kaijibot 2026.5.22 → 2026.5.24

package/README.en.md

@@ -17,13 +17,13 @@ Every AI assistant you've used follows the same pattern: you ask, it answers. Yo
 
 KaijiBot is different. After a few conversations on Feishu, it starts **reaching out to you** proactively — not with spam or hydration reminders, but with things you'd actually find interesting.
 
-| | Typical Chatbot | KaijiBot |
-|---|---|---|
-| **Interaction** | Reactive — you ask, it answers | Proactive insights + normal Q&A |
-| **User Understanding** | Stateless, starts from zero every time | Continuously learns your interests, domains, preferences |
-| **Timing Awareness** | Doesn't care what you're doing | Respects active hours, trust level, conversation cadence |
-| **Chinese Support** | English-first, Chinese often breaks | Chinese-native: pattern routing, persona extraction optimized for Chinese |
-| **Integration** | Requires Web/SDK integration | Feishu is the terminal — just send a message |
+|                        | Typical Chatbot                        | KaijiBot                                                                  |
+| ---------------------- | -------------------------------------- | ------------------------------------------------------------------------- |
+| **Interaction**        | Reactive — you ask, it answers         | Proactive insights + normal Q&A                                           |
+| **User Understanding** | Stateless, starts from zero every time | Continuously learns your interests, domains, preferences                  |
+| **Timing Awareness**   | Doesn't care what you're doing         | Respects active hours, trust level, conversation cadence                  |
+| **Chinese Support**    | English-first, Chinese often breaks    | Chinese-native: pattern routing, persona extraction optimized for Chinese |
+| **Integration**        | Requires Web/SDK integration           | Feishu is the terminal — just send a message                              |
 
 ## ✨ Core Features
 
@@ -82,8 +82,8 @@ Step into a pit once. That's enough.
 
 Not locked into any single provider. Switch between domestic and international at will. `kaijibot onboard` wizard auto-discovers configured API keys.
 
-| China (Recommended) | International | Aggregator / Self-hosted |
-|---|---|---|
+| China (Recommended)                                                                                                             | International                                                                        | Aggregator / Self-hosted                                                                                    |
+| ------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------ | ----------------------------------------------------------------------------------------------------------- |
 | Zhipu GLM · DeepSeek · Qwen · Kimi · MiniMax · Baidu Qianfan · StepFun · Volcengine · BytePlus · Kimi Coding · Xiaomi · Alibaba | Claude · Gemini · Grok · Mistral · Perplexity · Groq · Nvidia · HuggingFace · OpenAI | OpenRouter · LiteLLM · Together · Fireworks · Cloudflare AI · Vercel AI · SGLang · vLLM · Ollama · LMStudio |
 
 Switch models with a single command:
@@ -233,18 +233,18 @@ Project scale: `src/agents/` (762 files), `src/infra/` (484), `src/gateway/` (35
 
 **62 extensions** covering all capability layers:
 
-| Category | Extensions |
-|---|---|
-| **Messaging** | feishu (Feishu/Lark) |
-| **China LLMs** | Zhipu GLM · DeepSeek · Qwen · Kimi · MiniMax · Baidu Qianfan · StepFun · Volcengine · BytePlus · Kimi Coding · Xiaomi · Alibaba |
-| **International LLMs** | Claude · Gemini · Grok · Mistral · Perplexity · Groq · Nvidia · HuggingFace · OpenAI |
+| Category                   | Extensions                                                                                                                                 |
+| -------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------ |
+| **Messaging**              | feishu (Feishu/Lark)                                                                                                                       |
+| **China LLMs**             | Zhipu GLM · DeepSeek · Qwen · Kimi · MiniMax · Baidu Qianfan · StepFun · Volcengine · BytePlus · Kimi Coding · Xiaomi · Alibaba            |
+| **International LLMs**     | Claude · Gemini · Grok · Mistral · Perplexity · Groq · Nvidia · HuggingFace · OpenAI                                                       |
 | **Aggregators / Gateways** | OpenRouter · LiteLLM · Together · Fireworks · Cloudflare AI · Vercel AI · Copilot Proxy · Microsoft · Microsoft Foundry · Anthropic Vertex |
-| **Self-hosted** | Ollama · LMStudio · SGLang · vLLM |
-| **Dev Tools** | OpenCode · OpenCode-Go · Open-Prose · OpenShell · Kilocode · Arcee · Chutes · Venice · Vydra · Runway |
-| **Search / Browser** | Exa · Tavily · Browser (Playwright) |
-| **Memory** | Memory-Core · Memory-LanceDB · Memory-Wiki |
-| **Voice / Media** | Speech-Core · Talk-Voice · Media-Understanding · Image-Generation |
-| **Utilities** | Diffs · LLM-Task · Device-Pair · Webhooks · Shared · GitHub-Copilot |
+| **Self-hosted**            | Ollama · LMStudio · SGLang · vLLM                                                                                                          |
+| **Dev Tools**              | OpenCode · OpenCode-Go · Open-Prose · OpenShell · Kilocode · Arcee · Chutes · Venice · Vydra · Runway                                      |
+| **Search / Browser**       | Exa · Tavily · Browser (Playwright)                                                                                                        |
+| **Memory**                 | Memory-Core · Memory-LanceDB · Memory-Wiki                                                                                                 |
+| **Voice / Media**          | Speech-Core · Talk-Voice · Media-Understanding · Image-Generation                                                                          |
+| **Utilities**              | Diffs · LLM-Task · Device-Pair · Webhooks · Shared · GitHub-Copilot                                                                        |
 
 **21 built-in skills**: github, gh-issues, weather, summarize, coding-agent, mcporter, skill-creator, session-logs, healthcheck, notion, obsidian, canvas, nano-pdf, taskflow, taskflow-inbox-triage, clawhub, video-frames, gifgrep, node-connect, blogwatcher, sherpa-onnx-tts.
 
@@ -260,23 +260,23 @@ The cognitive system design draws on the following research:
 
 **Foundational Theory**
 
-- Green, D. M., & Swets, J. A. (1966). *Signal detection theory and psychophysics*. Wiley.
-- Thompson, W. R. (1933). On the likelihood that one unknown probability exceeds another in view of the evidence of two samples. *Biometrika*, 25(3/4), 285–294.
-- Altman, I., & Taylor, D. A. (1973). *Social penetration: The development of interpersonal relationships*. Holt, Rinehart & Winston.
-- Gentner, D. (1983). Structure-mapping: A theoretical framework for analogy. *Cognitive Science*, 7(2), 155–170.
+- Green, D. M., & Swets, J. A. (1966). _Signal detection theory and psychophysics_. Wiley.
+- Thompson, W. R. (1933). On the likelihood that one unknown probability exceeds another in view of the evidence of two samples. _Biometrika_, 25(3/4), 285–294.
+- Altman, I., & Taylor, D. A. (1973). _Social penetration: The development of interpersonal relationships_. Holt, Rinehart & Winston.
+- Gentner, D. (1983). Structure-mapping: A theoretical framework for analogy. _Cognitive Science_, 7(2), 155–170.
 
 **Human-Computer Relationships & Recommender Systems**
 
-- Bickmore, T. W., & Picard, R. W. (2005). Establishing and maintaining long-term human-computer relationships. *ACM Transactions on Computer-Human Interaction*, 12(2), 293–327.
-- Kotkov, D., Wang, S., & Veijalainen, J. (2016). A survey of serendipity in recommender systems. *Knowledge-Based Systems*, 111, 180–192.
+- Bickmore, T. W., & Picard, R. W. (2005). Establishing and maintaining long-term human-computer relationships. _ACM Transactions on Computer-Human Interaction_, 12(2), 293–327.
+- Kotkov, D., Wang, S., & Veijalainen, J. (2016). A survey of serendipity in recommender systems. _Knowledge-Based Systems_, 111, 180–192.
 
 **LLM Persona & Memory**
 
-- DEEPER: Directed Persona Refinement. (2025). *Proceedings of ACL 2025*. 32.2% error reduction via active contradiction resolution in persona maintenance.
-- PERSONAMEM: Persona-Aware Memory in LLMs. (2025). *Proceedings of COLM 2025*. Benchmark showing LLMs achieve ~50% accuracy on evolving profile tasks.
-- DV365: Dynamic User Representations over 365 Days. (2025). *Proceedings of KDD 2025*. Instagram's multi-slicing user embedding architecture.
+- DEEPER: Directed Persona Refinement. (2025). _Proceedings of ACL 2025_. 32.2% error reduction via active contradiction resolution in persona maintenance.
+- PERSONAMEM: Persona-Aware Memory in LLMs. (2025). _Proceedings of COLM 2025_. Benchmark showing LLMs achieve ~50% accuracy on evolving profile tasks.
+- DV365: Dynamic User Representations over 365 Days. (2025). _Proceedings of KDD 2025_. Instagram's multi-slicing user embedding architecture.
 - GemiRec: Gemini-Powered Recommendations. (2025). Xiaohongshu's multi-interest vector architecture with codebook quantization.
-- PIE: Personalized Interest Exploration. (2023). *Proceedings of WWW 2023*. Personalized PageRank with bandit exploration.
+- PIE: Personalized Interest Exploration. (2023). _Proceedings of WWW 2023_. Personalized PageRank with bandit exploration.
 - ProfiLLM: Fully Implicit User Profiling from Chatbot Interactions. (2025).
 
 ### Open Source Dependencies

package/README.md

@@ -18,13 +18,13 @@ Fork of [OpenClaw](https://github.com/openclaw/openclaw) · 飞书 + 30+ LLM 提
 
 KaijiBot 不一样。它在飞书里跟你聊了几次之后，会开始**主动**给你发消息。不是广告，不是提醒喝水，而是你真正可能感兴趣的东西。
 
-| | 普通聊天机器人 | KaijiBot |
-|---|---|---|
-| **交互方式** | 你问它才答 | 主动推送洞察 + 正常对话 |
-| **用户理解** | 无状态，每次从零开始 | 持续学习你的兴趣、领域、偏好 |
-| **时机感知** | 不管你在干嘛 | 尊重活跃时段、信任阶段、对话频率 |
+|              | 普通聊天机器人       | KaijiBot                                       |
+| ------------ | -------------------- | ---------------------------------------------- |
+| **交互方式** | 你问它才答           | 主动推送洞察 + 正常对话                        |
+| **用户理解** | 无状态，每次从零开始 | 持续学习你的兴趣、领域、偏好                   |
+| **时机感知** | 不管你在干嘛         | 尊重活跃时段、信任阶段、对话频率               |
 | **中文支持** | 英文优先，中文常掉队 | 中文原生优化，模式路由、画像提取均针对中文设计 |
-| **渠道集成** | 需要 Web/SDK 接入 | 飞书即终端，发消息就能用 |
+| **渠道集成** | 需要 Web/SDK 接入    | 飞书即终端，发消息就能用                       |
 
 ## ✨ 核心特性
 
@@ -83,8 +83,8 @@ KaijiBot 不会这样。它有一套纠错记忆系统，保证同样的错误
 
 不绑死任何一家。国内国际随意切换，`kaijibot onboard` 向导自动发现已配置的 API Key。
 
-| 国内（推荐） | 国际主流 | 聚合 / 自部署 |
-|---|---|---|
+| 国内（推荐）                                                                                                               | 国际主流                                                                             | 聚合 / 自部署                                                                                               |
+| -------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------ | ----------------------------------------------------------------------------------------------------------- |
 | 智谱 GLM · DeepSeek · 通义千问 · Kimi · MiniMax · 百度千帆 · 阶跃星辰 · 火山引擎 · BytePlus · Kimi Coding · 小米 · Alibaba | Claude · Gemini · Grok · Mistral · Perplexity · Groq · Nvidia · HuggingFace · OpenAI | OpenRouter · LiteLLM · Together · Fireworks · Cloudflare AI · Vercel AI · SGLang · vLLM · Ollama · LMStudio |
 
 切换模型只需一行：
@@ -189,7 +189,7 @@ export TAVILY_API_KEY="your-key"
         → 语义新鲜度检查（LLM 判定与近期洞察是否语义重复）
           → 统一管线生成洞察（模式路由 via timestamp % 100）：
               知识模式（40%）：Web 搜索 + LLM → LLM 自我精炼（评审→改写）→ 质量重试 → LLM 验证
-              模式模式（50%）：对话碎片聚类 → LLM 行为洞察 → 部分验证
+              模式（Pattern）模式（50%）：对话碎片聚类 → LLM 行为洞察 → 部分验证
               延伸模式（10%）：用户已知领域 → LLM 深度建议
             → 对比反重复 + 安全网去重
               → 投递到你的飞书会话
@@ -235,18 +235,18 @@ Agent 系统实现完整的推理循环：系统提示组装（上下文文件 +
 
 **62 个扩展**覆盖全部能力层：
 
-| 类别 | 扩展 |
-|---|---|
-| **消息渠道** | feishu（飞书） |
-| **国内 LLM** | 智谱 GLM · DeepSeek · 通义千问 · Kimi · MiniMax · 百度千帆 · 阶跃星辰 · 火山引擎 · BytePlus · Kimi Coding · 小米 · Alibaba |
-| **国际 LLM** | Claude · Gemini · Grok · Mistral · Perplexity · Groq · Nvidia · HuggingFace · OpenAI |
-| **聚合 / 网关** | OpenRouter · LiteLLM · Together · Fireworks · Cloudflare AI · Vercel AI · Copilot Proxy · Microsoft · Microsoft Foundry · Anthropic Vertex |
-| **自部署** | Ollama · LMStudio · SGLang · vLLM |
-| **开发工具** | OpenCode · OpenCode-Go · Open-Prose · OpenShell · Kilocode · Arcee · Chutes · Venice · Vydra · Runway |
-| **搜索 / 浏览器** | Exa · Tavily · Browser（Playwright） |
-| **记忆** | Memory-Core · Memory-LanceDB · Memory-Wiki |
-| **语音 / 媒体** | Speech-Core · Talk-Voice · Media-Understanding · Image-Generation |
-| **工具类** | Diffs · LLM-Task · Device-Pair · Webhooks · Shared · GitHub-Copilot |
+| 类别              | 扩展                                                                                                                                       |
+| ----------------- | ------------------------------------------------------------------------------------------------------------------------------------------ |
+| **消息渠道**      | feishu（飞书）                                                                                                                             |
+| **国内 LLM**      | 智谱 GLM · DeepSeek · 通义千问 · Kimi · MiniMax · 百度千帆 · 阶跃星辰 · 火山引擎 · BytePlus · Kimi Coding · 小米 · Alibaba                 |
+| **国际 LLM**      | Claude · Gemini · Grok · Mistral · Perplexity · Groq · Nvidia · HuggingFace · OpenAI                                                       |
+| **聚合 / 网关**   | OpenRouter · LiteLLM · Together · Fireworks · Cloudflare AI · Vercel AI · Copilot Proxy · Microsoft · Microsoft Foundry · Anthropic Vertex |
+| **自部署**        | Ollama · LMStudio · SGLang · vLLM                                                                                                          |
+| **开发工具**      | OpenCode · OpenCode-Go · Open-Prose · OpenShell · Kilocode · Arcee · Chutes · Venice · Vydra · Runway                                      |
+| **搜索 / 浏览器** | Exa · Tavily · Browser（Playwright）                                                                                                       |
+| **记忆**          | Memory-Core · Memory-LanceDB · Memory-Wiki                                                                                                 |
+| **语音 / 媒体**   | Speech-Core · Talk-Voice · Media-Understanding · Image-Generation                                                                          |
+| **工具类**        | Diffs · LLM-Task · Device-Pair · Webhooks · Shared · GitHub-Copilot                                                                        |
 
 **21 个内置技能**：github、gh-issues、weather、summarize、coding-agent、mcporter、skill-creator、session-logs、healthcheck、notion、obsidian、canvas、nano-pdf、taskflow、taskflow-inbox-triage、clawhub、video-frames、gifgrep、node-connect、blogwatcher、sherpa-onnx-tts。
 
@@ -262,23 +262,23 @@ Agent 系统实现完整的推理循环：系统提示组装（上下文文件 +
 
 **基础理论**
 
-- Green, D. M., & Swets, J. A. (1966). *Signal detection theory and psychophysics*. Wiley.
-- Thompson, W. R. (1933). On the likelihood that one unknown probability exceeds another in view of the evidence of two samples. *Biometrika*, 25(3/4), 285–294.
-- Altman, I., & Taylor, D. A. (1973). *Social penetration: The development of interpersonal relationships*. Holt, Rinehart & Winston.
-- Gentner, D. (1983). Structure-mapping: A theoretical framework for analogy. *Cognitive Science*, 7(2), 155–170.
+- Green, D. M., & Swets, J. A. (1966). _Signal detection theory and psychophysics_. Wiley.
+- Thompson, W. R. (1933). On the likelihood that one unknown probability exceeds another in view of the evidence of two samples. _Biometrika_, 25(3/4), 285–294.
+- Altman, I., & Taylor, D. A. (1973). _Social penetration: The development of interpersonal relationships_. Holt, Rinehart & Winston.
+- Gentner, D. (1983). Structure-mapping: A theoretical framework for analogy. _Cognitive Science_, 7(2), 155–170.
 
 **人机关系与推荐系统**
 
-- Bickmore, T. W., & Picard, R. W. (2005). Establishing and maintaining long-term human-computer relationships. *ACM Transactions on Computer-Human Interaction*, 12(2), 293–327.
-- Kotkov, D., Wang, S., & Veijalainen, J. (2016). A survey of serendipity in recommender systems. *Knowledge-Based Systems*, 111, 180–192.
+- Bickmore, T. W., & Picard, R. W. (2005). Establishing and maintaining long-term human-computer relationships. _ACM Transactions on Computer-Human Interaction_, 12(2), 293–327.
+- Kotkov, D., Wang, S., & Veijalainen, J. (2016). A survey of serendipity in recommender systems. _Knowledge-Based Systems_, 111, 180–192.
 
 **LLM 画像与记忆**
 
-- DEEPER: Directed Persona Refinement. (2025). *Proceedings of ACL 2025*. 32.2% error reduction via active contradiction resolution in persona maintenance.
-- PERSONAMEM: Persona-Aware Memory in LLMs. (2025). *Proceedings of COLM 2025*. Benchmark showing LLMs achieve ~50% accuracy on evolving profile tasks.
-- DV365: Dynamic User Representations over 365 Days. (2025). *Proceedings of KDD 2025*. Instagram's multi-slicing user embedding architecture.
+- DEEPER: Directed Persona Refinement. (2025). _Proceedings of ACL 2025_. 32.2% error reduction via active contradiction resolution in persona maintenance.
+- PERSONAMEM: Persona-Aware Memory in LLMs. (2025). _Proceedings of COLM 2025_. Benchmark showing LLMs achieve ~50% accuracy on evolving profile tasks.
+- DV365: Dynamic User Representations over 365 Days. (2025). _Proceedings of KDD 2025_. Instagram's multi-slicing user embedding architecture.
 - GemiRec: Gemini-Powered Recommendations. (2025). Xiaohongshu's multi-interest vector architecture with codebook quantization.
-- PIE: Personalized Interest Exploration. (2023). *Proceedings of WWW 2023*. Personalized PageRank with bandit exploration.
+- PIE: Personalized Interest Exploration. (2023). _Proceedings of WWW 2023_. Personalized PageRank with bandit exploration.
 - ProfiLLM: Fully Implicit User Profiling from Chatbot Interactions. (2025).
 
 ### 开源依赖

package/dist/.buildstamp

@@ -1 +1 @@
-{"builtAt":1779441315549,"head":"e635848e21d4d3de9e95ed35a790e4798225f02b"}
+{"builtAt":1779612003655,"head":"888f27c2a68eecd89d7a403d499485c91a2f6d40"}

package/dist/abort-ByczOLAa.js

@@ -0,0 +1,202 @@
+import { r as formatErrorMessage } from "./errors-CGMew-1w.js";
+import { o as normalizeOptionalLowercaseString, s as normalizeOptionalString } from "./string-coerce-CIT1-v_a.js";
+import { S as parseAgentSessionKey } from "./session-key-NGoeHrFI.js";
+import { v as resolveSessionAgentId } from "./agent-scope-DBfnuGjF.js";
+import { r as logVerbose } from "./globals-D7B2Z7Cq.js";
+import { a as resolveSessionStoreEntry, c as updateSessionStore } from "./store-in8V6KgD.js";
+import "./sessions-DKqgEiPs.js";
+import { l as resolveStorePath } from "./paths-je0O0nOS.js";
+import { t as loadSessionStore } from "./store-load-Cv6vCPDa.js";
+import { o as getLatestSubagentRunByChildSessionKey, p as markSubagentRunTerminated, u as listSubagentRunsForController } from "./subagent-registry-_DLmdEei.js";
+import { b as resolveMainSessionAlias, y as resolveInternalSessionKey } from "./subagents-format-BdPgiCkD.js";
+import { i as clearSessionQueues } from "./queue-BMWETISN.js";
+import { h as replyRunRegistry, t as abortEmbeddedPiRun } from "./runs-Bk3tcSZy.js";
+import { n as getAcpSessionManager } from "./manager-DpCxUErb.js";
+import "./pi-embedded-NCafz2Bo.js";
+import { i as setAbortMemory, n as isAbortRequestText } from "./abort-primitives-zUIoyBbh.js";
+import { o as stripMentions, s as stripStructuralPrefixes } from "./mentions-D-La2eW8.js";
+import { t as resolveCommandAuthorization } from "./command-auth-C6ggRah-.js";
+import { a as shouldPersistAbortCutoff, i as resolveAbortCutoffFromContext, t as applyAbortCutoffToSessionEntry } from "./abort-cutoff-C9Oge2mp.js";
+const abortDeps = {
+	getAcpSessionManager,
+	abortEmbeddedPiRun,
+	getLatestSubagentRunByChildSessionKey,
+	listSubagentRunsForController,
+	markSubagentRunTerminated
+};
+function formatAbortReplyText(stoppedSubagents) {
+	if (typeof stoppedSubagents !== "number" || stoppedSubagents <= 0) return "⚙️ Agent was aborted.";
+	return `⚙️ Agent was aborted. Stopped ${stoppedSubagents} ${stoppedSubagents === 1 ? "sub-agent" : "sub-agents"}.`;
+}
+function resolveSessionEntryForKey(store, sessionKey) {
+	if (!store || !sessionKey) return {};
+	const resolved = resolveSessionStoreEntry({
+		store,
+		sessionKey
+	});
+	if (resolved.existing) return resolved.legacyKeys.length > 0 ? {
+		entry: resolved.existing,
+		key: resolved.normalizedKey,
+		legacyKeys: resolved.legacyKeys
+	} : {
+		entry: resolved.existing,
+		key: resolved.normalizedKey
+	};
+	return {};
+}
+function normalizeRequesterSessionKey(cfg, key) {
+	const cleaned = normalizeOptionalString(key);
+	if (!cleaned) return;
+	const { mainKey, alias } = resolveMainSessionAlias(cfg);
+	return resolveInternalSessionKey({
+		key: cleaned,
+		alias,
+		mainKey
+	});
+}
+function stopSubagentsForRequester(params) {
+	const requesterKey = normalizeRequesterSessionKey(params.cfg, params.requesterSessionKey);
+	if (!requesterKey) return { stopped: 0 };
+	const dedupedRunsByChildKey = /* @__PURE__ */ new Map();
+	for (const run of abortDeps.listSubagentRunsForController(requesterKey)) {
+		const childKey = normalizeOptionalString(run.childSessionKey);
+		if (!childKey) continue;
+		const latest = abortDeps.getLatestSubagentRunByChildSessionKey(childKey);
+		if (!latest) {
+			const existing = dedupedRunsByChildKey.get(childKey);
+			if (!existing || run.createdAt >= existing.createdAt) dedupedRunsByChildKey.set(childKey, run);
+			continue;
+		}
+		const latestControllerSessionKey = normalizeOptionalString(latest?.controllerSessionKey) ?? normalizeOptionalString(latest?.requesterSessionKey);
+		if (latest.runId !== run.runId || latestControllerSessionKey !== requesterKey) continue;
+		const existing = dedupedRunsByChildKey.get(childKey);
+		if (!existing || run.createdAt >= existing.createdAt) dedupedRunsByChildKey.set(childKey, run);
+	}
+	const runs = Array.from(dedupedRunsByChildKey.values());
+	if (runs.length === 0) return { stopped: 0 };
+	const storeCache = /* @__PURE__ */ new Map();
+	const seenChildKeys = /* @__PURE__ */ new Set();
+	let stopped = 0;
+	for (const run of runs) {
+		const childKey = normalizeOptionalString(run.childSessionKey);
+		if (!childKey || seenChildKeys.has(childKey)) continue;
+		seenChildKeys.add(childKey);
+		if (!run.endedAt) {
+			const cleared = clearSessionQueues([childKey]);
+			const parsed = parseAgentSessionKey(childKey);
+			const storePath = resolveStorePath(params.cfg.session?.store, { agentId: parsed?.agentId });
+			let store = storeCache.get(storePath);
+			if (!store) {
+				store = loadSessionStore(storePath);
+				storeCache.set(storePath, store);
+			}
+			const entry = store[childKey];
+			const sessionId = replyRunRegistry.resolveSessionId(childKey) ?? entry?.sessionId;
+			const aborted = (childKey ? replyRunRegistry.abort(childKey) : false) || (sessionId ? abortDeps.abortEmbeddedPiRun(sessionId) : false);
+			if (abortDeps.markSubagentRunTerminated({
+				runId: run.runId,
+				childSessionKey: childKey,
+				reason: "killed"
+			}) > 0 || aborted || cleared.followupCleared > 0 || cleared.laneCleared > 0) stopped += 1;
+		}
+		const cascadeResult = stopSubagentsForRequester({
+			cfg: params.cfg,
+			requesterSessionKey: childKey
+		});
+		stopped += cascadeResult.stopped;
+	}
+	if (stopped > 0) logVerbose(`abort: stopped ${stopped} subagent run(s) for ${requesterKey}`);
+	return { stopped };
+}
+async function tryFastAbortFromMessage(params) {
+	const { ctx, cfg } = params;
+	const targetKey = normalizeOptionalString(ctx.CommandTargetSessionKey) ?? normalizeOptionalString(ctx.SessionKey);
+	const raw = stripStructuralPrefixes(ctx.CommandBody ?? ctx.RawBody ?? ctx.Body ?? "");
+	if (!isAbortRequestText(normalizeOptionalLowercaseString(ctx.ChatType) === "group" ? stripMentions(raw, ctx, cfg, resolveSessionAgentId({
+		sessionKey: targetKey ?? ctx.SessionKey ?? "",
+		config: cfg
+	})) : raw)) return {
+		handled: false,
+		aborted: false
+	};
+	const commandAuthorized = ctx.CommandAuthorized;
+	const auth = resolveCommandAuthorization({
+		ctx,
+		cfg,
+		commandAuthorized
+	});
+	if (!auth.isAuthorizedSender) return {
+		handled: false,
+		aborted: false
+	};
+	const agentId = resolveSessionAgentId({
+		sessionKey: targetKey ?? ctx.SessionKey ?? "",
+		config: cfg
+	});
+	const abortKey = targetKey ?? auth.from ?? auth.to;
+	const requesterSessionKey = targetKey ?? ctx.SessionKey ?? abortKey;
+	if (targetKey) {
+		const storePath = resolveStorePath(cfg.session?.store, { agentId });
+		const store = loadSessionStore(storePath);
+		const { entry, key, legacyKeys } = resolveSessionEntryForKey(store, targetKey);
+		const resolvedTargetKey = key ?? targetKey;
+		const acpManager = abortDeps.getAcpSessionManager();
+		if (acpManager.resolveSession({
+			cfg,
+			sessionKey: resolvedTargetKey
+		}).kind !== "none") try {
+			await acpManager.cancelSession({
+				cfg,
+				sessionKey: resolvedTargetKey,
+				reason: "fast-abort"
+			});
+		} catch (error) {
+			logVerbose(`abort: ACP cancel failed for ${resolvedTargetKey}: ${formatErrorMessage(error)}`);
+		}
+		const sessionId = replyRunRegistry.resolveSessionId(resolvedTargetKey) ?? entry?.sessionId;
+		const aborted = replyRunRegistry.abort(resolvedTargetKey) || (sessionId ? abortDeps.abortEmbeddedPiRun(sessionId) : false);
+		const cleared = clearSessionQueues([resolvedTargetKey, sessionId]);
+		if (cleared.followupCleared > 0 || cleared.laneCleared > 0) logVerbose(`abort: cleared followups=${cleared.followupCleared} lane=${cleared.laneCleared} keys=${cleared.keys.join(",")}`);
+		const abortCutoff = shouldPersistAbortCutoff({
+			commandSessionKey: ctx.SessionKey,
+			targetSessionKey: resolvedTargetKey
+		}) ? resolveAbortCutoffFromContext(ctx) : void 0;
+		if (entry && key) {
+			entry.abortedLastRun = true;
+			applyAbortCutoffToSessionEntry(entry, abortCutoff);
+			entry.updatedAt = Date.now();
+			store[key] = entry;
+			for (const legacyKey of legacyKeys ?? []) if (legacyKey !== key) delete store[legacyKey];
+			await updateSessionStore(storePath, (nextStore) => {
+				const nextEntry = nextStore[key] ?? entry;
+				if (!nextEntry) return;
+				nextEntry.abortedLastRun = true;
+				applyAbortCutoffToSessionEntry(nextEntry, abortCutoff);
+				nextEntry.updatedAt = Date.now();
+				nextStore[key] = nextEntry;
+				for (const legacyKey of legacyKeys ?? []) if (legacyKey !== key) delete nextStore[legacyKey];
+			});
+		} else if (abortKey) setAbortMemory(abortKey, true);
+		const { stopped } = stopSubagentsForRequester({
+			cfg,
+			requesterSessionKey
+		});
+		return {
+			handled: true,
+			aborted,
+			stoppedSubagents: stopped
+		};
+	}
+	if (abortKey) setAbortMemory(abortKey, true);
+	const { stopped } = stopSubagentsForRequester({
+		cfg,
+		requesterSessionKey
+	});
+	return {
+		handled: true,
+		aborted: false,
+		stoppedSubagents: stopped
+	};
+}
+//#endregion
+export { tryFastAbortFromMessage as i, resolveSessionEntryForKey as n, stopSubagentsForRequester as r, formatAbortReplyText as t };

package/dist/abort-cutoff.runtime-DKpU2Syx.js

@@ -0,0 +1,20 @@
+import { c as updateSessionStore } from "./store-in8V6KgD.js";
+import { n as hasAbortCutoff, t as applyAbortCutoffToSessionEntry } from "./abort-cutoff-C9Oge2mp.js";
+//#region src/auto-reply/reply/abort-cutoff.runtime.ts
+async function clearAbortCutoffInSessionRuntime(params) {
+	const { sessionEntry, sessionStore, sessionKey, storePath } = params;
+	if (!sessionEntry || !sessionStore || !sessionKey || !hasAbortCutoff(sessionEntry)) return false;
+	applyAbortCutoffToSessionEntry(sessionEntry, void 0);
+	sessionEntry.updatedAt = Date.now();
+	sessionStore[sessionKey] = sessionEntry;
+	if (storePath) await updateSessionStore(storePath, (store) => {
+		const existing = store[sessionKey] ?? sessionEntry;
+		if (!existing) return;
+		applyAbortCutoffToSessionEntry(existing, void 0);
+		existing.updatedAt = Date.now();
+		store[sessionKey] = existing;
+	});
+	return true;
+}
+//#endregion
+export { clearAbortCutoffInSessionRuntime };

package/dist/abort-cutoff.runtime.js

@@ -1 +1 @@
-export * from "./abort-cutoff.runtime-Bid5Yzv_.js";
+export * from "./abort-cutoff.runtime-DKpU2Syx.js";

package/dist/abort-primitives-zUIoyBbh.js

@@ -0,0 +1,93 @@
+import { i as normalizeLowercaseStringOrEmpty } from "./string-coerce-CIT1-v_a.js";
+import { p as normalizeCommandBody } from "./commands-registry-DCry6Ori.js";
+//#region src/auto-reply/reply/abort-primitives.ts
+const ABORT_TRIGGERS = new Set([
+	"stop",
+	"esc",
+	"abort",
+	"wait",
+	"exit",
+	"interrupt",
+	"detente",
+	"deten",
+	"detén",
+	"arrete",
+	"arrête",
+	"停止",
+	"やめて",
+	"止めて",
+	"रुको",
+	"توقف",
+	"стоп",
+	"остановись",
+	"останови",
+	"остановить",
+	"прекрати",
+	"halt",
+	"anhalten",
+	"aufhören",
+	"hoer auf",
+	"stopp",
+	"pare",
+	"stop kaijibot",
+	"kaijibot stop",
+	"stop action",
+	"stop current action",
+	"stop run",
+	"stop current run",
+	"stop agent",
+	"stop the agent",
+	"stop don't do anything",
+	"stop dont do anything",
+	"stop do not do anything",
+	"stop doing anything",
+	"do not do that",
+	"please stop",
+	"stop please"
+]);
+const ABORT_MEMORY = /* @__PURE__ */ new Map();
+const ABORT_MEMORY_MAX = 2e3;
+const TRAILING_ABORT_PUNCTUATION_RE = /[.!?…,，。;；:：'"’”)\]}]+$/u;
+function normalizeAbortTriggerText(text) {
+	return normalizeLowercaseStringOrEmpty(text).replace(/[’`]/g, "'").replace(/\s+/g, " ").replace(TRAILING_ABORT_PUNCTUATION_RE, "").trim();
+}
+function isAbortTrigger(text) {
+	if (!text) return false;
+	const normalized = normalizeAbortTriggerText(text);
+	return ABORT_TRIGGERS.has(normalized);
+}
+function isAbortRequestText(text, options) {
+	if (!text) return false;
+	const normalized = normalizeCommandBody(text, options).trim();
+	if (!normalized) return false;
+	const normalizedLower = normalizeLowercaseStringOrEmpty(normalized);
+	return normalizedLower === "/stop" || normalizeAbortTriggerText(normalizedLower) === "/stop" || isAbortTrigger(normalizedLower);
+}
+function getAbortMemory(key) {
+	const normalized = key.trim();
+	if (!normalized) return;
+	return ABORT_MEMORY.get(normalized);
+}
+function pruneAbortMemory() {
+	if (ABORT_MEMORY.size <= ABORT_MEMORY_MAX) return;
+	const excess = ABORT_MEMORY.size - ABORT_MEMORY_MAX;
+	let removed = 0;
+	for (const entryKey of ABORT_MEMORY.keys()) {
+		ABORT_MEMORY.delete(entryKey);
+		removed += 1;
+		if (removed >= excess) break;
+	}
+}
+function setAbortMemory(key, value) {
+	const normalized = key.trim();
+	if (!normalized) return;
+	if (!value) {
+		ABORT_MEMORY.delete(normalized);
+		return;
+	}
+	if (ABORT_MEMORY.has(normalized)) ABORT_MEMORY.delete(normalized);
+	ABORT_MEMORY.set(normalized, true);
+	pruneAbortMemory();
+}
+//#endregion
+export { setAbortMemory as i, isAbortRequestText as n, isAbortTrigger as r, getAbortMemory as t };

package/dist/abort.runtime-BbfUtrW4.js

@@ -0,0 +1,2 @@
+import { i as tryFastAbortFromMessage, t as formatAbortReplyText } from "./abort-ByczOLAa.js";
+export { formatAbortReplyText, tryFastAbortFromMessage };

package/dist/abort.runtime.js

@@ -1 +1 @@
-export * from "./abort.runtime-BcGW7zNU.js";
+export * from "./abort.runtime-BbfUtrW4.js";