Skip to content

[PR] wave-11 batch: 7 community fixes (opencode provider, preflight epoch, output budget, +4) — #6797/#6818 excluded - #6827

Merged
ranxianglei merged 23 commits into
masterfrom
batch11-wave11
Nov 18, 2026
Merged

ranxianglei merged 23 commits into
masterfrom
batch11-wave11

Conversation

@ranxianglei

@ranxianglei ranxianglei commented Nov 15, 2026 •

Copy link
Copy Markdown
Owner

Wave-11 batch: 7 community fixes — opencode provider routing, incremental-timestamp journal, preflight epoch pause, output-budget restore, and 3 more.

Author commits are kept verbatim (five PRs land as real two-parent merge commits; #6775/#6772/#6766 as their plain commits after the rebase onto current master — same bytes either way). Per-PR review done; independent reviewer passes over the largest (#6773). #6773 is not in this batch — held pending a fix or explicit ack (detailed comment on that PR). #6797 (pi thread negotiation) and #6818 (incremental-timestamp journal) are excluded per owner preference this round; both are reviewed and correct and can land separately.

PR Author What it fixes Closes
#6819 sx4im P1: OpenCode 2.x's built-in Claude Agent provider (claude-agent/*) drives a local Claude Code CLI that calls Anthropic directly, so the HTTP-hook routing never sees it — silently uncompressed. bili plugin install opencode now routes it via providers.claude-agent.settings.baseURL (live lane port, merge-preserving, never clobbers a foreign baseURL); doctor warns on unrouted/foreign/stale-port states (2.x only). #6816
#6807 Maziar123 P1: once the epoch-confirm latch (#3641) is set and history grows past the model window, every turn is refused (preflight_compress_failed, non-retryable) and the session can never recover — the latch only clears when the old history comes back, which it won't. The verbatim-pause now applies only while the payload still fits the window; over-window payloads fold as usual. #6069
#6803 Maziar123 omp native (0.1.214): a degraded direct-send retry keeps the plugin-stamped prompt_cache_key on a payload whose fingerprint no longer matches it → strict upstreams 400. The direct-send guard now drops only a key this process stamped (LRU set, cap 1024); host-set keys and the normal incremental path are untouched. #6802
#6801 Maziar123 Fold-reconcile residual work retokenized every (anchor, candidate) pair → resumed sessions with heavy stored history blocked the event loop >300s. Digest/sketch computation is now memoized per candidate and pairing scans only the matching bucket (order-preserving, first-match semantics unchanged). #6800
#6775 mariopasquali Restart loses two persisted stats fields (same family as #2129/#5914): the k̂ calibration model provenance (calibratedEstimateModel, calibrationRing.model) — so a k̂ learned on one model leaks onto another after restart — and the emergency-shrink marker (emergencyShrinkAt/Estimate). Both restored on load with type validation. —
#6772 OMSociety #3962's dsh compaction-coverage readout had no reachable exit (its profile-bundle note never fires on desktop/mobile hosts). A shared compactionCoverageText() formatter now emits the readout once per process on the first routed request. #6770
#6766 sx4im P1-class: the watermark fastpath re-parsed the wire text on a HIT and discarded the output budget that restoreOutputBudget had restored into the parsed object — the restored max_tokens never reached upstream while the log said success. The plumbing moved to src/server/budget.ts (side-request re-exports it) and the fastpath now carries the restored field. A/B golden e2e pins it. #6752

Review notes

  • [PR] fix(storage): keep CCR originals safe during bulk migration (#5599) #6773 (excluded, independent deep review): coherent and well-tested, no data-loss path, but two jsonWriter=off regressions: converge-to-disk can permanently wedge a session under the multi-relay topology, and startup pre-warm silently skips sessions with originals. Full findings in the PR comment; waiting on author/owner decision.

Verification

Closes #6816
Closes #6069
Closes #6802
Closes #6800
Closes #6770
Closes #6752

sx4im and others added 13 commits November 14, 2026 17:15
)

On a watermark fastpath hit, prepareOpenai re-parses scalar fields from
the raw wire text, silently discarding the max_tokens restore that
restoreOutputBudget had just applied upstream of it. The log still
reports the restore while the forwarded body carries the client's
shrunk value (max_tokens: 1).

Snapshot the restored budget from the pre-rebind body and write it
back after the rebind. The output-budget field plumbing moves from
side-request.ts to budget.ts so the prepare modules can use it without
reaching into the side-lane engine (#1440 P2 cut 2); side-request.ts
re-exports it, so existing importers are untouched.

Regression test: hermetic A/B golden over two proxies (rollout off vs
on) asserting the starved turn forwards the restored ceiling on both
paths, with a residency check proving the fastpath actually hit.
…#6752)

The #1440 P2 cut 2 move (output-budget plumbing side-request.ts -> budget.ts)
trips the lane-guard (#2170): side-request.ts changed without a side-lane suite
update. Pin the re-export identity so the re-export and the definition can never
silently diverge.
The shape pass and the residual census in planReconciliation compare
every unclaimed anchor with every candidate. For each pair they ran
canonicalTextOf, wordBagDigest and simhashHex on the candidate text
again. These values depend only on the candidate text, so the work was
anchors x candidates full-text tokenizations.

A resumed session where every fold is inactive hits this hard: about
9.6k stored ids against 1.3k inbound messages kept the event loop busy
for more than 300 s, so the watchdog killed the proxy. The next request
for the same session started the loop again.

Now each candidate's canonical text, word-bag digest and simhash are
computed once. Candidates are grouped by word-bag digest (and, for user
text, by unwrapped identity), keeping their original order. Each anchor
only scans its own group, so the first match and the near-duplicate
answer are the same as before.

Measured on a copy of the affected store: the request that never
finished now takes about 0.8 s in bili. The new test (3k anchors x 1.5k
candidates) runs in about 1 s; before the change it did not finish in
300 s.
While a proxy is claimed (BILLION_CONTEXT_PROXY), the omp plugin stamps
the session id as a top-level prompt_cache_key. The proxy strips it
before it forwards. But when one request falls back to a direct send
(the proxy is not ready, or its respawn failed), the claim is still
set, so the body was already stamped and nothing strips the field.
A strict upstream then rejects the request:

  400 prompt_cache_key: Extra inputs are not permitted

Record the values the plugin stamps, and give the omp lane its own
direct-send guard: it runs the shared incremental guard first, then
removes a top-level prompt_cache_key only when its value is one this
process stamped. A key the host set itself is left alone.
…is pending (#6069)

The #3641 epoch-confirm latch pauses preflight folding and passes the
request through verbatim. An over-window payload cannot pass through, so
the paused preflight reported 0 compressible ranges and every turn was
refused with preflight_compress_failed. The latch clears only after a
turn is forwarded, so the session could not recover.

Keep the pause while the verbatim payload fits the window, judged on the
same calibrated scale as the fold loop, and fold as usual once it does not.
…s gate)

The constant is referenced only inside src/plugin-install.ts; the
unused-exports CI lane (INTERNAL_ONLY) fails on it. Drop the leading
export — no behavior change.
#6816)

The routing PR writes providers.claude-agent.settings.baseURL on 2.x
hosts; this note documents the new behavior in en+zh: re-run to route
existing installs, port drift + the doctor warning, foreign baseURL
preservation, and the mcp__opencode__* tools residual.
@github-actions

github-actions Bot commented Nov 15, 2026 •

Copy link
Copy Markdown
Contributor

📦 Built Package Artifact

Branch: batch11-wave11 (75a4538)

Latest build — install this exact version

npm install -g billion-context@0.0.10835-pr.6827.7ed0148

This exact version is immutable and reproducible. The rolling pr tag tracks the newest test build across all PRs.

Recent builds (commit → version)

commit version build
75a4538 billion-context@0.0.10835-pr.6827.7ed0148 build
05d75f2 billion-context@0.0.10692-pr.6827.e3b0de0 build
ae09131 billion-context@0.0.10649-pr.6827.b21baf4 build
6d2e2a1 billion-context@0.0.10635-pr.6827.2db219b build
0a9866a billion-context@0.0.10506-pr.6827.719d7ff build
25a89b1 billion-context@0.0.10200-pr.6827.ce26d87 build
1f29f9e billion-context@0.0.10193-pr.6827.3c6b43f build

Download artifact

  1. Download the artifact from the Actions run
  2. Extract the tarball and install:
tar xzf billion-context-pr6827.tgz
npm install -g package

Install instructions are refreshed on each push.

@ranxianglei ranxianglei changed the title wave-11 batch: 9 community fixes (opencode provider, incremental-ts journal, preflight epoch, output budget, +5) [PR] wave-11 batch: 9 community fixes (opencode provider, incremental-ts journal, preflight epoch, output budget, +5) Nov 15, 2026
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

Review of #6827 (wave-11 batch: 9 community fixes)

Verdict: mergeable — one CI blocker found and fixed (pushed), pending CI green.

I reviewed all 9 sub-PRs in detail (per-merge diffs against 2e3cfc97, the current master tip — the batch is up to date, no rebase drift). Each is a real 2-parent merge of a distinct, well-scoped community fix. The merge-base audit claim (all lines verbatim at tip) is consistent with the per-merge diffs I read.

CI blocker — found and fixed. The architecture-docs lane was red: FAIL: reference/architecture.md missing: src/incremental-first-seen.ts. PR #6818 added the new module but only updated the pi.ts row. Per the Documentation Layout rule, a new src/ module must be listed in the same PR. I verified this was the only missing entry (exactly one new src/ file, exactly one missing row). I pushed a one-line fix as an additive commit (25a89b1e) to batch11-wave11; the lane passes locally (architecture-docs guard: all green). Live CI on the new head: 0 failed so far, 10 pending.

Per-PR findings (code review, ③):

Regression review (②) — behavior changes, all intentional and disclosed:

Test effectiveness (④): each sub-PR ships a regression test — 9 new files (incremental-first-seen, issue6069-epoch-pause-over-window, issue6752-watermark-output-budget-restore, omp-direct-send-stamped-pck, fold-reconcile-residual-scale, pi-message-threads, persist-calibration-model, persist-emergency-marker, issue6770-compaction-coverage-text) plus doctor/plugin-install-opencode/opencode-rolling-cache-marks/side-lane-invariants/storage-batched-verification updates and the 4 golden files. The issue-numbered tests pin the original repro. Real assertions, not tautologies.

Diff cleanliness (⑤/⑤b): clean — every line maps to one of the 9 fixes; no unrelated churn, no mass whitespace. Code quality is high: fail-closed validation, merge-preserving writes, advisory-only doctor, no dead code, named constants. The #6819 unused export OPENCODE_CLAUDE_AGENT_PROVIDER was already caught and dropped by the unused-exports gate (9dc914cc).


8问(结构化审核结论)

  1. 是什么问题: 一个批量 PR,合并 9 个独立的社区修复——增量时间戳恢复日志、输出预算跨水位恢复、preflight 纪元暂停、omp prompt_cache_key、fold-reconcile 性能、pi anthropic 线程续接、persist 统计恢复、dsh 压缩覆盖率读数、opencode 2.x claude-agent 路由。
  2. 什么原因导致的: 9 个各自独立的根因——[PR] fix(incremental): keep message arrival timestamps across a crash before the SQLite save #6818 增量车道在规范行落库前时间戳丢失;[PR] fix: carry the restored output budget across the watermark rebind (#6752) #6766 水位快路径重解析时丢弃已恢复的 max_tokens;[PR] fix(preflight): fold an over-window payload while epoch confirmation is pending (#6069) #6807 纪元暂停闩锁在超窗会话上永久卡死;[PR] fix(omp): drop the stamped prompt_cache_key on a direct send (#6802) #6803 退化的 direct-send 保留插件盖戳的 prompt_cache_key 导致严格上游 400;[PR] fix(reconcile): compute candidate word evidence once per pass (#6800) #6801 恢复会话时 O(anchors×candidates) 全量重分词阻塞事件循环 >300s;[PR] fix(pi): negotiate full Anthropic requests for bili #6797 bili 传输改写历史后 pi 原生线程续接重新指纹化导致 429;[PR] fix(persist): keep k̂ model provenance and the emergency-shrink marker across restarts #6775 persist 恢复时丢弃校准/紧急收缩字段;[PR] feat(dsh): emit compression coverage readout on the first routing request (#6770) | feat(dsh): 在首个路由请求上输出压缩覆盖读数 (#6770) #6772 压缩覆盖率读数无可达出口;[PR] fix: route OpenCode 2.x built-in Claude Agent provider through the proxy (#6816) #6819 opencode 2.x 内置 claude-agent provider 不发 HTTP 请求导致静默不压缩。
  3. 影响范围与严重程度: 9 个修复影响面互不重叠。最严重:[PR] fix(preflight): fold an over-window payload while epoch confirmation is pending (#6069) #6807(P1,超窗会话每轮被拒,完全不可用)、[PR] fix: route OpenCode 2.x built-in Claude Agent provider through the proxy (#6816) #6819(内置 provider 静默不压缩,用户无感知地失去压缩)。[PR] fix(incremental): keep message arrival timestamps across a crash before the SQLite save #6818 是基础设施级(增量车道时间戳恢复,影响所有增量会话)。其余为性能/正确性/体验修复。
  4. 怎么修复的: 各修复都在根因层动手——[PR] fix(incremental): keep message arrival timestamps across a crash before the SQLite save #6818 内容寻址恢复日志 + 失败关闭校验;[PR] fix: carry the restored output budget across the watermark rebind (#6752) #6766 预算跨重绑定携带;[PR] fix(preflight): fold an over-window payload while epoch confirmation is pending (#6069) #6807 把纪元暂停限定在"载荷放得下"时;[PR] fix(omp): drop the stamped prompt_cache_key on a direct send (#6802) #6803 只丢弃本进程盖戳的 key;[PR] fix(reconcile): compute candidate word evidence once per pass (#6800) #6801 按候选记忆化 + 桶索引;[PR] fix(pi): negotiate full Anthropic requests for bili #6797 禁用 pi 原生续接并在 finally 恢复 thinking 级别;[PR] fix(persist): keep k̂ model provenance and the emergency-shrink marker across restarts #6775 严格类型校验后恢复字段;[PR] feat(dsh): emit compression coverage readout on the first routing request (#6770) | feat(dsh): 在首个路由请求上输出压缩覆盖读数 (#6770) #6772 共享格式化器 + 每进程一次出口;[PR] fix: route OpenCode 2.x built-in Claude Agent provider through the proxy (#6816) #6819 把内置 provider 指向实时 lane 端口且保留外部 baseURL。
  5. 是否有回归问题: 审了 9 个修复触碰的既有行为。已披露的有意变更:[PR] fix(persist): keep k̂ model provenance and the emergency-shrink marker across restarts #6775 的 stats 恢复改变 state digest(仅 stats 段,消息内容不变,golden 已更新);[PR] fix(preflight): fold an over-window payload while epoch confirmation is pending (#6069) #6807 超窗会话从"每轮被拒"改为"警告 + 折叠";[PR] fix(omp): drop the stamped prompt_cache_key on a direct send (#6802) #6803 direct-send 现在丢弃本进程盖戳的 key;[PR] fix(pi): negotiate full Anthropic requests for bili #6797 禁用 pi 原生 anthropic 线程续接;[PR] fix(incremental): keep message arrival timestamps across a crash before the SQLite save #6818 flushSync 现在排空等待者(修复潜在挂起)。未发现 fail-fast→吞错、阈值/时序漂移、输出格式意外变化。另发现并修复 1 处 CI 阻塞:architecture-docs 缺 src/incremental-first-seen.ts 条目(已推一行修复,本地 lane 已绿)。
  6. 是否破坏缓存: 不破坏。[PR] fix: carry the restored output budget across the watermark rebind (#6752) #6766 修复的是预算被丢弃的 bug(恢复缓存预算),[PR] fix(omp): drop the stamped prompt_cache_key on a direct send (#6802) #6803 修复的是严格上游 400(正确丢弃盖戳 key)。其余不涉及缓存失效策略。
  7. 是否改了设计方向: 不改。9 个修复都在既有架构内收敛,无新边界/数据流转向。[PR] fix(incremental): keep message arrival timestamps across a crash before the SQLite save #6818 的恢复日志是增量车道既有恢复机制的补全,不是新方向。
  8. 是否引入了新的配置: 不引入新的 bili 用户配置。[PR] fix: route OpenCode 2.x built-in Claude Agent provider through the proxy (#6816) #6819 的 claude-agent baseURL 是安装时写入宿主配置的(merge-preserving,不覆盖外部值),doctor 警告是 advisory-only。无新 env/config/默认值。

结论:可合。 9 个修复方向正确、代码质量高、各自有回归测试、无无意回归。唯一的 CI 阻塞(architecture-docs 缺条目)我已推一行修复,本地 lane 已绿;CI 在新 head 25a89b1e 上重跑中(0 失败,10 待跑)。等 CI 全绿即可合。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

Conflict resolved — final head 0a9866ab93 pushed to batch11-wave11 (fast-forward, no force-push).

What I did this run:

  1. Wake state: PR branch batch11-wave11 was 68 commits behind master (merge base 2e3cfc97), 30 ahead, merge state dirty. The two earlier CI-engineer sessions (11-16) died mid-run with no progress (badge reset 11-16 19:49Z); nothing had been pushed.
  2. Merged origin/master (301e1a5459) into the branch. Exactly ONE conflict: src/agent/pi.ts — both sides added independent module-level code at the same spot (between createPiDirectSendGuard and createBiliPlugin): the PR side's createOmpDirectSendGuard() ([PR] fix(omp): drop the stamped prompt_cache_key on a direct send (#6802) #6803, drops this-process-stamped prompt_cache_key) and master's nudge-debug module state (nudgeDebugWired / nudgeNotifiedTurns / nudgeFullPending, [PR] feat: pi lane — show the injected nudge text verbatim in the terminal when debug is on #6151). Neither references the other → resolved by keeping both (PR guard first, adjacent to its sibling, then master's state). All import hunks auto-merged; I verified the union of both sides' import lines survived (nudge-debug.js, debugEnabled, createPrepareDirectSend).
  3. Found master itself red on the architecture-docs lane and pre-empted it: src/responses-system-sections.ts (added by [PR] fix(responses): share developer segment concatenation between compressed re-request and steady state, fix prefix re-bill | fix(responses): 压缩重请求与稳态共用 developer 段拼接,修复每次折叠后的前缀重计费 (#6922) #6944) has no row in reference/architecture.md — reproduced on a clean master worktree (architecture-docs guard: FAILED). Inherited failure, not caused by this merge. Per the CI-coordination policy (2026-10-24) I merged origin/ci-fixes, which carries the fix (8ed56be997 docs(architecture): add src/responses-system-sections.ts to the module map (#6922)).
  4. Local verification on the final head: npm run build ✓ · npm run typecheck ✓ · tools/check-architecture-docs.mjs all green ✓ · targeted tests 59/59 (pi.ts merge area: nudge-debug, omp-direct-send-stamped-pck, pi-message-threads, issue3304-incremental-direct-send, issue3433-pi-incremental-guard, issue4516-pi-direct-fallback, issue5025-pi-direct-window, issue4894-direct-window-guard) + 36/36 re-run after the ci-fixes merge (incl. issue2929-codex-fold-double-miss, the only test file ci-fixes touched) ✓.
  5. Sanity: diff of the new head vs the old PR head contains only master/ci-fixes-side additions — no PR-side content lost; kernel (0.0.198) and pi-subagents (0.2.4) versions identical across base/master/branch, so no version-discipline exposure.
  6. Live status (verified this run): GitHub API single read → head 0a9866ab93, mergeable: true (conflict cleared), mergeable_state: blocked = pending CI/reviews on the fresh head, normal right after push. The mirror check cache was stale (last fetched 2026-11-15), so I could not read the new run's checks from it — CI verdict will come from the fresh run.

八问(结构化结论)

  1. 是什么问题:PR 分支落后 master 68 个提交且处于合并冲突态(dirty),无法合并。
  2. 什么原因导致的:批量 PR 自分叉点 2e3cfc97 后 master 继续合入 68 个提交;src/agent/pi.ts 同一位置被两侧各自独立新增代码(PR 侧 [PR] fix(omp): drop the stamped prompt_cache_key on a direct send (#6802) #6803 的 createOmpDirectSendGuard,master 侧 [PR] feat: pi lane — show the injected nudge text verbatim in the terminal when debug is on #6151 的 nudge-debug 模块状态),形成纯文本冲突。
  3. 影响范围与严重程度:仅影响本 PR 的可合并性,无用户侧行为影响;冲突为两侧纯新增、无语义竞争。
  4. 怎么修复的:merge origin/master 进 PR 分支(保留 9 个子 PR 的原始 merge 历史,不 rebase 改写);唯一冲突按「两侧都保留」解决;同时发现 master 自身 architecture-docs 红([PR] fix(responses): share developer segment concatenation between compressed re-request and steady state, fix prefix re-bill | fix(responses): 压缩重请求与稳态共用 developer 段拼接,修复每次折叠后的前缀重计费 (#6922) #6944 新增文件缺模块图条目,干净 master worktree 复现),按 CI 协调政策 merge origin/ci-fixes(含该条修复)。
  5. 是否有回归问题:无。本次只带入 master 既有变更(已合入已审)+ ci-fixes(哨兵收敛分支);PR 侧唯一改动是 pi.ts 冲突解法,已核对新 head 相对旧 PR head 的 diff 只含 master 侧新增、PR 内容零丢失;build/typecheck/定向测试(59+36)全绿。
  6. 是否破坏缓存:不涉及。
  7. 是否改了设计方向:不涉及。
  8. 是否引入了新的配置:不涉及。

判定:冲突已解,等新 head 0a9866ab93 的 CI 全绿即可合。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

No action needed this run — the conflict was already resolved in the previous run; verified live just now.

Live state (checked this run, not from memory):

The full CI pass is still running on 0a9866ab93. Nothing to fix or re-resolve; if any lane goes red, this lane will re-trigger and I'll fix it.

八问(结构化结论)

  1. 是什么问题: 系统转发报告 PR 与 base 分支冲突(dirty)——经实时核验,该状态已过期,冲突在前一轮已解决。
  2. 什么原因导致的: 转发基于镜像缓存的旧状态(缓存自 2026-11-15 起未刷新);实际分支已合入 master + ci-fixes,冲突已解。
  3. 影响范围与严重程度: 无实际影响——PR 当前 mergeable: true,仅等新 head CI 跑完。
  4. 怎么修复的: 本轮无需修复;核验了 head、merge 状态、check runs 三处实时数据。
  5. 是否有回归问题: 不涉及(本轮零改动)。
  6. 是否破坏缓存: 不涉及。
  7. 是否改了设计方向: 不涉及。
  8. 是否引入了新的配置: 不涉及。

判定: 无新问题,CI 进行中(0 失败),若变红本车道会自动再触发处理。

@ranxianglei ranxianglei changed the title [PR] wave-11 batch: 9 community fixes (opencode provider, incremental-ts journal, preflight epoch, output budget, +5) wave-11 batch: 8 community fixes (opencode provider, incremental-ts journal, preflight epoch, output budget, +4) — #6797 excluded Nov 17, 2026
@ranxianglei ranxianglei changed the title wave-11 batch: 8 community fixes (opencode provider, incremental-ts journal, preflight epoch, output budget, +4) — #6797 excluded [PR] wave-11 batch: 8 community fixes (opencode provider, incremental-ts journal, preflight epoch, output budget, +4) — #6797 excluded Nov 17, 2026
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

(ai: vllm-qwen/qwen3.8-27b)

Red architecture-docs lane fixed — new head ae0913189e pushed (fast-forward, no force-push).

What I did this run:

  1. Wake state: the trigger reported "merge state: dirty", but live verification showed that was stale — the branch had moved. Another session had pushed a newer merge 6d2e2a13e1 (upstream master post-[PR] docs(clients): dsh-go host | docs(clients): dsh-go 宿主 #6592/[Design][dsh] Stop guessing dsh's auxiliary requests — the in-process plugin can declare each request's purpose explicitly #6880 + storage-render golden regen) that diverged from my previous head 0a9866ab93. The real current blocker was a red architecture-docs CI lane on 6d2e2a13e1 (reproduced locally: FAIL: reference/architecture.md missing: src/incremental-first-seen.ts).
  2. Root cause: the re-merge 2d53c13dde was based on e45ad01695, which predates the 25a89b1e fix that originally listed src/incremental-first-seen.ts in reference/architecture.md. The re-merge dropped that entry, so the lane (which requires every src/ module to be listed) failed.
  3. Fix: re-added the one-line entry for src/incremental-first-seen.ts (alphabetical, before incremental-lane.ts), byte-identical to the original 25a89b1e fix.
  4. Verification: tools/check-architecture-docs.mjs → all green locally; diff is exactly one line, no other churn.
  5. Push discipline: re-fetched before push — remote head still 6d2e2a13e1 (unchanged), clean fast-forward 6d2e2a13e1..ae0913189e.
  6. Live CI on the new head ae0913189e (single read): architecture-docs → success ✓, 0 failed (27 success / 15 skipped / 13 in progress).

八问(结构化结论)

  1. 是什么问题:PR 分支的 architecture-docs CI 车道红——reference/architecture.md 缺 src/incremental-first-seen.ts 的条目。
  2. 什么原因导致的:另一会话做的 post-[PR] docs(clients): dsh-go host | docs(clients): dsh-go 宿主 #6592/[Design][dsh] Stop guessing dsh's auxiliary requests — the in-process plugin can declare each request's purpose explicitly #6880 重合并(2d53c13dde)基于 e45ad01695,该提交早于列出该文件的修复 25a89b1e,重合并把那条目弄丢了,guard 因此失败。
  3. 影响范围与严重程度:仅影响本 PR 的 CI 可合并性(一条车道红),无用户侧行为影响。
  4. 怎么修复的:重新加回那一行条目(字母序、在 incremental-lane.ts 之前),与原始 25a89b1e 修复逐字一致。
  5. 是否有回归问题:无。纯文档一行,guard 本地全绿,diff 恰好一行,无其他改动。
  6. 是否破坏缓存:不涉及。
  7. 是否改了设计方向:不涉及。
  8. 是否引入了新的配置:不涉及。

判定:architecture-docs 红已修,新 head ae0913189e 上该车道已绿、0 失败;其余 CI 跑完全绿即可合。

# Conflicts:
#	tests/golden/storage-render/anthropic.json
#	tests/golden/storage-render/chat.json
#	tests/golden/storage-render/google.json
#	tests/golden/storage-render/responses.json
# Conflicts:
#	src/server/budget.ts
…6775 persisted fields) — wire shas unchanged, persisted digests only
@ranxianglei ranxianglei changed the title [PR] wave-11 batch: 8 community fixes (opencode provider, incremental-ts journal, preflight epoch, output budget, +4) — #6797 excluded wave-11 batch: 7 community fixes (opencode provider, preflight epoch, output budget, +4) — #6797/#6818 excluded Nov 18, 2026
@ranxianglei ranxianglei changed the title wave-11 batch: 7 community fixes (opencode provider, preflight epoch, output budget, +4) — #6797/#6818 excluded [PR] wave-11 batch: 7 community fixes (opencode provider, preflight epoch, output budget, +4) — #6797/#6818 excluded Nov 18, 2026
@ranxianglei
ranxianglei merged commit a446a3c into master Nov 18, 2026
54 checks passed
@github-actions

Copy link
Copy Markdown
Contributor

📦 Merged Build Published

This merge to master was published as 0.0.1233-master.a446a3c (rolling master tag).

npm install -g billion-context@0.0.1233-master.a446a3c  # this exact build
npm install -g billion-context@master  # always the newest merged build

npm view billion-context@master prints the current version at any time.

Louie0806 added a commit to Louie0806/billion-context that referenced this pull request Nov 18, 2026
Re-sync the drifted-name / draft-envelope lane onto upstream master
a446a3c (PR ranxianglei#6827, 24 commits / 33
files). No lane file is touched by this delta (overlap 0), so the merge
is a pure history advance: no conflict resolution was needed.

Verification on the merged tree:
- kernel build rc=0
- tsc --noEmit: only the 5 pre-existing @deepseek-ai/cordis host-only
  errors in tests/dsh-tool-registration-lifetime.test.ts (also present
  on a pristine master worktree)
- gates: check-unused-exports, config-surface-guard,
  lifecycle-ledger-guard, check-test-import-graph,
  check-architecture-docs all rc=0
- gen-config-docs generate: working tree clean (docs fixed point holds)
- suites: issue5175 66/0, tag-echo 79/0, issue2190-degen-close 32/0,
  issue3905-announcement-tail 19/0, issue2929-codex-fold-double-miss 1/0,
  issue2303-draft-tail 14/0, native-compact-terminal 1/0,
  notice-registry 5/0, side-lane-invariants 18/0,
  issue5403-watermark-fastpath 1/0,
  issue6752-watermark-output-budget-restore 1/0
ranxianglei pushed a commit that referenced this pull request Nov 18, 2026
…nder golden conflict

Master's batch11-wave11 (#6827) regenerated the storage-render goldens for new persisted fields; the PR's lastFoldDrift field also touches them. Regenerated from the merged code so the snapshots reflect both; golden suite 5/5, PR test 3/3, typecheck clean.
ranxianglei pushed a commit that referenced this pull request Nov 18, 2026
…lict at 0.0.200

Master advanced past the PR base again (batch11 wave11 #6827, lean-compress
receipt #7105, wire-client-sidecars #7083, p1 batch b2/b3, #7087, #7101 ->
kernel 0.0.199, storage-render goldens regenerated for #6775 persisted
fields, 400 files). Content hunks auto-merged (kernel/src/compress.ts,
kernel/src/types.ts, all four storage-render goldens); the only conflicts
were the three kernel version lines, resolved at 0.0.200 = master's 0.0.199
+ this PR's kernel/src change, per the CI-enforced kernel version discipline.

Targeted local verification: tests/golden-storage-render.test.ts 5/5,
tests/storage-state-fields.test.ts 2/2, kernel issue6116/issue3769/
hide-consumed suites 27/27.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment