docs(maintainer): handoff updated to the final state, with the production record by Ikalus1988 · Pull Request #1795 · Ikalus1988/MisakaNet

Ikalus1988 · 2026-09-17T07:36:06Z

User description

Handoff updated to the state after every workstream landed, plus the first production verification.

§3 rewritten — no longer "in flight": each of the seven workstreams that merged, with the one-line reason it matters, and the hunk-splitting method recorded for next time (the same one-file/one-owner rule that came out of today's incident).

§7 rewritten — what is genuinely next, in order: the two maintainer actions (first /adopt, and a token-bearing production re-verification), waiting for the hit-rate denominator to have data before B1 can be evaluated at all, the two deliberate tails of #1780 (FAQ expansion; trigger in the index text, which needs an INDEX_TEXT_VERSION bump and therefore a full rebuild — to be weighed, not done by reflex), the disposition of #1713–#1716, and the six verified-but-unclaimed severe findings from the misakanet/ scan in #1775.

New §8 — production record, honestly split:

query	result
`如何切换识图模型`	✅ two hits, correct lesson on top — this returned zero before #1794, so the acceptance point passed
`定时任务不执行`	⚠️ hits, but the top one is wrong: aliasing buys recall, not precision for every phrasing
two more probes	❓ `hits=0`, consistent with the anonymous 5-reads/day/IP quota being used up by the probe itself; the with-token retry hit TLS timeouts here. Recorded as 未判定, with copy-pasteable commands and the token path for the next attempt

That distinction is deliberate: this stretch already cost hours because a green signal was hiding a real failure, so an inconclusive probe is written down as inconclusive.

Two process lessons added, both paid for today:

never git checkout <ref> -- <file> in a shared working tree to get a clean copy — it erased another agent's uncommitted work here; use a separate worktree or git show <ref>:<path>;
"deploy succeeded" is a statement about plumbing, not behaviour — the only verdict is a real query against production.

PR Type

Documentation, Other

Description

Rewrite handoff §3 from "在途" to "全部落地" with seven landed workstreams
Add new §8 recording first production verification of alias wiring
Document two process lessons on shared worktree and deploy vs behaviour
Rewrite §7 as concrete next steps ordered by dependency

Diagram Walkthrough

flowchart LR
  A["§3 在途 → 全部落地<br/>七个工作流汇总"] --> B["§4 维护者介入<br/>首次 /adopt 待做"]
  B --> C["§6 流程教训<br/>+2 条"]
  A --> D["§7 真正的下一步<br/>按依赖排序"]
  A --> E["§8 新增<br/>生产验证记录"]
  E --> F["✅ 验收点达成<br/>如何切换识图模型"]
  E --> G["⚠️ 召回≠精度<br/>定时任务不执行"]
  E --> H["❓ 未判定<br/>配额耗尽 + TLS 抖动"]

File Walkthrough

Relevant files

Documentation

handoff-2026-09-16.md `Handoff final state + production verification record` docs/maintainer/handoff-2026-09-16.md Header rewritten to span 09-16 → 09-17 with summary of "六个工作流全部落地" §3 rewritten from "在途" (in flight) to "全部落地" listing seven landed workstreams with key points and hunk-splitting method §4 adds a 09-17 update note: PAT not needed for `/adopt`; `auto-merge-lesson` label created §6 adds two process lessons: never `git checkout --` in shared worktrees; "deploy succeeded" ≠ behaviour verified §7 rewritten as five concrete next steps ordered by dependency (not as PR-merge checklist) New §8 documents the first production verification of alias wiring (`如何切换识图模型` ✅, `定时任务不执行` ⚠️, two probes ❓未判定) with copy-pasteable retry commands	+53/-17

…tion record Every workstream that was "in flight" when this file was written has landed, so §3 is rewritten as what shipped (with the one-line reason each matters), §7 as what is genuinely next, and a new §8 records the first production verification of the alias wiring: - `如何切换识图模型` — a plain Chinese question that returned **zero** hits before #1794 — now returns two, with the correct lesson on top. That was the acceptance point, and it passed. - `定时任务不执行` returns hits but the **top one is wrong**: aliasing buys recall, not precision for every phrasing. - Two further probes were inconclusive (`hits=0`), consistent with the anonymous 5-reads/day/IP quota being used up by the probe itself, and the with-token retry hit TLS timeouts from this machine. Recorded as **未判定**, with the exact commands to re-run with a token — because the lesson of this stretch is precisely that a green signal hid a real failure for hours. Two process lessons added, both paid for today: never use `git checkout <ref> -- <file>` in a shared working tree to get a clean copy (it erased another agent's uncommitted work here; use a separate worktree or `git show <ref>:<path>`), and treat "deploy succeeded" as a statement about plumbing rather than behaviour — the only verdict is a real query against production. Signed-off-by: Ikalus1988 <136884451+Ikalus1988@users.noreply.github.com>

cloudflare-workers-and-pages · 2026-09-17T07:36:19Z

Deploying with Cloudflare Workers

The latest updates on your project. Learn more about integrating Git with Workers.

Status	Name	Latest Commit	Preview URL	Updated (UTC)
✅ Deployment successful! View logs	misakanet-web	`7eb45a0`	Commit Preview URL Branch Preview URL	Sep 17 2026, 07:36 AM

Ikalus1988 · 2026-09-17T07:37:51Z

🧾 Audit Report — PR #1795 (`7eb45a0`)

📊 Quality Score

⚠️ Quality score unavailable; continuing with hard gates.

🔏 DCO Audit

✅ All commits signed-off.

📏 PR Size

Metric	Value
Files Changed	1
Lines Added	53

🔐 Secret Scan

✅ No hardcoded secrets detected.

📦 Dependency Audit

⏭️ Skipped; no Python/JS dependency files changed.

🧪 Test Suite

✅ PASS — 53% coverage

📋 Lesson Schema

✅ All lessons valid.

⚖️ Verdict

✅ All gates passed. Ready for merge.

Scope: full | Triggered by 7eb45a0 | View run

github-actions · 2026-09-17T07:38:35Z

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

🎫 Ticket compliance analysis ❌ 1780 - Partially compliant Compliant requirements: Documents (handoff §3) that [Retrieval] 接线近义词表：worker + 本地 CLI，含 INDEX_TEXT_VERSION 决策 #1780 has landed and reports the key constraint (INDEX_TEXT_VERSION kept at 3 because expansion is query-side) Notes in §7.3 the two remaining tails (FAQ matcher does not yet expand; `trigger` not yet merged into BM25 index text — needs version bump + rebuild, intentionally deferred) Non-compliant requirements: No actual wiring code added in this PR (implementation is reported as already merged in PR feat(retrieval): wire the alias table into both search paths, and unify two expansion layers #1794, which is not in this diff) No env-var switch, no `MISAKANET_QUERY_ALIASES` reference, no unit/regression test changes No three-step rollback instructions in this PR description Requires further human verification: Verify that the referenced PR feat(retrieval): wire the alias table into both search paths, and unify two expansion layers #1794 actually contains the wiring implementation and matches all acceptance criteria (env-var name, default-on, rollback path) Verify that the commit SHA `8d362241a` cited in §8 corresponds to the documented merge state Verify `INDEX_TEXT_VERSION` is in fact still 3 in production after the wiring merged 1713 - Not compliant Non-compliant requirements: No lesson file added or modified by this PR; ticket is only mentioned in §7.4 as "pending disposition" 1716 - Not compliant Non-compliant requirements: No lesson file added or modified by this PR; ticket is only mentioned in §7.4 as "pending disposition"
⏱️ Estimated effort to review: 1 🔵⚪⚪⚪⚪
🧪 No relevant tests
🔒 No security concerns identified
⚡ Recommended focus areas for review Factual claim to verify The doc references PR #1794 as the one that wired the synonym table (the work ticket #1780 asked for). Since this PR is purely a status doc and contains none of the wiring code, an experienced reviewer should independently confirm that #1794 (a) actually exists, (b) is merged, (c) implements the `MISAKANET_QUERY_ALIASES` switch with default-on, (d) keeps `INDEX_TEXT_VERSION` at 3, and (e) covers both worker and `search_knowledge.py`. Otherwise the handoff's "已合并" claim in §3 row 7 is unsupported by anything in this diff. The same caveat applies to the specific commit SHA `8d362241a` and deploy run "143" cited in §8. \| #1780 接线近义词表 \| 已合并 \| worker + 本地 CLI 同表同开关；统一旧 `_SYNONYM_MAP`（30 条迁移为 `related` 降级层）；`INDEX_TEXT_VERSION` 保持 3（查询期展开，索引文本未变） \| Cross-reference to verify §7.3 and §8 point to `docs/maintainer/query-alias-design-2026-09-16.md §8.7` for the rationale on FAQ expansion and `trigger` indexing. If that section does not actually exist (file may not yet have §8.7, or may be named differently), readers following the breadcrumb will hit a dead end. Confirm the target section exists before merging. 3. #1780 的两个尾巴：FAQ 匹配器是否跟随展开；`trigger` 是否并入 BM25 索引文本（并入需要 bump `INDEX_TEXT_VERSION` 并承担一次全量重建——要把这个代价和收益一起评估，别顺手做）。 4. #1713–#1716 的处置：改一行 provenance 即可合并；对方不响应时决定关闭还是补写。 5. 零散但值得：`workers/register-proxy-sw.js`（4475 行）的 ocr 切片扫描结果在 issue #1775，其中 6 条已核实的严重项仍待认领（freshness 列表解析、heal.py 写原始日志、remote.py 路径拼接等）。 --- ## 8. 生产验证记录（2026-09-17 上午，首次） `#1794`（接线近义词表）合并后 worker 自动部署成功（deploy-worker run 143 @ `8d362241a`），随即在生产端点 `https://misakanet.org/mcp` 上验证： \| 查询 \| 结果 \| 读法 \| \|---\|---\|---\| \| `如何切换识图模型` \| ✅ 2 条命中，top = 《多模型Switch脚本模式 — 双 Agent 模型管理》 \| 这句纯中文问句在接线前返回 0 条。这是 #1794 的验收点，已达成 \| \| `定时任务不执行` \| ⚠️ 有命中（2 条）但 top 是错的（GBK 编码那篇，不是 cron 那篇） \| 别名给的是召回，不是这个句式下的精度；与评测中"仍有 4 条失败"一致 \| \| `switch vision model` / `如何切换视觉模型` \| ❓ `no_match=False hits=0` \| 与匿名配额（5 次/天/IP）被本轮探测用尽的表现一致；改用本机 token 复验时 `misakanet.org` 出现 TLS 握手超时（本会话反复出现，属网络抖动），未判定 \| 下次复验请带 token（本机可用：`/tmp/pub-cc/.misakanet-agent/token` 或安装器写入的 `~/.misakanet-agent/token`），并顺手确认： ```bash curl -sS "https://misakanet.org/api/search-signals/stats?days=7" \| python3 -m json.tool # 期望 200 + rows 非空且在增长 curl -sS https://misakanet.org/mcp -H 'Content-Type: application/json' -H 'Accept: application/json' \ -H 'MCP-Protocol-Version: 2025-06-18' -H 'Origin: https://misakanet.org' \ -H "Authorization: Bearer $TOKEN" \ -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"misakanet_search","arguments":{"query":"定时任务不执行","top":2}}}' 不要把"两次无法判定"写成通过——这一轮已经因为"绿灯掩盖真红"付过代价（见 §6）。仍未做的：`#1780` 的 FAQ 匹配器没有跟着展开（阈值语义会变，离线评测看不到 FAQ 语料）；`trigger` 尚未并入 BM25 索引文本。两条都写在 `docs/maintainer/query-alias-design-2026-09-16.md` §8.7。 </details> </td></tr> </table> <!-- pr-agent-review-state:v1 {"findings":[{"body":"Factual claim to verify The doc references PR #1794 as the one that wired the synonym table (the work ticket #1780 asked for). Since this PR is purely a status doc and contains none of the wiring code, an experienced reviewer should independently confirm that #1794 (a) actually exists, (b) is merged, (c) implements the `MISAKANET_QUERY_ALIASES` switch with default-on, (d) keeps `INDEX_TEXT_VERSION` at 3, and (e) covers both worker and `search_knowledge.py`. Otherwise the handoff's \"已合并\" claim in §3 row 7 is unsupported by anything in this diff. The same caveat applies to the specific commit SHA `8d362241a` and deploy run \"143\" cited in §8.","finding_id":"12643064cbe5","first_seen":"2026-09-17T07:38:34.421088Z","last_seen":"2026-09-17T07:38:34.421088Z","last_seen_head_sha":"7eb45a0496c3887c264b28009cb58f2db3b9526b","line_end":48,"line_start":48,"path":"docs/maintainer/handoff-2026-09-16.md","state":"ACTIVE"},{"body":"Cross-reference to verify §7.3 and §8 point to `docs/maintainer/query-alias-design-2026-09-16.md §8.7` for the rationale on FAQ expansion and `trigger` indexing. If that section does not actually exist (file may not yet have §8.7, or may be named differently), readers following the breadcrumb will hit a dead end. Confirm the target section exists before merging.","finding_id":"d72dcc215899","first_seen":"2026-09-17T07:38:34.421088Z","last_seen":"2026-09-17T07:38:34.421088Z","last_seen_head_sha":"7eb45a0496c3887c264b28009cb58f2db3b9526b","line_end":116,"line_start":88,"path":"docs/maintainer/handoff-2026-09-16.md","state":"ACTIVE"}],"last_run":{"complete":true,"excluded_files":[],"head_sha":"7eb45a0496c3887c264b28009cb58f2db3b9526b","kind":"full","run_id":"https://github.com/Ikalus1988/MisakaNet/commit/7eb45a0496c3887c264b28009cb58f2db3b9526b"},"schema_version":1} -->

github-actions · 2026-09-17T07:38:47Z

PR Code Suggestions ✨

No code suggestions found for the PR.

github-actions · 2026-09-17T07:39:21Z

✅ Merged! Thanks again, @Ikalus1988.

docs(maintainer): handoff updated to the final state, with the production record (+53 lines, 1 files)

Quick question — did any MisakaNet lesson help you this time?
→ Share feedback

No need to reply if nothing comes to mind. ⚡

github-actions · 2026-09-17T07:39:22Z

🎉 Merged — Thank you!

Your contribution has been merged into main.

PR: #1795 — docs(maintainer): handoff updated to the final state, with the production record

What's next:

Your code is now part of MisakaNet's failure-lesson corpus (now 393 lessons)
Feel free to pick up another issue labeled good first issue or status: competition
Questions? Ask in this thread or open a Discussion

Welcome to the MisakaNet contributor community! 🧠

Ikalus1988 added area:docs docs-only Only documentation files changed shape-safe labels Sep 17, 2026

github-actions Bot added the Review effort 1/5 label Sep 17, 2026

Ikalus1988 merged commit 835b260 into main Sep 17, 2026
26 checks passed

Ikalus1988 mentioned this pull request Sep 19, 2026

chore(main): release 2.31.0 #1862

Merged

Navigation Menu

docs(maintainer): handoff updated to the final state, with the production record - #1795

docs(maintainer): handoff updated to the final state, with the production record#1795
Ikalus1988 merged 1 commit into
mainfrom
docs/handoff-final

Ikalus1988 commented Sep 17, 2026 •

edited by github-actions Bot

Loading

Uh oh!

cloudflare-workers-and-pages Bot commented Sep 17, 2026

Uh oh!

Ikalus1988 commented Sep 17, 2026

Uh oh!

github-actions Bot commented Sep 17, 2026

Uh oh!

github-actions Bot commented Sep 17, 2026

Uh oh!

Uh oh!

github-actions Bot commented Sep 17, 2026

Uh oh!

github-actions Bot commented Sep 17, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

Ikalus1988 commented Sep 17, 2026 • edited by github-actions Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

User description

PR Type

Description

Diagram Walkthrough

File Walkthrough

Uh oh!

cloudflare-workers-and-pages Bot commented Sep 17, 2026

Deploying with Cloudflare Workers

Uh oh!

Ikalus1988 commented Sep 17, 2026

🧾 Audit Report — PR #1795 (7eb45a0)

📊 Quality Score

🔏 DCO Audit

📏 PR Size

🔐 Secret Scan

📦 Dependency Audit

🧪 Test Suite

📋 Lesson Schema

⚖️ Verdict

Uh oh!

github-actions Bot commented Sep 17, 2026

PR Reviewer Guide 🔍

Uh oh!

github-actions Bot commented Sep 17, 2026

PR Code Suggestions ✨

Uh oh!

Uh oh!

github-actions Bot commented Sep 17, 2026

Uh oh!

github-actions Bot commented Sep 17, 2026

🎉 Merged — Thank you!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Ikalus1988 commented Sep 17, 2026 •

edited by github-actions Bot

Loading

🧾 Audit Report — PR #1795 (`7eb45a0`)