dsh-vibe-math 2.0.19 → 2.0.20

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "dsh-vibe-math",
   "description": "Multi-agent mathematical problem-solving & verification frameworks for DeepSeek Harness — THREE agent presets in one install: vibe-math-v2 (probability-driven: qs.json + Propos knowledge base + explorer→solver→review/debate verdict), vibe-math-v3 (THIRD-generation, recommended: paper-style Markdown knowledge base with Problems/Progress/Propos/Methods/Verified + planner-agent scheduling that decides the next N actions + universal theory/method invention library + agents write their own Markdown directly via a per-file write lock), and vibe-math-v4 (FOURTH-generation: persistent self-organizing resident subagents that message & meet to decide all tasks, verify only by unanimous consensus, /compact at a context threshold, and stop only when all agree the problem is solved). Installing this bundle auto-installs all three presets (v1 was removed at v2.0.0).",
-  "version": "2.0.19",
+  "version": "2.0.20",
   "type": "module",
   "engines": {
     "node": "^22.19.0 || >=24.0.0"

package/vibe-math-v4/vibe-math-v4.js

@@ -338,8 +338,10 @@ export function apply(ctx) {
     async function writeTaskboard(){ const lines=['# 任务板','']; for(const t of taskboard){ lines.push('- ['+t.status+'] '+t.title+(t.claimer?('（认领:'+t.claimer+'）'):'')+(t.proposer?('（提议:'+t.proposer+'）'):'')+(t.description?('：'+t.description):'')) } await writeText('Shared/taskboard.md',lines.join('\n')) }
     async function proposeTask(title,description,proposer){ const id='t-'+shortId(); taskboard.push({id,title:String(title),description:String(description||''),status:'open',proposer:proposer||'',claimer:'',source:''}); await saveTaskboard(); markProgress(); logActivity('task','proposed '+id+'「'+title+'」'); return {ok:true,id} }
     async function claimTask(id,claimer){ const t=taskboard.find(x=>x.id===id); if(!t) return {ok:false,message:'no such task'}; if(t.status!=='open') return {ok:false,message:'task already '+t.status}; t.status='claimed'; t.claimer=claimer; await saveTaskboard(); markProgress(); logActivity('task',claimer+' claimed '+id); 
-      // wake the claimer to work on it (framework moves the task, resident decides how)
-      const r=residents.get(claimer); if(r && !busy.has(claimer)){ currentResident=claimer; await wakeResident(r, (await normalPrompt(r))+'\n\n[YOU CLAIMED TASK '+id+'] '+t.title+' — '+t.description,'normal'); await saveAll() }
+      // wake the claimer to work on it (framework moves the task, resident decides how).
+      // NOT while paused/stopped: a paused run must not start new work — the claim is recorded on the
+      // board and the resident (who claimed it) picks it up again after resume.
+      const r=residents.get(claimer); if(r && !busy.has(claimer) && running && !autoDone){ currentResident=claimer; await wakeResident(r, (await normalPrompt(r))+'\n\n[YOU CLAIMED TASK '+id+'] '+t.title+' — '+t.description,'normal'); await saveAll() }
       return {ok:true} }
     async function taskDone(id,claimer){ const t=taskboard.find(x=>x.id===id); if(!t) return {ok:false}; t.status='done'; t.doneBy=claimer; await saveTaskboard(); await writeTaskboard(); markProgress(); logActivity('task','done '+id); return {ok:true} }
     async function saveTaskboard(){ await writeJson('State/taskboard.json',taskboard); await writeTaskboard() }
@@ -348,7 +350,7 @@ export function apply(ctx) {
     // Apply context/compact bookkeeping from a resident's reply, so the flag can clear even when
     // the reply came through a meeting/verify branch (defensive) as well as the normal branch.
     function postmark(r, parsed){
-      if(typeof parsed.contextPct==='number') r.contextPct=clPct(parsed.contextPct)
+      const cp=Number(parsed.contextPct); if(Number.isFinite(cp)) r.contextPct=clPct(cp)   // tolerate numeric strings ("40")
       if(parsed.compacted===true || (r.needCompact && parsed.summary)){
         r.contextSeed=String(parsed.summary||r.contextSeed||'')
         r.contextPct=Math.min(r.contextPct||15,25)
@@ -444,7 +446,10 @@ export function apply(ctx) {
         const lines=['# 会议 '+st.id+'｜'+fmtTime(),'','**议程**：'+st.agenda,'']
         for(const [id,iv] of Object.entries(st.inputs)){ lines.push('### '+id); lines.push(iv.input||''); lines.push('') }
         await writeText('Shared/meetings/'+st.id+'.md', lines.join('\n'))
-        meetings.push({id:st.id,agenda:st.agenda,at:now(),inputs:st.inputs})
+        // the full transcript lives on disk (Shared/meetings/<id>.md); the State array keeps a small
+        // index (id/agenda/at) so session.json does not carry a second copy of every transcript and
+        // rewrite it on EVERY saveAll during long runs (reports below are capped for the same reason).
+        meetings.push({id:st.id,agenda:st.agenda,at:now()}); if(meetings.length>200) meetings.shift()
         logDecision('meeting',st.agenda)
         // handle what the meeting produced: task proposals/claims, verify targets, stop vote
         for(const [id,iv] of Object.entries(st.inputs)){
@@ -823,7 +828,7 @@ export function apply(ctx) {
       }
       // normal turn
       if(r.status==='brainstorm'){ r.insight=parsed.summary||output; r.status='active' }
-      if(typeof parsed.solved==='boolean') reports.push({rId:r.rId,solved:parsed.solved,summary:parsed.summary||'',at:now()})
+      if(typeof parsed.solved==='boolean'){ reports.push({rId:r.rId,solved:parsed.solved,summary:parsed.summary||'',at:now()}); if(reports.length>100) reports.shift() }   // solved-signal ring (not surfaced in status/report)
       if(parsed.propose_verify) maybeQueueVerify(parsed.propose_verify, r.rId)
       // group-conversation relay: the resident may choose to speak to the whole team (input) —
       // forward it to the others so this is a real discussion group, not private monologues.
@@ -865,7 +870,15 @@ export function apply(ctx) {
       await saveAll(); return {ok:true,message:'v4 started: '+params.residentCount+' resident(s) brainstorming',project:currentProject}
     }
     async function resume(){
-      currentProject=await readCurrentProject(); await ensureDirs(); await loadAll(); await loadSettings()
+      currentProject=await readCurrentProject(); await ensureDirs()
+      // Resume is only meaningful for a stopped/paused/crashed run. If THIS process is already driving
+      // a live run whose disk state belongs to it, loadAll below would overwrite the in-memory state
+      // with a slightly stale snapshot (busy marks, wake round counters, mailbox contents from the last
+      // saveAll) — a silent clobber for a useless "kick". No-op instead. A cross-process restart is
+      // always allowed: its disk epoch differs, so the in-memory state is empty/stale anyway.
+      const pre=await readJson('State/session.json')
+      if(running && !autoDone && pre && pre.processEpoch===processEpoch) return {ok:true,message:'already running (no-op)'}
+      await loadAll(); await loadSettings()
       // A run the group CONCLUDED (unanimous voteSolved → autoDone) must not be silently revived into
       // a zombie that keeps waking residents with no consensus that it should still run. The group
       // decided it is done; continuing means a NEW run (vibe_v4_start / vibe_v4_configure).
@@ -953,6 +966,11 @@ export function apply(ctx) {
     // Create/configure a project and set params/problem WITHOUT starting any resident.
     // The intended flow: vibe_v4_configure {project?, problem?, params?}  →  vibe_v4_start {}.
     async function configure(cfg){
+      // configure is the PRE-START setup tool (project/problem/params). Switching the project while a
+      // run is LIVE would split the run's state across two trees: residents' briefs & libraries point
+      // at the OLD frameworkRoot while every subsequent saveAll/transcript/Verified card would go to the
+      // NEW project. Params tuning mid-run belongs to vibe_v4_set.
+      if(running && !autoDone) return {ok:false,message:'cannot configure while a run is running (pause or abort first; use vibe_v4_set to tune params)'}
       if(cfg && cfg.project && String(cfg.project).trim()) currentProject=String(cfg.project).trim()
       if(cfg && cfg.problem) problemText=String(cfg.problem)
       if(cfg && cfg.params && typeof cfg.params==='object') setParams(cfg.params)
@@ -962,7 +980,13 @@ export function apply(ctx) {
       await saveAll()
       return {ok:true,project:currentProject,problem:problemText?problemText.slice(0,60):'',params:Object.keys(params).map(k=>k+'='+params[k]).join(', ')}
     }
-    async function initAbort(){ clearHeartbeat(); running=false; phase='idle'; autoDone=false; for(const [,r] of residents){ if(r.childId){ try{ subagents.interrupt(r.childId,{kind:'ancestor',agent:rootAgent}) }catch(e){} } r.childId=''; r.lastActiveAt=0; r.roundsSinceCompact=0 } await saveAll(); return {ok:true,message:'aborted'} }
+    async function initAbort(){ clearHeartbeat(); running=false; phase='idle'; autoDone=false; for(const [,r] of residents){ if(r.childId){ try{ subagents.interrupt(r.childId,{kind:'ancestor',agent:rootAgent}) }catch(e){} } r.childId=''; r.lastActiveAt=0; r.roundsSinceCompact=0 }
+      // Wipe the coordination state too: an aborted run must not report an in-flight meeting/verify,
+      // a parked meeting, a verify queue, or busy residents (their childIds are gone, so no end event
+      // can ever clear those marks). resume()/start() re-initialize anyway; this keeps status truthful
+      // between abort and the next action.
+      meetingState=null; verifyState=null; pendingMeeting=null; pendingVerify=[]; busy=new Set(); wakeKind=new Map(); currentResident=''; finalizeLock=null
+      await saveAll(); return {ok:true,message:'aborted'} }
     function setPause(){ clearHeartbeat(); running=false; return {ok:true,message:'paused'} }
 
     return {

package/vibe-math-v4//345/256/236/347/216/260/346/226/271/346/241/210.md

@@ -570,7 +570,22 @@ VibeMath/Projects/<project>/
 >
 > 说明：无法对"某常驻真的拒绝/掉线"的情况达成全体一致时，看门狗会把该对象保留为"未定论/带概率"，这是哲学上期望的诚实结果。`session.json` 仍不持久化 `meetingState`（进行中的会议不跨进程恢复），B 的停滞看门狗在重启后仍会触发，因此重启也能自愈。
 >
-> 全套测试仍全绿：`selfdrive-v4` 21/21、`e2e-v4-fixes` 94/94（T1–T32，后续审计回归见 §27/§28）、v3 100/100、v2 business 18/regression 14、multisession 25/25。
+> 全套测试仍全绿：`selfdrive-v4` 21/21、`e2e-v4-fixes` 107/107（T1–T37，后续审计回归见 §27/§28/§29）、v3 100/100、v2 business 18/regression 14、multisession 25/25。
+
+## 29. 第五轮全面深度审计（npm v2.0.20，遗留区域 + 未修改逻辑）
+
+对历轮尚未触达的"牵连对象/未改逻辑"做复查，发现并修复 5 处（新增回归 T33–T37）：
+
+- **运行中 resume 用陈旧磁盘快照覆盖内存态（静默 clobber）**：resume 先 `loadAll` 再用磁盘状态覆盖内存——若某进程已在本进程内驱动着活 run（有人把 resume 当"踢一脚"用），内存里比磁盘新一个事件的 busy/轮次/邮箱状态会被回退到最近一次 saveAll 的快照。修复：resume 开头预读 `session.json`，若内存 `running && !autoDone` **且磁盘 epoch 属于本进程**（同进程活 run），直接返回 `{ok:true,message:'already running (no-op)'}`——不做任何加载；跨进程重启（epoch 不同）不受影响（A4 场景依旧放行）。**T33**。
+- **configure 在运行中切换项目 → 状态分裂到两棵树**：configure 是"启动前"工具，但运行中调用会把 currentProject 改到新项目：常驻的简报/资料库仍在旧树，而之后的 saveAll/会议纪要/Verified 卡全写到新树。修复：`running && !autoDone` 时拒绝并提示改用 `vibe_v4_set` 调参数。**T34**。
+- **abort 后 status 谎报在途协调态**：initAbort 只停调度并清 childId，会议/验证/暂存会议/验证队列/busy 全部残留——被中断常驻的 end 因 childId 已清无法命中，busy 印记永远无人清除，status 显示"会议进行中/验证中/busy=[r-1]"，而 run 早已 running=false。修复：initAbort 一并清空 `meetingState/verifyState/pendingMeeting/pendingVerify/busy/wakeKind/currentResident/finalizeLock`。**T35**。
+- **暂停期间 claim 任务会僵尸唤醒认领者**：v2.0.18 的暂停门控只覆盖了 end 处理器；`claimTask` 的直接唤醒绕过了它——暂停时某个 end 回复里 `claim_task`（或宿主工具认领）会把认领者拉起来做一轮新工作。修复：认领记录照常入板，但唤醒加 `running && !autoDone` 门控（认领者自己知道认领了什么，恢复后自然接续）。**T36**。
+- **reports/meetings 无界膨胀（session.json 每事件全量重写）**：`reports` 只写不读，且每个普通轮 end（模板自带 `solved:false`）都 push 一条；`meetings` 数组把**完整会议全文**再存一份（磁盘 `Shared/meetings/<id>.md` 已有），长 run 中 session.json 越来越大、每次 saveAll 都重写整份。修复：reports 上限 100（环形），meetings 只存 {id,agenda,at} 小索引（上限 200）——全文仍在磁盘，会议计数/报告不受影响。**T37**。
+- **连带小修**：`postmark` 的 contextPct 只认 number，模型报数字字符串（"40"）时不生效；现兼容数字字符串。
+
+> 边界复查（未改，记录）：暂停中 `postMessage`（显式宿主动作）仍会投递唤醒（与"认领"不同，是人的明确指令）；brainstorm 期首轮 end 永不触发、以及"邮箱在超长共识链后送达"维持 §27/§28 的设计边界。
+>
+> 全套测试仍全绿：`selfdrive-v4` 21/21、`e2e-v4-fixes` 107/107（T1–T37）、v3 100/100、v2 business 18/regression 14、multisession 25/25、selfdrive-v3 0 异常。
 
 ## 27. 第三轮全面深度审计（npm v2.0.18，牵连对象/逻辑）