Skip to content

[PR] feat: host message ids into the canonical store — reconcile becomes exact set algebra (#3464 R1) - #3473

Closed
ranxianglei wants to merge 3 commits into
masterfrom
2026-10-17_host-message-ids
Closed

ranxianglei wants to merge 3 commits into
masterfrom
2026-10-17_host-message-ids

Conversation

@ranxianglei

Copy link
Copy Markdown
Owner

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

Closes #3464

What this does (R1 of the #2960 identity pipeline)

Hosts that serialize stable per-message ids (dsh Message.id, opencode msg_…) now ride those ids through bili end-to-end. On stamped lanes, message identity stops being byte-hash guessing and becomes exact set algebra:

  • Transport: plugins passively read each wire message's own id off the outgoing body and stamp it in the reserved top-level field _bili_message_ids — only when the proxy advertises capabilities.hostMessageIds in its manifest. The proxy extracts + unconditionally strips the field at the single body-parse chokepoint (src/server/handle.ts), so the reserved namespace can never reach upstream.
  • Storage: schema v6 adds messages.host_id / host_slot plus partial unique index (session_id, host_id, host_slot) WHERE host_id IS NOT NULL. planIngest dedups stamped legs by host key: one row per host message across drifted resends (fresher bytes win, first-seen position kept). Store reuses the host-keyed seq across byte drift and refreshes the row in place.
  • Reconcile: new top-priority Pass -1 claims missing covered ids by exact host-id equality (role-gated, k-th-to-k-th within multi-core wire groups) ahead of every byte-based pass; anchor.h seeds and backfills across passes so a later unstamped request degrades gracefully to the heuristic ladder.
  • Fork: snapshot responses gain an additive hostIds array mapping each ordered message back to its bili ref (outside every hash by construction).

Lanes without host ids (pi/omp message model has no id field per the #2938 survey; claude/codex/kimi/zcode) keep today's byte-keyed behavior bit-for-bit — gradual adoption, not protocol-breaking. No kernel changes, no new config/env surface.

Behavior changes (old → new)

# Change
a Any JSON body carrying _bili_message_ids: old = unknown top-level field could reach upstream via re-serialization (accidental leak surface); new = extracted + stripped pre-forward on every path. All other bytes unchanged.
b Storage schema v5 → v6: additive (two nullable columns + partial unique index), probe-guarded ALTERs, older builds no-op at the version gate (rollback-safe both directions).
c Reconcile gains Pass -1 (byHost counter added to every result literal): active ONLY when the request carries valid host ids AND anchors recorded them; unstamped lanes produce bit-identical plans/counters (pinned by test).
d Stamped-lane rows may REFRESH bytes in place on host-key collision — a documented append-only exception, required because post-#2960⑤ the DB view must equal the in-memory tail bytes or sampled digest parity false-alarms. Unstamped rows are never refreshed.
e Stamped-lane seq assignment prefers host-key continuity: a drifted resend keeps its seq instead of minting a new one (no watermark churn).
f Plugins stamp only when the manifest advertises capabilities.hostMessageIds; old proxies never see the field, old plugins against new proxies simply don't stamp.
g Fork snapshot response gains an additive hostIds array (old clients ignore unknown fields; matching/hash logic untouched).

Scope interpretation note

Issue acceptance says "dsh(pi 亦然)", but pi's message model carries no per-message id (#2938 lane survey) — so the mechanism is implemented generic-per-lane with adoption following host serialization capability: dsh + opencode get stamps, pi/omp deliberately NOT wired (a reader that can never fire is dead code). The passive read lights up automatically for any future host (or host build) that starts serializing message ids into the anthropic body. Active injection (forcing ids into the body via waterfall/transform hooks) is a follow-up pending evidence of which hosts serialize them — the passive path makes that optional.

Verification

  • npm run typecheck clean; npm run build clean (zero kernel changes — no kernel-change marker needed).
  • New suite tests/issue3464-host-ids.test.ts (27 tests): channel extract/validate/strip matrix, slot semantics, schema v5→v6 migration + fresh DB, planIngest twin elimination / no-id control / slot disambiguation / fork ref mapping, store seq continuity + in-place refresh, Pass -1 claim/role-guard/k-th/unstamped-inert/anchor-persistence.
  • Targeted regression suites green locally: fold-reconcile (+positional/perf), issue2915, fold-drift, lineage-reconcile-boot, storage-unified, storage-ingest-watermark, storage-ext, storage-engine-bootstrap, persist-tail, multi-instance-persist, plugin-fork, identity-proof, plugin-protocol, plugin-agent, dsh-title-sidecar-handle, omp-native-identity, persist-stats-roundtrip, storage-save-path, storage-review-fixes, cache-friendly-plugin, proxy-persist.
  • Fixture updates forced by the result-shape change: byHost: 0 added to one deep-equal literal in tests/fold-reconcile.test.ts; user_version assertions 5→6 in three storage tests (the watermark test's synthesized DDL now strips the v6 additions too, so it exercises the real v4→v6 path).
  • Pre-existing local failures NOT caused by this PR (verified identical on pristine master 269accd via detached worktree): proxy-persist.test.ts scheduleSave/flushSync entry-count assertions (environmental temp-dir behavior); full-suite verdict = remote CI.

Deferred (tracked, not silent)

@github-actions

github-actions Bot commented Oct 17, 2026 •

Copy link
Copy Markdown
Contributor

📦 Built Package Artifact

Branch: 2026-10-17_host-message-ids (a94d371)

Option A — Install this exact build (recommended)

npm install -g billion-context@0.0.4116-pr.3473.e7a646a

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-pr3473.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] 🏷 (ai: vllm-qwen/qwen3.8-27b)

查重: 不重复。 全仓检索(open+closed)「host message id / _bili_message_ids / 正史收编」仅命中本 issue 与其母 issue #3464。相关项只有 PR #2981(closed unmerged,#2938 Tier-1 tool-call-id 通道半边)——已被本 PR 的 per-message 设计取代,master 上无同类在途实现,无并行工作冲突。

开始 review:已拉取分支 pr-3473(单 commit 2ff663d67,base 269accde8),正在做方向/回归/代码/测试/diff 五关审查 + CI 状态核对,结论按八问体例回帖。

ranxianglei pushed a commit that referenced this pull request Oct 17, 2026
Review fix for #3473 (separate from the feature commit): the interface is
only used as extractHostIds' return type inside this file, so the leading
export trips scripts/check-unused-exports.mjs (INTERNAL_ONLY). Same class as
master's 8e110a6. Drops the export; no behavior change.
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

Review 结论:暂不可合(拦下)。 方向正确、降级路径干净、无回归,但核心配对函数把「wire 对齐的 id 列表」和「展开后的 core 数组」当成同一个东西在按位置 zip——实跑证明:任何含多块消息(text+tool_use、text+tool_result,即任何带工具调用的回合)的会话里,整个 host-id 通道静默失效。验收标准 #1/#3 在真实流量下不成立。另有 1 处 CI 红已顺手修掉推上分支,1 处前提需要补端到端证据。

F1(拦下):coreHostRefs 的配对契约与生产数据形状不符

生产链路永远是这两个形状:

  • extractHostIds(handle.ts 解析收口点)/collectWireHostIds(插件侧)产出的列表按 wire message 对齐——长度校验就是对着 messages 数组做的,一条 wire 消息一个条目;
  • anthropicToCore(kernel/src/wire/anthropic.ts:93-207)对每条 wire 消息的每个 content block 产出一个 core,同一组 core 共享同一个 wireIndex。

而 coreHostRefs(src/host-id-channel.ts:131-150)要求 ids.length === msgs.length 且用 ids[i] 按位置取 id。只要存在一条多块消息,两个数组长度不等 → 返回全 null → 通道关闭。我在 PR head 上用真实代码路径实测:

3 条 wire 消息(user text / assistant text+tool_use / tool_result)+ stamp ["H1","H2","H3"]
→ extractHostIds OK → anthropicToCore 展开出 4 个 core
→ coreHostRefs 结果: [null,null,null,null]   ← 通道死

后果(全部下游都挂在同一个函数上):

测试把这个错误契约钉成了绿的:tests/issue3464-host-ids.test.ts 的 fixture 传的是按 core 对齐的列表(如 :300 两条 core 来自一条 wire 消息却给 ["HP","HP"];:186 三条 core 来自两条 wire 却给 ["H1","H1","H2"])——生产永远不会产出这种形状。27 个用例全过,但对真实流量是空转。同族问题还有一处:src/persist.ts buildSession 用 pluginSnapshotHostIds.length === pluginSnapshot.length(core 数)校验持久化配对,多块快照的正确 wire 对齐列表会在加载时被整段丢弃——而 handlePluginFork 自己又是按 maxWire 切片(wire 语义)的,两处互相矛盾。文档反而写对了:MESSAGE-IDENTITY.md 与各处注释都说「via their wireIndex sidecars」「k-th pairing within multi-core groups」——实现与自家文档相悖。

修法方向(给作者,几行级但牵动测试面):coreHostRefs 改为按 m.wireIndex 查表(h = wi>=0 && wi<ids.length ? ids[wi] : null),去掉严格等长守卫;buildSession 校验改成对照覆盖的 wire 跨度(复用 handlePluginFork 已有的 maxWire 推导);fixture 全部改回 wire 对齐形状;并补一个端到端用例:真实 anthropic body 走 extractHostIds → anthropicToCore → coreHostRefs(我的探针可直接当种子)。纯单块会话在新旧两种实现下结果一致,所以这个改法不破坏现有绿测。

F2(已直接修,commit 472fb02)

CI unused-exports job 红:HostIdExtract 导出但仅文件内自用 → 去掉 export(与 master 8e110a6 同类)。本地已复验 gate 通过 + typecheck 干净,CI 会在新 head 上重跑。

F3(需补证据):「宿主真的把 id 序列化进出站 body」是承重前提,仓内零证据

_bili_message_ids 在全仓只出现在 2 个源文件,没有任何测试覆盖「插件 stamp → 代理提取 → ingest」整链。dsh 的 MessageBase.id(以及 opencode 的 msg_… 是否真进 anthropic body 的每个 message 对象)若实际不存在,collectWireHostIds 返回 null,整个 R1 连单块会话都不会激活。被动读取的设计本身没问题(文档也写了「when the host serializer includes it」),但验收要求至少一条 lane 有实证:建议补一个 hermetic e2e(fake dsh host,body 消息带 id → 断言 [hostids] host message-id channel active 日志行或 DB 行 host_id 非空)。

F4(合并前注意)

分支落后 master 5 个 commit(fa41483),其中三个动过 src/agent/dsh-native.ts(#3451 web profile compaction、#3393 重注册自锁、8e110a696 unused-exports)。当前无文本冲突(mergeable:true),合并前请 rebase 到最新 master,我的修复 commit 骑在顶上不受影响。

本地定向验证(PR head 2ff663d)

新套件 27/27 过;触碰的既有套件(fold-reconcile/storage-ext/storage-ingest-watermark/storage-unified)99 过 1 skip 0 败;typecheck、build 干净。全量回归以远程 CI 为准(已在 472fb02 上重跑)。


八问

  1. 是什么问题:本 PR 是功能不是缺陷——[identity] R1: ingest host message ids into the canonical store — reconcile becomes exact set algebra instead of byte-hash guessing (#2960 L3) #3464/Identity pipeline: key sessions by (agent, conversation id) in the unified store — decouple fold ledgers from protocol/host origin (#2938 / #2847 / #2671 defect-1) #2960 的 R1:把宿主稳定消息 id(dsh MessageBase.id / opencode msg_…)经保留 body 字段 _bili_message_ids 送进正史库(schema v6),让 fold reconcile 从字节哈希猜测定格为精确集合代数(Pass -1)。review 的问题是:它是否兑现了声明的能力。
  2. 什么原因导致的(我找到的缺陷之根因):实现者混淆了两个数组类——wire 对齐的 stamp 列表(每 wire 消息一条)与展开后的 core 数组(每 content block 一条)——zip 写成了位置索引+严格等长守卫,而所有文档注释描述的都是 wireIndex 查表的 k-th 配对;单测 fixture 照抄了同一个错误契约,于是错的契约被钉成绿的。
  3. 影响范围与严重程度:对用户无回归(未 stamp/降级路径逐位不变,有测试钉住),但功能本体对其目标流量失效:dsh/opencode 任何带工具调用的回合都会静默关掉整条通道(存储去重、Pass -1、anchor.h、fork 映射全部不生效),且只有 info 日志一行「channel active」与真实行为矛盾。严重度:验收 fix: keep system prompt byte-stable for prefix cache #1/Use kernel renderTags:"text-only" + forbid tag echoing in prompt #3 不达,静默降级使其不易被发现。
  4. 怎么修复的:见 F1 修法方向(wireIndex 查表 + 放宽守卫 + buildSession 跨度校验 + fixture 改写 + 端到端用例);F2 的 gate 红已由我直接修掉(472fb02)。
  5. 是否有回归问题:审过的既有行为——未 stamp lane 逐位不变(所有新逻辑都挂在 hostIds !== undefined / hostId !== null 门后,测试钉住);v5→v6 迁移为加性可空列+部分唯一索引,probe 守卫、双向兼容(旧构建在版本门 no-op);append-only 日志的唯一例外(host-key 行原地刷新)边界收紧:同字节 no-op、refcount 释放、seq 稳定保住 tailSeqs 指针;fork 协议 hostIds 加性且在所有 hash 之外,新旧客户端对的 branch-point 比较不受影响;manifest capabilities 加性字段;chain/bypass/unattributed 三条 raw-forward 早退路径位于解析收口点之前,构造上不可能携带 stamp 流量。未发现回归。
  6. 是否破坏缓存:不涉及缓存/重试/增量游标失效策略;watermark 连续性由 host-key seq 稳定性设计保证(F1 修复前该机制休眠);前缀缓存侧:未 stamp 流量字节不变(字段在场才重新序列化),stamp 流量新增的顶层字段在转发上游前被删除,上游字节不变。
  7. 是否改了设计方向:是,且属 owner 已拍板的方向(dsh plugin lane: the host already has a stable MessageId — a per-request identity side-channel would eliminate the #2396 family of fold-identity drift #2938/[identity] R1: ingest host message ids into the canonical store — reconcile becomes exact set algebra instead of byte-hash guessing (#2960 L3) #3464):MESSAGE-IDENTITY.md 旧「future direction」(宿主 id 至多是 join hint)被取代——stamp lane 上宿主 id 成为行主身份,字节降为内容列。我的异议在执行不在方向。
  8. 是否引入了新的配置:无 env/config/CLI 新增。新增的是线级保留字段 _bili_message_ids(转发前剥离)与 manifest capability hostMessageIds——均加性、能力门控、非用户配置面。

判定:拦下,待 F1 修复(含 fixture 改写)+ F3 端到端证据 + F4 rebase 后复审。 F2 已落地(472fb02,CI 重跑中)。

ranxianglei pushed a commit that referenced this pull request Oct 17, 2026
… capability

Review fix for #3473 (separate from the feature commit): the feature adds
capabilities.hostMessageIds to the plugin manifest but did not regenerate
the golden snapshots, so the wire-contract-golden gate fails on CI
(#1304: ledger only grows; goldens change only via explicit regeneration).

Justification for the schema change: additive capability advertisement for
the #3464 R1 host message-id side channel — old plugins ignore the unknown
capability and never stamp; the field name and cap are the channel contract
plugins learn at bootstrap. Regenerated with
node --import tsx scripts/update-wire-contract-goldens.ts; the only delta is
the hostMessageIds block in both manifest snapshots.
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

补一条 review 发现 + 已直接修(commit b143d69): 上一条回复后我核对了新 head 的全量 CI,test lane 还有 2 个红——wire-contract-golden 的 manifest-default / manifest-optin 快照漂移。原因:本 PR 给 plugin manifest 加了 capabilities.hostMessageIds,但没按 #1304 流程重新生成 golden(node --import tsx scripts/update-wire-contract-goldens.ts)。我已本地复现、按官方脚本重生成并核对过 diff:唯一变化就是两个 manifest 快照里新增的 hostMessageIds 块(共 8 行插入),kernel.json/bili.json 无实质变化。schema 变更理由(按 #1304 要求在此声明):这是 #3464 R1 host-id 通道的加性能力广告——旧插件不认识该 capability 就永不 stamp,字段名与上限是插件 bootstrap 时学习的通道契约。golden 测试本地已转绿(26/26)。

至此我在本分支上共推了 2 个独立标注的 review 修复 commit:472fb0261(unused-exports gate)与 b143d6967(golden 重生成)。CI 会在 b143d6967 上全量重跑。合并判定维持不变:仍被 F1(coreHostRefs 配对契约与生产数据形状不符,验收 #1/#3 在真实流量下不成立)拦下,待作者修复 + F3 端到端证据 + rebase 最新 master 后复审。

@ranxianglei
ranxianglei force-pushed the 2026-10-17_host-message-ids branch from b143d69 to 2ef562a Compare October 17, 2026 04:25
ranxianglei pushed a commit that referenced this pull request Oct 17, 2026
Review fix for #3473 (separate from the feature commit): the interface is
only used as extractHostIds' return type inside this file, so the leading
export trips scripts/check-unused-exports.mjs (INTERNAL_ONLY). Same class as
master's 8e110a6. Drops the export; no behavior change.
ranxianglei pushed a commit that referenced this pull request Oct 17, 2026
… capability

Review fix for #3473 (separate from the feature commit): the feature adds
capabilities.hostMessageIds to the plugin manifest but did not regenerate
the golden snapshots, so the wire-contract-golden gate fails on CI
(#1304: ledger only grows; goldens change only via explicit regeneration).

Justification for the schema change: additive capability advertisement for
the #3464 R1 host message-id side channel — old plugins ignore the unknown
capability and never stamp; the field name and cap are the channel contract
plugins learn at bootstrap. Regenerated with
node --import tsx scripts/update-wire-contract-goldens.ts; the only delta is
the hostMessageIds block in both manifest snapshots.
awork added 3 commits October 17, 2026 12:26
…comes exact set algebra (#3464 R1)

Plugin lanes (dsh/opencode) passively read each wire message's own host id
off the outgoing body and stamp it in the reserved top-level field
_bili_message_ids when the proxy advertises capabilities.hostMessageIds.
The proxy extracts+strips the field at the body-parse chokepoint, zips the
wire-aligned list onto cores via wireIndex sidecars, and persists it:

- schema v6: messages gains host_id/host_slot + partial unique index
  (session_id, host_id, host_slot); migration additive/probe-guarded
- planIngest dedups stamped legs by host key — one row per host message
  across drifted resends (fresher bytes win, first-seen position kept);
  unstamped lanes keep today's byte-keyed rows bit-for-bit
- store reuses the host-keyed seq across byte drift and refreshes the row
  in place (documented append-only exception for #2960(5) digest parity)
- fold-reconcile Pass -1 claims missing covered ids by exact host-id
  equality (role-gated, k-th within multi-core groups) ahead of every
  byte-based pass; anchor.h seeds/backfills across passes; unstamped
  lanes are bit-identical
- fork snapshots expose an additive hostIds array mapping each ordered
  message back to its bili ref (outside every hash)

No new config/env surface. No kernel changes.
Review fix for #3473 (separate from the feature commit): the interface is
only used as extractHostIds' return type inside this file, so the leading
export trips scripts/check-unused-exports.mjs (INTERNAL_ONLY). Same class as
master's 8e110a6. Drops the export; no behavior change.
… capability

Review fix for #3473 (separate from the feature commit): the feature adds
capabilities.hostMessageIds to the plugin manifest but did not regenerate
the golden snapshots, so the wire-contract-golden gate fails on CI
(#1304: ledger only grows; goldens change only via explicit regeneration).

Justification for the schema change: additive capability advertisement for
the #3464 R1 host message-id side channel — old plugins ignore the unknown
capability and never stamp; the field name and cap are the channel contract
plugins learn at bootstrap. Regenerated with
node --import tsx scripts/update-wire-contract-goldens.ts; the only delta is
the hostMessageIds block in both manifest snapshots.
@ranxianglei
ranxianglei force-pushed the 2026-10-17_host-message-ids branch from 2ef562a to a94d371 Compare October 17, 2026 04:26
@ranxianglei

Copy link
Copy Markdown
Owner Author

🤖 Powered by ework · qwen3.8-27b

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

冲突已解决,rebase 完成。 新 head:a94d37105(原 b143d6967 → rebase 到最新 master 74a79ab45),分支已 force-push,GitHub 侧 mergeable 已转干净(blocked 仅剩新 head 上的 CI 重跑)。

冲突实况:全分支只有 1 处内容冲突,在 src/agent/dsh-native.ts——master 的 #3329/#3340 把增量 turn 的 ack 归属从「按 URL 取 pending」改成了「按 per-request turn id 取 pending」(onResponse 头三行 + incPendingAcks 键位),而本 PR 恰好把新函数 maybeStampHostIds 插在同一个位置。位置碰撞,不是语义碰撞。解法取两侧意图之并:PR 的 maybeStampHostIds(含注释)逐字保留,master 的 turn-id 归属逻辑逐字保留;beforeSend 里对 maybeStampHostIds 的调用(在增量接管之后、只 stamp 全量 anthropic body)自动合并成功,与两侧语义都自洽。rebase 期间 master 又进了 2 个 commit(#3348 SQLite bootstrap 超时,仅动 engine-bootstrap,与本 PR 零文件重叠),顺手一并 rebase 干净。

定向验证(本地只跑改动相关范围,全量以远程 CI 为准):typecheck 干净;issue3464-host-ids(27)+ dsh-native + incremental-client-recovery + incremental-lane 共 102 过 0 败;fold-reconcile + wire-contract(golden)共 87 过 0 败。

八问

  1. 是什么问题:系统转发的上游 PR 冲突唤醒——本 PR 分支与 master 产生合并冲突,需要 rebase 恢复可合并状态。
  2. 什么原因导致的:master 在本 PR base 之后改了同一文件的相邻区域(增量 ack 归属改按 turn id 键控,[bug] incremental turns (#3248): pending-ack maps are keyed by request URL, not conversation id — a late response mis-attributes x-bili-applied-seq to another conversation and kills incremental turns for both #3329/fix: attribute incremental-turn acks by per-request turn id, not URL (#3329) #3340),与 PR 新增函数插入点重叠,git 无法自动合并该 hunk。
  3. 影响范围与严重程度:仅影响本 PR 分支的可合并性,用户面零影响;不处理则 PR 永远停在 dirty 无法合并。
  4. 怎么修复的:rebase 到最新 master(含 rebase 途中新并入的 fix: align SQLite engine self-download timeout with the updater (10s -> 60s) #3348);唯一冲突按「两侧意图之并」手工解决——新功能函数与 master 的新归属逻辑各自完整保留,无取舍丢弃任何一方行为。
  5. 是否有回归问题:冲突区外的所有文件全部由 git 自动合并(两侧改动零重叠区);冲突区的解决结果已逐行核对为「PR 原文 + master 原文」的精确并集,未引入任何第三形态。既有行为逐项过:增量 lane 的 turn-id 归属(master 语义)不变、host-id stamp 通道(PR 语义)不变、stamp 仍排在增量接管之后只对全量 body 生效。未发现回归。
  6. 是否破坏缓存:不涉及。
  7. 是否改了设计方向:不涉及,纯机械 rebase + 冲突并集。
  8. 是否引入了新的配置:不涉及。

收工判定:冲突已清除,分支骑在最新 master 上,CI 在新 head 上重跑中。注意:这不改变 review 结论——F1(coreHostRefs 配对契约与生产数据形状不符,验收 #1/#3 在真实流量下不成立)仍是拦下项,待作者修复 + F3 端到端证据后复审;本次 rebase 未触碰 F1 相关代码。

@ranxianglei

Copy link
Copy Markdown
Owner Author

Superseded as an implementation: written against the pre-#3642/#3658 identity world (whole-store re-key assumptions, pre-snapshot-render reconcile). The R1 direction — host-declared message ids into the canonical store, exact set-algebra reconcile — remains tracked by #3464, and the current architecture (canonical seeds, media fingerprints, snapshot render seams) is the base it should be rebuilt on. Closing the branch; reopen as a fresh PR against master when R1 is scheduled.

@ranxianglei ranxianglei changed the title feat: host message ids into the canonical store — reconcile becomes exact set algebra (#3464 R1) [PR] feat: host message ids into the canonical store — reconcile becomes exact set algebra (#3464 R1) Oct 20, 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.

[identity] R1: ingest host message ids into the canonical store — reconcile becomes exact set algebra instead of byte-hash guessing (#2960 L3)

1 participant