JSON 转换，为 tuple、set、float 及其他非 JSON 标量提供 Python string/repr 事实，并保留分批／私有快照、拒绝传播及可变行 attachment identity。既有 handoff-gate 来源 IO／兼容适配保留。不新增 Python 决策回退或逐字段 RPC。 |
+| Facade 删除条件 | 最后一个 Python 展示、索引、quota、管理及规划调用方改为在进程内消费既有 typed owner 后，删除摘要／handoff 传输 facade 及原始事实请求载体。`todo_text_content_revision` 仅在最后一个兼容调用方退出后删除。本切片不退役其他 context-page 调用方、委派事实、Turn recovery 或 monitor-poll facade。 |
+| 格式／回滚 | 不改变持久化格式、authority、CAS、writer fence、lease、receipt、provider 选择或默认行为。仅回退代码即可，无需数据迁移。 |
+
+## 成本与真实路径证据
+
+生产源码：`summary_item.py` 从 329 降至 100 行，`handoff_note.py` 从 150
+降至 112 行；Python 净减 273 行，TypeScript 净增 157 行，生产代码净减
+116 行。七个生产文件共增加 242／删除 358 行。相关重构复用既有 handoff／read
+codec，不新增 owner 或 RPC leaf。更大范围调用方退役沿用既有退出条件。
+
+对每条命令、每个夹具规模（N=1／N=3）各执行一次真实 CLI 跟踪，结果如下：
+
+| 命令 | 跨运行时调用：基线 → 候选 | Stdout 字节 N=1 | Stdout 字节 N=3 |
+| --- | --- | --- | --- |
+| `agent-context --phase before_plan` | 1 → 1 | 1885 → 1885 | 2247 → 2247 |
+| `agent-context --phase after_delegate_result` | 3 → 3 | 1612 → 1612 | 1612 → 1612 |
+| 同一 result 阶段，带 Turn identity | 3 → 3 | 2594 → 2594 | 2594 → 2594 |
+| `quota should-run` | 75 → 74 | 25520 → 25520 | 25904 → 25904 |
+
+预估的 73 次目标**尚未达到**：首次规划针对空 User 来源，没有展示行，因此原本
+就不存在可删除的前置展示 RPC。仅退役非空 Agent 规划的展示调用。不得通过删除
+第二次规划决策或合并无关调用方来凑足预估。调用减少不证明延迟改善；同条件
+25 次性能对照仍由外部独立验收。
+
+退役前的合成 quota Todo 行及其 Python 展示结果固定在
+`tests/control_plane_ts/fixtures/todo_summary_display.json`。在外部隔离 runtime 中固定 Python／Node 墙钟后，八条完整 CLI 输出与重放基线
+Python owner 的输出逐字节一致（包括 `context_id`）。独立 TS 用例覆盖
+嵌套字段优先级、旧 identity、Unicode 与截断前凭据排除、有界且公开安全的阻塞
+原因、仅 open advancement 可见的 hint、声明规范化、同 scope 最后 outcome 优先、
+有序来源去重、显式 false readiness 以及终态 frontier 字段。Python 保留真实
+File／SQLite CLI 恢复、可变行 identity、大／超大传输及 fail-closed 适配测试，
+包括 tuple／set／float 事实。
+
+聚焦验证：89 个 TS 测试通过；typecheck、变更 Python 的 ruff、配置内 mypy
+（17 个文件）、diff 空白检查和语义 advisory 通过。要求的 Python 集合为 467 个通过、1 个临时路径深度限制的 skip，并存在四个
+既有 legacy 路径失败（三个 quota legacy-provider 用例，以及
+`test_absent_fence_preserves_legacy_path_without_starting_typescript`），在相同
+环境加载基线 Python owner 后复现。它们仍是失败，不计为通过的 skip。mypy
+配置文件列表不覆盖本次修改的 Python 模块。真实 installed-wheel、最低 Node
+版本及隔离 PostgreSQL CI 尚未验证；未修改 PostgreSQL 代码。本源码检查点不
+认证完整 CI、已安装交付或延迟收益。
+
+审查修正：摘要展示复用 completion writer 的共享常量，将 `decision_scope_outcomes[*].schema_version` 保持为 canonical 记录中的 `todo_decision_scope_outcome_v0`，替换未注册的 `decision_scope_outcome_v0`；新增 TS 回归测试及 File／SQLite `quota should-run` 覆盖。
@@ -0,0 +1,73 @@
+# Todo summary-display follow-up qualification (2026-10-27)
+
+This entry supplements the [original retirement record](2026-10-26-todo-summary-display-projection.md).
+The production boundary remains summary display through `todo.context.page` and
+in-process quota planning; both complete planning decisions remain necessary.
+There is no new provider, format, permission or RPC owner.
+
+## Preserve the earlier evidence
+
+The independent review of head `c242ca5f439c269c313329c883299a28d95a55dc`
+against `be1dc37a008be7017b64b622f653b6707df5c544` found:
+
+- the registered `todo_decision_scope_outcome_v0` correction passed canonical
+  File/SQLite source and installed wheel/sdist checks;
+- one SQLite CLI fixture used 83 → 82 requests, including 22 → 21
+  `todo.context.page` calls. Its workload differs from the original N1/N3
+  75 → 74 trace; neither trace permits removing another planning decision;
+- a fixed 32-pair CLI cohort had base/head p50/p95 336.97/521.98 and
+  345.96/836.79 ms, a 60.31% candidate p95 increase;
+- the predeclared 64-pair rerun had 301.56/400.98 and 305.88/391.08 ms.
+  This later result does not erase the earlier failed cohort. The difference
+  did not establish a repeatable regression or explain the earlier tail;
+- cold startup, warm typed requests and idle/burst daemon RSS were still missing.
+  Python thread profiling did not measure Node CPU or establish tail causality.
+
+These historical cohorts keep their original status. They are not relabelled
+as passes, and the default SQLite limits remain unchanged.
+
+## Rebase and real recovery
+
+The follow-up baseline is `ff8d9932294e9004b9ffeb2fb473d00354397eb8`.
+Six File/SQLite recovery test cases initially stopped on
+`canonical_authority_required`, before their intended assertions. The fixture
+now initializes the real provider from authored source facts **before** calling
+the production CLI; it does not restore a legacy read fallback. All eight
+receipt/next-action cases pass, including version preservation, rejection of
+wrong Todo step updates, legitimate original-Turn settlement and replay.
+The earlier six failures remain recorded as fixture failures, not proof of
+previously passing recovery behavior.
+
+## Final-head measurement contract
+
+[PR #7411](https://github.com/loopx-project/loopx/pull/7411) records the immutable
+final head, execution commands, fixture fingerprint, environment and results.
+Its bounded offline qualification uses the same synthetic canonical SQLite
+Todo and interpreter on both source revisions, with no active Goal mutation:
+
+- 24 managed-process cold-start pairs, distinguishing warm filesystem/compile
+  cache from a fresh process;
+- 64 complete `quota should-run` CLI pairs with alternating AB/BA order, all
+  samples retained and nearest-rank p50/p95;
+- 256 warm typed requests and 256 complete quota-planning projections per
+  revision, separating transport from the Python facade;
+- idle, sampled burst and post-burst daemon RSS over 2,000 bounded projections;
+- real CLI identity/result assertions, focused TS/Python coverage and packaged
+  readback, reported separately from timing.
+
+Do not overlap owned heavy builds/tests with timed cohorts. Keep private raw
+captures outside Git and publish only synthetic aggregates. A single synthetic
+workload does not establish sustained large-Goal performance. Any miss retains
+an owner-review hold; a faster microbenchmark cannot waive a full-CLI failure.
+
+## Rebase onto the subsequently merged main
+
+The final integration base advances to
+`17b2a0f9b818dc6e6fe77ccc567b20ff91b3bc40`. Main now includes #7459's
+canonical fixture repair and changes to the real quota/context path. The
+conflict resolution keeps that shared fixture repair instead of duplicating it;
+this PR retains the File/SQLite receipt-version regression. The earlier
+`ff8d9932294e9004b9ffeb2fb473d00354397eb8` comparison and
+`e7a54b8f32a648124133b02e11404b75a421fdac` head remain historical evidence.
+The final PR reply records fresh exact-head correctness, package and matched
+performance results for the new base. No earlier cohort is rewritten.
@@ -0,0 +1,57 @@
+# Todo 摘要展示后续验收（2026-10-27）
+
+本记录补充[原退役记录](2026-10-26-todo-summary-display-projection.zh-CN.md)。生产边界仍是
+`todo.context.page` 的摘要展示，以及 quota planning 内的直接组合；两个完整规划决策
+都仍有必要。不新增 provider、格式、权限或 RPC owner。
+
+## 保留此前证据
+
+独立评审对照 `be1dc37a008be7017b64b622f653b6707df5c544` 与
+`c242ca5f439c269c313329c883299a28d95a55dc` 得到：
+
+- 已注册 `todo_decision_scope_outcome_v0` 的修正通过真实 File/SQLite 来源及
+  wheel/sdist 安装验证；
+- 一个 SQLite CLI fixture 请求为 83 → 82，其中 `todo.context.page` 为 22 → 21。
+  它与原 N1/N3 的 75 → 74 负载不同；两者均不支持删除另一个规划决策；
+- 固定 32 对 CLI 样本的基线/候选 p50/p95 为 336.97/521.98、345.96/836.79 ms，
+  候选 p95 增加 60.31%；
+- 预先声明的 64 对复测为 301.56/400.98、305.88/391.08 ms。后一次结果不能抹掉
+  前一次失败；不一致尚未证明可重复回归，也没有解释先前尾延迟；
+- 冷启动、温态 typed request、daemon 空闲/突发 RSS 仍缺失。Python 线程 profile
+  没有测得 Node CPU，也未证明尾延迟成因。
+
+上述历史批次保留原结论，不改写成通过，默认 SQLite 限值也不放宽。
+
+## Rebase 与真实恢复
+
+本次后续工作的基线为 `ff8d9932294e9004b9ffeb2fb473d00354397eb8`。六个 File/SQLite
+恢复用例最初在目标断言前因 `canonical_authority_required` 失败。现在 fixture 先用
+已写好的合成来源初始化真实 provider，再调用生产 CLI；没有恢复旧读取 fallback。
+八个 receipt/next-action 用例全部通过，覆盖版本保留、错误 Todo 步骤更新拒绝、
+原 Turn 合法结算与重放。之前六项保留为 fixture 失败，不声称先前已验证恢复。
+
+## 最终 head 测量契约
+
+[PR #7411](https://github.com/loopx-project/loopx/pull/7411) 记录不可变最终 head、执行命令、
+fixture 指纹、环境与结果。有界离线验收在两个修订上使用相同解释器和合成 canonical
+SQLite Todo，不修改活跃 Goal：
+
+- 24 对受管进程冷启动，区分已温热的文件系统/编译缓存和新进程；
+- 64 对完整 `quota should-run` CLI，AB/BA 交错、不删样本，按 nearest-rank 算 p50/p95；
+- 每版本 256 次温态 typed request、256 次完整 quota-planning 投影，区分 transport
+  与 Python facade；
+- 2,000 次有界投影的 daemon 空闲、突发采样峰值和突发后 RSS；
+- 真实 CLI 身份/结果断言、聚焦 TS/Python 与安装回读，与耗时分别报告。
+
+计时期间不并行执行本 Agent 的重型构建/测试。原始私有记录留在 Git 外，只发布合成
+汇总。单个合成负载不能证明持续大 Goal 性能。任何未达标保留 owner-review hold；
+微基准更快不能豁免完整 CLI 失败。
+
+## 同步随后合入的主线
+
+最终集成基线更新为 `17b2a0f9b818dc6e6fe77ccc567b20ff91b3bc40`。主线已包含
+#7459 的 canonical fixture 修复，也改动了真实 quota/context 路径。冲突解决沿用
+共同 fixture 修复，不重复补丁；本 PR 保留 File/SQLite 的 receipt-version 回归。
+此前 `ff8d9932294e9004b9ffeb2fb473d00354397eb8` 对照和
+`e7a54b8f32a648124133b02e11404b75a421fdac` head 保留为历史证据。最终 PR 回复
+记录新基线与新精确 head 的正确性、安装和匹配性能复测，不改写旧批次。