Skip to content

[PR] fix: strip ACP panels and status markers on the google lane before projection (#2798) - #2805

Closed
ranxianglei wants to merge 7 commits into
masterfrom
2026-10-11_google-acp-strip
Closed

ranxianglei wants to merge 7 commits into
masterfrom
2026-10-11_google-acp-strip

Conversation

@ranxianglei

@ranxianglei ranxianglei commented Oct 11, 2026 •

Copy link
Copy Markdown
Owner

Closes #2798

What this fixes

The Google (Gemini native) lane never ran the incoming ACP strip passes: src/server/prepare-google.ts:23 imported stripAcpPanelMessages + stripAcpStatusMarkers but never called either — not in prepareGoogle, not in prepareGoogleCountTokens. All three sibling lanes strip before projection (prepare-openai.ts:82,86, prepare-anthropic.ts:124,128, prepare-responses.ts:114). Consequence: ACP panel user messages (#359) and ephemeral [ACP] status marker lines (#717/#1029, incl. the multi-line acp_status reports fixed in #2789/#2797) rode unstripped into the compression state on every /v1beta turn — billed and replayed forever.

Not a regression from the #1440 extraction (per the issue's provenance): pre-extraction src/server.ts had exactly the same three call sites; the google branch never called them. The extraction carried over the shared import list, leaving the vestigial import.

Why it is not a one-line fix

stripAcpPanelMessages/stripAcpStatusMarkers operate on {role, content} arrays; google contents are {role, parts} — passing parsed.contents would be a silent no-op. This PR adds parts-aware variants instead.

Changes

Commit 1 — main fix (#2798)

  • src/acp-panel.ts: two new exported strippers mirroring sibling semantics:
    • stripAcpPanelGoogleContents(contents) — removes whole user-role contents whose plain-text join is an ACP panel (box / fallback / cache-report / rule-report signatures, same as the messages variant). Any non-text or signatured part makes a content ineligible → preserved. Never deletes non-panel contents.
    • stripAcpStatusMarkersGoogleContents(contents) — strips MARKER_LINE lines from text parts of user/model/role-less contents. Skips thought: true parts and parts carrying thoughtSignature (KDD release v0.1.3 #10: thinking-family blocks are signature-verified and must stay byte-exact — unlike anthropic, where the signature sits on a separate field name, so google needs an explicit guard). Marker-only parts degrade to " "; never deleted.
  • src/server/prepare-google.ts: both passes added to prepareGoogle (before the chain-carrier strip — order panels → markers → carriers matches the openai lane) and to prepareGoogleCountTokens, with the same log lines as the siblings so the lane is observable identically.
  • src/fork-adoption.ts: the same two passes mirrored in incomingCoreMessages's google branch. Required because maybeAdoptForkBlocks runs before prepare* on the raw body (first request of a fresh session); without the mirror, id derivation would diverge for marker/panel-bearing messages after this fix and silently drop ref inheritance.
  • tests/issue2798-google-acp-strip.test.ts (new, 11 tests): stripper units (panel signatures, multi-part join, non-text/signatured part preservation, KDD release v0.1.3 #10 byte-exactness, multi-line acp_status report shape), fork-adoption google parity, prepareGoogleCountTokens read-only measurement, and a full-server proof that upstream /v1beta traffic arrives stripped while still answering.

Commit 2 — in-passing fix (same mechanism family, labeled)

  • src/server.ts prepareCountTokens (the anthropic /v1/messages/count_tokens twin): forwarded the raw body for measurement while prepareAnthropic strips panels/markers before projection — the preview measured a payload the proxy never sends. Now runs the same two passes at the top of the preview (still read-only, same log lines). OpenAI/Responses have no count_tokens endpoint (existing isCountTokensRequest tests pin that), so this twin completes the set.
  • Also drops the unused stripAcpPanelResponsesInput import from server.ts (vestigial; tsc does not flag it because noUnusedLocals is off).
  • Regression test added to tests/count-tokens.test.ts (natural home; keeps commit 2 self-contained).

Behavior changes (old → new)

Surface Old New Why
/v1beta …:generateContent / :streamGenerateContent ACP panel user messages + [ACP] marker lines forwarded verbatim into compression state every turn (billed/replayed forever) stripped before projection, matching the other three lanes closes the lane gap (#2798)
/v1beta …:countTokens measured payload included panels/markers measured payload now equals what the main lane forwards measurement fidelity
Fork/resume adoption, google wire id-pass mirrored only the chain-carrier strip mirrors panels + markers + carriers in order adopted refs stay in parity with post-fix projection
Anthropic /v1/messages/count_tokens (in-passing) measured the raw body (panels/markers present) measures the stripped body, matching what /v1/messages forwards same mechanism family

No new config/env/CLI surface. No kernel change (kernel-change: not applicable — host-side only). No persistence format change.

Regression review (existing behaviors audited)

  • Prefix affinity compares raw request bytes before prepare — unaffected by stripping.
  • Thinking-family bytes (KDD release v0.1.3 #10): the new google marker stripper explicitly skips thought:true and thoughtSignature parts; pinned by a byte-exactness test.
  • Image index: buildIncomingImageIndex runs pre-prepare on the raw body by design, but refreshIncomingImageIndex re-runs post-prepare within the same request — self-heals exactly like all four lanes today. No change.
  • outboundContextEstimates: estimate-only fallback projection; same trait across all lanes. No change.
  • Agent-side fork-adopt: chatBodyToCore returns null for google bodies by design ([Bug][pi] fork adoption masks the root cause of snapshot 409; lineage-only subtask falsely triggered and Responses clien | [Bug][pi] fork adoption 隐藏 snapshot 409 原因,lineage-only 子任务误触发且 Responses 客户端未支持 #2469) — nothing to mirror client-side.
  • Compression modes: the strip applies to incoming history uniformly, at the same position as the sibling lanes, before any mode-specific handling — reasoned in both plugin mode and proxy mode (§6).
  • Cache/retry/cursors: systemInstruction anchor untouched; the strips remove proxy-injected ephemeral artifacts from history, making the forwarded prefix more stable, never less. Session identity/persistence untouched.

Verification

Local targeted runs (full suite via CI): typecheck clean; tests/issue2798-google-acp-strip.test.ts 11/11, tests/acp-panel.test.ts, tests/count-tokens.test.ts 9/9 (incl. new), tests/issue1029-acp-marker-stale-refs.test.ts, tests/fork-adoption-wires.test.ts, tests/pi-omp-fork-adopt.test.ts, tests/dsh-fork-adopt.test.ts, tests/issue1204-rules-google-plugin.test.ts — 99/99 pass.

Golden regeneration (commit 3) — intentional drift, justified

First CI run was red on exactly two tests, both asserting against the shared golden-google pin set (tests/golden/storage-render/google.json):

  • storage-render golden (google): wire bodies + persisted state match committed snapshot
  • storage engine: full-mode restart is invisible on the google wire (== legacy restart, wire == committed goldens)

Why the pins moved: the golden drive runs with compress: { injectTool: true, injectNudge: true }, so once the fold fires at body #9 the proxy injects its ephemeral [ACP] status-marker line (#717/#1029) into the next request's incoming history. Pre-fix that line rode unstripped into the google lane's projection on every subsequent turn — i.e. the pin set had codified the exact defect #2798 removes. Post-fix prepareGoogle strips it before projection, so bodies #10+ and the final persisted state legitimately differ. This is the fix working as designed, made visible by the pin set; per the gate's own instruction the goldens were regenerated with node --import tsx scripts/update-storage-render-goldens.ts.

The delta is exactly the strip effect and nothing else (reviewed field-by-field):

  • outbound bodies release v0.1.3 #10–release v0.1.4 #14 each shrink by a constant 268 bytes (one fixed marker content removed from contents; constant across all five bodies rules out nondeterministic drift);
  • moved digests: stateDigest, sectionDigests.{state,stats,metadata}, stateKeyDigests.{messageRefs,tokenSnapshot,lastPassIds};
  • unchanged: blockContents digest, blocks/nudge/nextBlockId/nextRunId/deadRefs/hiddenOrphanRefs/terminalStreak digests, foldAt=9/editedAt=12/summaryFrom=9, and stateShape incl. the refs count (the marker rides inside an existing message as a line, so no ref is minted/freed);
  • chat/anthropic/responses goldens regenerate byte-identical (no diff) — confirming zero cross-lane effect;
  • local re-run of both previously-failing test files: 15/15 pass.

Rebase onto current master + #2797 block-semantics adaptation (commits 1cf37f907 → a48d5e099)

Conflict wake-up arrived while master had merged #2797 (3185aee4a, acp_status marker = BLOCK not line): the platform sync had merged a stale master missing it, so the google stripper was adapted in the resolution — it now calls the shared stripAcpStatusBlocks verbatim, exactly like the three sibling lanes.

  • Behavior change (old→new): on /v1beta traffic the google lane previously would have stripped only the [ACP] head line of an acp_status report (pre-fix(#2789): strip the whole acp_status marker block from incoming history #2797 semantics); it now consumes the whole block — the sentinel-wrapped [acp-status]…[/acp-status] body exactly, or the legacy unwrapped report grammar up to the first prose line, or to end-of-message when cut mid-report — identical to openai/anthropic/responses post-fix(#2789): strip the whole acp_status marker block from incoming history #2797. Also: a marker at end of part leaves no trailing newline (shared split/join semantics).
  • Conflicts resolved: src/server.ts import line (kept — used by the in-passing count_tokens fix); tests/golden/storage-render/google.json (took master's, regenerated — same strip-only diff as before, stateDigest bafa686948387cd7…, chat/anthropic/responses byte-identical).
  • Test suite adapted (a48d5e099): mirrors tests/acp-status-marker-block.test.ts fixtures for the google parts shape — wrapped whole-block removal, legacy stop-at-prose, cut-mid-report consume-to-end, plus a signed-part-with-full-block byte-exact case (KDD release v0.1.3 #10).
  • Verification on the merged head: typecheck clean; targeted suites 123/123 (issue2798 15, acp-panel, acp-status-marker-block, count-tokens, golden-storage-render, storage-engine-proofs, fork-adoption-wires, pi-omp/dsh-fork-adopt).

ework-agent added 2 commits October 11, 2026 21:58
…ojection (#2798)

prepare-google.ts imported stripAcpPanelMessages/stripAcpStatusMarkers but never called them, so ACP panel user messages (#359) and ephemeral [ACP] status marker lines (#717/#1029, incl. multi-line acp_status reports #2789/#2797) rode unstripped into the compression state on every /v1beta turn and were billed/replayed forever while the openai/anthropic/responses lanes stripped them. The existing strippers only understand {role, content} arrays — google contents are {role, parts} — so passing parsed.contents would have been a silent no-op.

Add parts-aware variants to acp-panel.ts: stripAcpPanelGoogleContents (whole panel user contents removed; any non-text/signatured part makes a content ineligible) and stripAcpStatusMarkersGoogleContents (per-part marker-line stripping that skips thought/thoughtSignature parts per KDD #10). Call both in prepareGoogle and prepareGoogleCountTokens before the chain-carrier strip, with the same log lines as the siblings. Mirror the same two passes in fork-adoption's incomingCoreMessages google branch so adoption id derivation stays in parity with prepareGoogle (maybeAdoptForkBlocks runs pre-prepare on the raw body).

Behavior change (old->new): /v1beta generateContent/streamGenerateContent/countTokens previously forwarded and measured ACP panel messages + [ACP] marker lines verbatim; they are now stripped before projection/measurement, matching the other three lanes. No config surface change, no kernel change.
…s preview too (in-passing, #2798)

Found while working #2798: prepareCountTokens (/v1/messages/count_tokens) forwarded the raw body to upstream for measurement while prepareAnthropic strips panels (#359) and [ACP] marker lines (#717/#1029) before projection — so the preview measured a payload that is never what the proxy sends. Same mechanism family as the google gap, trivially fixable in passing: run the same two strip passes at the top of the preview, same log lines, still read-only. OpenAI/Responses have no count_tokens endpoint, so this twin completes the set.

Also drops the unused stripAcpPanelResponsesInput import from server.ts (vestigial; tsc does not flag it — noUnusedLocals is off).

Behavior change (old->new): /v1/messages/count_tokens previously measured panel messages + marker lines verbatim; they are now excluded from the measured payload, matching what /v1/messages forwards.
@github-actions

github-actions Bot commented Oct 11, 2026 •

Copy link
Copy Markdown
Contributor

📦 Built Package Artifact

Branch: 2026-10-11_google-acp-strip (a48d5e0)

Option A — Install this exact build (recommended)

npm install -g billion-context@0.0.3343-pr.2805.513d976

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

Option B — Download artifact

  1. Download the artifact from the Actions run
  2. Extract the tarball and install:
tar xzf billion-context-pr2805.tgz
npm install -g package

Install instructions are refreshed on each push.

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

[bot] 🏷 Reviewing #2805 now.

Pulled the branch (2026-10-11_google-acp-strip, 2 commits / 6 files) and read the full diff + tests. Code-level read is done and looks sound; final CI (the 4 full-test-matrix jobs on ubuntu/windows × node 22/24) is still running — I'll post the structured conclusion as soon as it's green.

Quick preview of what I verified so far:

  • Root cause is real & correctly located. prepare-google.ts imported both strippers but never called them; all three sibling lanes strip panels→markers→carriers before projection. The google lane now matches that exact order (panels → markers → carriers), confirmed against prepare-openai/anthropic/responses.
  • Not a one-line fix, correctly so. {role, parts} ≠ {role, content}, so the messages-array strippers would be silent no-ops on /v1beta. The new parts-aware variants are conservative: a content is panel-eligible only when every part is plain {text}; any image/functionCall/functionResponse/thought/thoughtSignature part makes it real content and it's preserved.
  • KDD release v0.1.3 #10 guard is present and load-bearing. Unlike anthropic (signature on a separate field), a Gemini thinking part carries its own .text, so stripAcpStatusMarkersGoogleContents explicitly skips thought:true and thoughtSignature parts — pinned by a byte-exactness test. This is the one place a naive port would brick signed-thinking sessions.
  • Fork-adoption mirror adds the same two passes to incomingCoreMessages's google branch in the same order, keeping id/ref-inheritance parity with post-fix projection.
  • In-passing twin (commit 2): anthropic /v1/messages/count_tokens preview measured the raw body while prepareAnthropic strips — now measures the stripped payload. Behavior change (old: count included panels/markers → new: count matches what's forwarded); disclosed in the PR table. OpenAI/Responses have no count_tokens endpoint, so the set is complete.
  • Tests have real teeth (whole-message discipline, split-panel join, non-text preservation, marker-only→space never-deleted, KDD#10, fork parity, countTokens read-only, plus a full-server proof that upstream /v1beta arrives stripped while still answering).

No new config/env/CLI surface, no kernel change, no persistence change. Awaiting the last 4 green checks before the formal verdict.

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

[bot] 🏷 PR review 结论 — 方向对、代码 sound,但 CI 红:拦下(待一次性 golden 重生成 + PR body 理由)

代码我逐段读过(diff 全量 + 相关既有行为),修复本身正确、定位准、测试有牙。唯一阻塞项不在代码逻辑,而在漏了一次 golden 快照重生成。下面按八问给结论。

1. 是什么问题:Google(Gemini 原生)lane 在投影前从没跑过 ACP panel / status-marker 剥离 —— prepare-google.ts 只 import 了 stripAcpPanelMessages/stripAcpStatusMarkers 却从不调用。后果:#359 的 panel 用户消息、#717/#1029(含 #2789/#2797 的多行 acp_status 报告)的 [ACP] marker 行,每一轮 /v1beta 都原样进压缩态、被计费并永久回放。属 lane 遗漏,不是 #1440 抽取引入的回归(抽取前同样缺这三处调用,只是把残留 import 带了进来)。

2. 什么原因导致:(a) 抽取时共享 import 列表被照搬,google 分支却没接调用;(b) 不是能一行修的原因 —— 那两个 stripper 吃 {role, content},google 是 {role, parts},直接传 parsed.contents 是静默 no-op,必须写 parts-aware 变体。

3. 影响范围与严重程度:所有走 /v1beta …:generateContent/:streamGenerateContent/:countTokens 的 Gemini 原生会话;每轮多计费 + 回放 panel/marker token,长会话累积放大。不是「完全不可用」,是特定条件下持续的成本/正确性瑕疵(计费虚高、压缩态被临时 artifact 污染)。中偏高。

4. 怎么修复的:新增两个 parts-aware 剥离器 stripAcpPanelGoogleContents / stripAcpStatusMarkersGoogleContents(语义对齐兄弟 lane);在 prepareGoogle 与 prepareGoogleCountTokens 里按 panels→markers→carriers 顺序、投影前执行(与 openai lane 同序)。fork-adoption.ts 的 google 分支同步镜像这两遍 —— 因为 maybeAdoptForkBlocks 在 prepare* 之前跑原始 body,不镜像会让带 marker/panel 的消息在本修复后 id 派生漂移、静默丢 ref 继承。anthropic /v1/messages/count_tokens 孪生(prepareCountTokens)补上同样两遍(in-passing,同一机制族)。KDD#10 关键点做对了:Gemini 的 thinking part 自带 .text(不像 anthropic 签名在独立字段),所以 marker 剥离器显式跳过 thought:true 与带 thoughtSignature 的 part,保证签名块字节不变 —— 这点若照抄 anthropic 版本会 brick signed-thinking 会话,PR 明确防住了。

5. 回归问题 — 这里有问题(CI 红):

  • 代码层面既有行为我逐项过过,均 OK:prefix affinity 用 prepare 之前的 raw bytes 比较,不受剥离影响;image index post-prepare 自愈;无新配置/持久化格式变更;两种压缩模式(plugin/proxy)都在同一位置统一作用于入站历史,推过。
  • 但全量 test matrix 挂了 3/4 job,最终 tally pass 6340 / fail 2,两个失败都是 google wire golden:
    • tests/golden-storage-render.test.ts → storage-render golden (google): wire bodies + persisted state match committed snapshot
    • tests/storage-engine-proofs.test.ts:161 → storage engine: full-mode restart is invisible on the google wire
  • 根因明确且确定(非 flaky):本 PR 改变了 google 出站 wire 内容(压缩 foldAt 之后回放的 marker/panel 现在被剥掉了),而提交里没重新生成对应的 golden。两个失败测试读的是同一个文件 tests/golden/storage-render/google.json;chat/anthropic/responses 三个 lane 的 golden 全绿(它们 wire 没动)。drift 指纹也吻合:blockContents 不变、只有 messageRefs/tokenSnapshot/lastPassIds 偏移 = 投影少了几条临时消息,压缩块本身完好。
  • 按 AGENTS.md §7.3:golden gate 触发就是 stop-the-line,不能放宽。

6. 是否破坏缓存:不破坏,反而让出站前缀更稳(去掉代理注入的临时 artifact)。session identity / persistence 不动。

7. 是否改设计方向:不改架构/边界/数据流,只是把 google lane 补齐到与其它三 lane 一致的既有机制(parts-aware 变体)。

8. 是否引入新配置:无新 env/config/CLI。


处置(合并前必办)

  1. node --import tsx scripts/update-storage-render-goldens.ts —— 它会重写四个 wire 的 golden,预期只有 tests/golden/storage-render/google.json 变化。
  2. git diff 该文件,人工确认:变化恰好是预期的 marker/panel 移除(post-fold 各轮的 outbound body sha / state digest),没有误删真实对话内容(Turn N: 语料本身不含 panel/marker 特征,理论上不该被动到)。
  3. 把这次 golden 重生成作为一个独立标注 commit 提上去,并在 PR body 写明理由(§7.3 要求):「google wire 现于投影前剥离 panel/marker,故 google render golden 相应更新;仅 google 变化,blockContents 不变」。
  4. 重新跑 CI 转绿。

修复质量我给好评:root cause 找得准、KDD#10 守卫到位、11 个新单测(含 KDD#10 字节精确性、fork parity、full-server proof)都有牙且在 CI 里全绿。只差这一次 golden 重生成 + 理由说明即可合。

判定:拦下(block) —— 待上述 golden 重生成合入并 CI 转绿后放行。

…tentional drift)

The golden-google drive runs compress.injectTool+injectNudge, so after the fold at body #9 the proxy injects its ephemeral [ACP] status-marker line into the next request's history (#717/#1029). Pre-fix that line rode into the google lane's projection on every subsequent turn; post-fix prepareGoogle strips it before projection — which is exactly what #2798 changes, now observable in the pin set.

Delta is exactly the strip effect and nothing else: outbound bodies #10-#14 each shrink by a constant 268 bytes (the fixed marker content removed from contents); stateDigest/sectionDigests(state,stats,metadata)/stateKeyDigests(messageRefs,tokenSnapshot,lastPassIds) move accordingly; blockContents digest, blocks/nudge/deadRefs/hiddenOrphanRefs/terminalStreak digests, foldAt/editedAt/summaryFrom and stateShape (incl. refs count) are unchanged. The other three wires' goldens regenerate byte-identical (no git diff for chat/anthropic/responses), confirming no cross-lane effect. Regenerated with node --import tsx scripts/update-storage-render-goldens.ts (#2798, Closes via PR #2805).
ework-agent added 3 commits October 13, 2026 08:51
#2798 conflict resolution: keep the count_tokens strip import in server.ts; adopt the #2797 block-marker semantics for the google stripper (stripAcpStatusBlocks); google storage-render golden regenerated post-merge (strip-caused drift only).
…k semantics

The merged master changed marker stripping from line-granular (stripMarkerLines) to block-granular (stripAcpStatusBlocks): an acp_status head now consumes its wrapped [acp-status]/[/acp-status] body or legacy report grammar, and end-of-part markers no longer leave a trailing newline. The google lane shares stripAcpStatusBlocks verbatim with the siblings (adapted in the merge commit), so this suite now pins the same shapes tests/acp-status-marker-block.test.ts does: wrapped whole-block removal, legacy grammar stop-at-prose, cut-mid-report consume-to-end, plus a signed-part-with-full-block byte-exact case (KDD #10).
@ranxianglei

Copy link
Copy Markdown
Owner Author

(ai: zhipuai-lb/glm-5.3)

陈旧票清理(第三批动作①):目标 issue #2798 已关。为避免重复劳动关闭本 PR;如判断有误请直接 reopen 并留言。

@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

结论:同意关闭,不 reopen。我回到当前 master 核对了代码 —— #2798 已被更新的统一机制修掉,本 PR 的方案被整体取代。

没有直接照单全收「目标 issue 已关」,而是查了当前 master:

1) #2798 的缺口确实已修(用的是另一套机制)
src/server/prepare-google.ts:112-115 有明确引用 #2798 的注释,并调用统一的入站剥离引擎:

// #2798: google previously only stripped chain carriers — panels, status
// marker lines and report bodies rode the rebuild unseen ...
const markerCounts = stripRawStage(parsed, "google");

该引擎来自更新的 PR #2815(marker-declarative)(src/marker/engine.ts),把 openai/anthropic/responses/google 四条 wire 的 panel/marker/report/nudge/carrier 剥离合并进一个声明式引擎。本 PR 想做的几件事它都覆盖了:

  • prepareGoogle 投影前剥离 → 引擎调用(prepare-google.ts:115);
  • fork-adoption 的 google 分支镜像 → fork-adoption.ts:124-125 同样走 stripRawStage(clone, "google");
  • 本 PR 特意加的 KDD#10 thought/thoughtSignature 字节精确守卫,引擎本来就有(engine.ts:141-143)。

2) 时间线对得上:本 PR 基于 2026-10-11 的旧 master(那时还没有这个引擎,所以走了「acp-panel.ts 加两个 parts-aware 变体」的老路子);issue #2798 于 2026-10-13 标 completed(对应 #2815 合入);master 此后大幅重构,本 PR 的 diff 现在也套不上去 —— 硬合会引入一条与引擎并行的重复/冲突剥离路径。

所以我上一轮「block、待 golden 重生成」是针对当时基线的评审,现在整个方案已被 #2815 取代,该结论作废。关闭是对的,我不 reopen、不动它。

顺带一个小的观察(不构成重开理由,单独留意即可):master 的 google countTokens 孪生(prepareGoogleCountTokens,prepare-google.ts:397)只做 fold-prune(processTurn + stripKernelSummaries),没有跑 stripRawStage,而主路径跑了。即 countTokens 测出来的量仍会把 panel/marker 计入,与主路径实际转发量不完全一致 —— 属测量保真度的小瑕疵。这是 #2815 之后的边角,若要处理应基于当前 master 用引擎补一处调用,而不是复活这个陈旧 PR;是否立项由你定。

判定:维持关闭,不 reopen。

@ranxianglei ranxianglei changed the title fix: strip ACP panels and status markers on the google lane before projection (#2798) [PR] fix: strip ACP panels and status markers on the google lane before projection (#2798) Oct 24, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Google (Gemini native) lane never strips ACP panels/status markers from incoming history — vestigial import in prepare-google

1 participant