billion-context-pi 0.1.70 → 0.1.71-pr.457.264

package/README.md

@@ -79,6 +79,8 @@ That's it. The extension auto-loads on next Pi startup. No configuration needed
 > **Using another sub-agent extension?** billion-context-pi ships its own `acp_delegate` sub-agent tool (see below) at a fraction of the context cost (~600 tok vs ~7K tok/turn). Two delegation tools in one session only make the model's choice noisier, so pick one:
 > - **Use ACP's delegate** — remove the other extension: `pi remove npm:pi-subagents`
 > - **Keep your own sub-agent** — turn ACP's delegate off in `acp.json`: `{ "delegate": false }` (see *Using your own sub-agent instead* below)
+>
+> If you keep `pi-subagents` installed, billion-context-pi detects it at session start: a **project-level** install (`<cwd>/.pi/npm` or the project extensions dir) automatically stands `acp_delegate` down for that project — a reminder then tells you how to give pi-subagents' agents ACP compression via `/acp-subagents`. A **user-level-only** install (`~/.pi/npm`, user extensions dir) leaves `acp_delegate` active and logs a warning instead. Set `"delegate": { "forceEnable": true }` in `acp.json` to keep `acp_delegate` active regardless of detection.
 
 ## How it works
 
@@ -120,6 +122,8 @@ billion-context-pi is built for the **Pi** coding agent (`@earendil-works/pi-cod
 
 - **Coexisting with the [billion-context](https://github.com/ranxianglei/billion-context) wire proxy** — running both on the same session double-compresses every request (wasted tokens, nested summaries, two ref coordinate systems). This is prevented automatically: launcher paths (`bili pi`, …) export `BILLION_CONTEXT_PROXY`, and models whose `baseUrl` routes through the proxy (`…/bili/https://upstream…`) are detected at session start — in both cases billion-context-pi stands down with a warning and leaves the proxy as the sole compressor. One exception: transparent mode, where traffic reaches the proxy via `HTTPS_PROXY` so the URL carries no `/bili/` prefix — that is undetectable from the URL, so export `BILLION_CONTEXT_PROXY=1` before starting pi in that case.
 
+- **Reverse collision — the `billion-context` thin plugin shadows `/acp`.** If the pi plugin shipped inside [`billion-context`](https://github.com/ranxianglei/billion-context) is also installed (an entry pointing at a `billion-context` install in `"packages"` of `~/.pi/agent/settings.json`, e.g. from `bili plugin install pi`), it registers its own `/acp` command, and Pi has no duplicate-command guard — which one answers depends on load order. Telling symptom: on a plain `pi` launch, typing `/acp` shows ``bili: no proxy detected (run via `bili <client>` or set a /bili/ baseURL)`` — that string is emitted by the thin plugin, **not** by billion-context-pi, and pi does not need `bili` running. Fix: the collision only bites when you launch plain `pi` — remove the thin plugin there (`bili plugin remove pi`, or drop the non-`npm:billion-context-pi` entry from that settings file) so billion-context-pi owns `/acp`. If you launch via `bili pi` instead, nothing needs removing: the launcher exports `BILLION_CONTEXT_PROXY` and billion-context-pi stands down entirely (bullet above), so both plugins may stay installed side by side. Note: `bili plugin install pi` removes an existing `npm:billion-context-pi` entry from that file without saying so ([billion-context#788](https://github.com/ranxianglei/billion-context/issues/788)).
+
 ## Model-facing tools
 
 | Tool | What it does |

package/README.zh-CN.md

@@ -78,6 +78,8 @@ pi install npm:billion-context-pi
 > **你另有子代理扩展?** billion-context-pi 自带 `acp_delegate` 子代理工具(见下文),上下文成本极低(~600 tok vs ~7K tok/轮)。同一会话里两套委派工具只会让模型的选择更混乱,二选一:
 > - **用 ACP 的 delegate** —— 卸载另一个扩展:`pi remove npm:pi-subagents`
 > - **保留你自己的子代理** —— 在 `acp.json` 里关掉 ACP 的 delegate:`{ "delegate": false }`(见下文*改用你自己的子代理*)
+>
+> 若保留已安装的 `pi-subagents`,billion-context-pi 会在会话启动时检测:**项目级**安装(`<cwd>/.pi/npm` 或项目内 extensions 目录)会自动停用该项目的 `acp_delegate`,并提醒你运行 `/acp-subagents` 让 pi-subagents 的子代理获得 ACP 压缩;仅**用户级**(全局)安装时只记一条警告日志,`acp_delegate` 保持启用。在 `acp.json` 中设置 `"delegate": { "forceEnable": true }` 可在检测到第三方子代理时仍强制保留 `acp_delegate`。
 
 ## 工作原理
 
@@ -119,6 +121,8 @@ billion-context-pi 面向 **Pi** 编码代理(`@earendil-works/pi-coding-agent`)
 
 - **与 [billion-context](https://github.com/ranxianglei/billion-context) 线代理共存** —— 两者同时作用于同一会话会对每个请求双重压缩(token 浪费、嵌套摘要、两套 ref 坐标系)。这会被自动防止:launcher 路径(`bili pi` 等)导出 `BILLION_CONTEXT_PROXY`;模型 `baseUrl` 经代理路由(`…/bili/https://upstream…`)时在会话开始即被检测到 —— 两种情况下 billion-context-pi 都会带警告让位,由代理独占压缩。一个例外:透明模式(流量经 `HTTPS_PROXY` 到达代理、URL 无 `/bili/` 前缀)无法从 URL 识别 —— 此时请在启动 pi 前导出 `BILLION_CONTEXT_PROXY=1`。
 
+- **反向冲突 —— `billion-context` 薄插件抢占 `/acp`。** 如果 [`billion-context`](https://github.com/ranxianglei/billion-context) 内置的 pi 插件也被安装了(`~/.pi/agent/settings.json` 的 `"packages"` 里存在指向某个 `billion-context` 安装目录的条目,例如来自 `bili plugin install pi`),它会注册自己的 `/acp` 命令,而 Pi 没有重复命令防护 —— 哪个生效取决于加载顺序。典型症状:在普通 `pi` 启动下输入 `/acp` 显示 ``bili: no proxy detected (run via `bili <client>` or set a /bili/ baseURL)`` —— 这行字符串由薄插件发出,**不是** billion-context-pi 发出的,而且 pi 并不需要运行 `bili`。修复:该冲突只在你用普通 `pi` 启动时才有实际影响 —— 移除薄插件(`bili plugin remove pi`,或从该设置文件中删除除 `npm:billion-context-pi` 之外的条目),让 `/acp` 由 billion-context-pi 接管;若改用 `bili pi` 启动则无需移除任何东西:launcher 会导出 `BILLION_CONTEXT_PROXY`,billion-context-pi 整体让位(见上一条),两个插件可以并存。注意:`bili plugin install pi` 会静默删除该文件中已有的 `npm:billion-context-pi` 条目且不提示([billion-context#788](https://github.com/ranxianglei/billion-context/issues/788))。
+
 ## 模型工具
 
 | 工具 | 作用 |

package/dist/config.d.ts

@@ -29,6 +29,13 @@ export interface DelegateConfig {
     /** Enable acp_delegate tools (delegate/wait/cancel) and their system-prompt
      *  section. Default: true. Set `enabled: false` to skip registering them. */
     enabled?: boolean;
+    /** Keep acp_delegate active even when a third-party subagent extension
+     *  (pi-subagents) is installed. Default: false — when pi-subagents is
+     *  detected at session start, acp_delegate stands down (tools, fleet
+     *  shortcut and system-prompt section skipped) to avoid two overlapping
+     *  sub-agent systems, and a reminder points at /acp-subagents so the
+     *  third-party agents can still get ACP compression tools (#415). */
+    forceEnable?: boolean;
     /** How delegate usage is reported back to the main session.
      *  "separate" (default) — delegate tokens tracked in a separate accumulator;
      *  main session totals stay clean, delegate usage shows as its own block in
@@ -89,6 +96,9 @@ export interface DelegateConfig {
  *  corresponding timeout/watchdog is disabled. */
 export interface DelegatePolicy {
     enabled: boolean;
+    /** Resolved delegate.forceEnable (default false): keep acp_delegate even
+     *  when a third-party subagent extension (pi-subagents) is installed (#415). */
+    forceEnable: boolean;
     displayUsage: "merged" | "separate";
     maxDepth: number;
     syncTimeoutMs: number | null;

package/dist/index.js

@@ -4532,6 +4532,7 @@ var debug = {
 var DEFAULT_FLEET_SHORTCUT = "ctrl+alt+d";
 var DEFAULT_DELEGATE_POLICY = {
   enabled: true,
+  forceEnable: false,
   displayUsage: "separate",
   maxDepth: 2,
   syncTimeoutMs: 5 * 6e4,
@@ -4565,19 +4566,28 @@ function resolveDelegate(adapter) {
     DEFAULT_DELEGATE_POLICY.asyncTimeoutMs
   );
   const maxConcurrent = resolveMaxConcurrent(process.env.PI_ACP_DELEGATE_MAX_CONCURRENT, cfg.maxConcurrent);
+  const forceEnable = resolveForceEnable(process.env.PI_ACP_DELEGATE_FORCE_ENABLE, cfg.forceEnable);
   if (idleMs === null) {
     logWarn("config", {
       event: "delegate-idle-watchdog-disabled",
       hint: "no-output watchdog is off; hung async runs must be cancelled manually via acp_delegate_cancel"
     });
   }
-  return { enabled, displayUsage, maxDepth, syncTimeoutMs, idleMs, asyncTimeoutMs, maxConcurrent, thinkingLevel: cfg.thinkingLevel, agents: cfg.agents, notifyIfRead: cfg.notifyIfRead ?? "skip", fleetShortcut: resolveFleetShortcut(cfg.fleetShortcut) };
+  return { enabled, forceEnable, displayUsage, maxDepth, syncTimeoutMs, idleMs, asyncTimeoutMs, maxConcurrent, thinkingLevel: cfg.thinkingLevel, agents: cfg.agents, notifyIfRead: cfg.notifyIfRead ?? "skip", fleetShortcut: resolveFleetShortcut(cfg.fleetShortcut) };
 }
 function resolveFleetShortcut(value) {
   if (typeof value === "string") return value;
   if (value !== void 0) logWarn("config", { event: "delegate-config-invalid", field: "fleetShortcut", value: String(value), fallback: DEFAULT_FLEET_SHORTCUT });
   return DEFAULT_FLEET_SHORTCUT;
 }
+function resolveForceEnable(envValue, cfgValue) {
+  if (envValue === "true") return true;
+  if (envValue === "false") return false;
+  if (envValue !== void 0) {
+    logWarn("config", { event: "delegate-config-invalid", field: "forceEnable", value: envValue, fallback: cfgValue === true });
+  }
+  return cfgValue === true;
+}
 function resolveMaxDepth(value) {
   if (value === void 0) return DEFAULT_DELEGATE_POLICY.maxDepth;
   const n = Number(value);
@@ -6016,6 +6026,25 @@ var OverflowEpisode = class {
   }
 };
 
+// src/turn-boundary.ts
+function isTurnBoundary(entry, policy = {}) {
+  if (entry.message?.role === "user") return true;
+  return policy.countCustomMessages === true && isCustomMessageEntry(entry);
+}
+function lastTurnBoundaryId(entries, policy) {
+  for (let i = entries.length - 1; i >= 0; i--) {
+    const e = entries[i];
+    if (isTurnBoundary(e, policy)) return e.id;
+  }
+  return void 0;
+}
+function lastTurnBoundaryIndex(entries, policy) {
+  for (let i = entries.length - 1; i >= 0; i--) {
+    if (isTurnBoundary(entries[i], policy)) return i;
+  }
+  return -1;
+}
+
 // src/runtime.ts
 function readContextEntries(sm) {
   const source = sm;
@@ -6027,6 +6056,9 @@ function isPiHost(sm) {
   const source = sm;
   return typeof source.buildContextEntries === "function";
 }
+function retryBreakerKey(sm, policy) {
+  return lastTurnBoundaryId(readContextEntries(sm), policy);
+}
 function mergeLiveEntries(entries, live, state, origins) {
   const persisted = entries.filter((e) => e.type === "message");
   const liveIdentities = live.map(messageIdentity);
@@ -6230,6 +6262,40 @@ function createRuntime(adapter) {
   function dropTokenScale(sid) {
     tokenScaleStale.delete(sid);
   }
+  const HOST_USAGE_WINDOW = 4;
+  const HOST_USAGE_MIN_SAMPLES = 3;
+  const HOST_USAGE_SPREAD_MAX = 0.25;
+  const hostUsageSamples = /* @__PURE__ */ new Map();
+  function noteHostUsage(sid, tokens) {
+    if (tokens <= 0) return false;
+    let win = hostUsageSamples.get(sid);
+    if (!win) {
+      win = [];
+      hostUsageSamples.set(sid, win);
+    }
+    win.push(tokens);
+    if (win.length > HOST_USAGE_WINDOW) win.shift();
+    if (win.length < HOST_USAGE_MIN_SAMPLES) return false;
+    const lo = Math.min(...win);
+    const hi = Math.max(...win);
+    return (hi - lo) / lo <= HOST_USAGE_SPREAD_MAX;
+  }
+  function dropHostUsageSamples(sid) {
+    hostUsageSamples.delete(sid);
+  }
+  const sizeDivergenceStreaks = /* @__PURE__ */ new Map();
+  function noteSizeDivergence(sid, divergent) {
+    if (!divergent) {
+      sizeDivergenceStreaks.delete(sid);
+      return false;
+    }
+    const streak = (sizeDivergenceStreaks.get(sid) ?? 0) + 1;
+    sizeDivergenceStreaks.set(sid, streak);
+    return streak === 3;
+  }
+  function dropSizeDivergence(sid) {
+    sizeDivergenceStreaks.delete(sid);
+  }
   const strictEchoLogged = /* @__PURE__ */ new Set();
   const compressOutcomes = /* @__PURE__ */ new Map();
   function compressTrackerFor(sid) {
@@ -6379,6 +6445,7 @@ function createRuntime(adapter) {
   }
   let refused = false;
   let refusalMessage = null;
+  let delegateStoodDown = false;
   return { core, store, get refused() {
     return refused;
   }, set refused(v2) {
@@ -6387,6 +6454,10 @@ function createRuntime(adapter) {
     return refusalMessage;
   }, set refusalMessage(v2) {
     refusalMessage = v2;
+  }, get delegateStoodDown() {
+    return delegateStoodDown;
+  }, set delegateStoodDown(v2) {
+    delegateStoodDown = v2;
   }, get adapter() {
     return adapterRef;
   }, setAdapter: (a) => {
@@ -6395,7 +6466,7 @@ function createRuntime(adapter) {
     return promptsRef;
   }, setPrompts: (p) => {
     promptsRef = p;
-  }, markNudgeShown, nudgeShownFor, nudgeShownTokensFor, clearNudgeTracking, clearNudgeTokenStamps, noteCompressOutcomes, compressRetryCappedFor, clearCompressRetryTracking, liveContextLimit, configFor, reasoningDropFor, reloadConfig, stateFor, save, deriveChildState: deriveChild, acquireLock, overflowFor, overflowDrop, noteDeadCompress, clearDeadCompress, throttleFor, throttleDrop, noteTokenScale, dropTokenScale };
+  }, markNudgeShown, nudgeShownFor, nudgeShownTokensFor, clearNudgeTracking, clearNudgeTokenStamps, noteCompressOutcomes, compressRetryCappedFor, clearCompressRetryTracking, liveContextLimit, configFor, reasoningDropFor, reloadConfig, stateFor, save, deriveChildState: deriveChild, acquireLock, overflowFor, overflowDrop, noteDeadCompress, clearDeadCompress, throttleFor, throttleDrop, noteTokenScale, dropTokenScale, noteHostUsage, dropHostUsageSamples, noteSizeDivergence, dropSizeDivergence };
 }
 
 // node_modules/typebox/build/system/memory/memory.mjs
@@ -10792,6 +10863,15 @@ __export(typebox_exports, {
   With: () => With2
 });
 
+// src/host.ts
+function isDeclaredForkHost() {
+  const v2 = process.env.PI_ACP_FORK_HOST;
+  return v2 === "1" || v2?.toLowerCase() === "true";
+}
+function isUnsupportedHost(sm) {
+  return !isPiHost(sm) && !isDeclaredForkHost();
+}
+
 // src/tokens.ts
 function collectCoveredMessageIds(state) {
   const ids = /* @__PURE__ */ new Set();
@@ -10840,25 +10920,6 @@ function adjustedTokenCount(core, messages, state, config, prelim, imageTokensBy
   return view.drifted ? view.viewTokens : prelim;
 }
 
-// src/turn-boundary.ts
-function isTurnBoundary(entry, policy = {}) {
-  if (entry.message?.role === "user") return true;
-  return policy.countCustomMessages === true && isCustomMessageEntry(entry);
-}
-function lastTurnBoundaryId(entries, policy) {
-  for (let i = entries.length - 1; i >= 0; i--) {
-    const e = entries[i];
-    if (isTurnBoundary(e, policy)) return e.id;
-  }
-  return void 0;
-}
-function lastTurnBoundaryIndex(entries, policy) {
-  for (let i = entries.length - 1; i >= 0; i--) {
-    if (isTurnBoundary(entries[i], policy)) return i;
-  }
-  return -1;
-}
-
 // src/compat.ts
 function normalizeSystemPrompt(input) {
   if (input === void 0) return "";
@@ -10887,6 +10948,12 @@ var UNSUPPORTED_HOST_MESSAGE = [
   "  bili omp",
   "Docs: https://github.com/ranxianglei/billion-context"
 ].join("\n");
+var FORK_HOST_WARNING_MESSAGE = [
+  '[billion-context-pi] Running on a declared fork host (PI_ACP_FORK_HOST). Known limitation: live-tail refs can drift as the session grows, so compress calls may start failing with "does not exist in this session" (#454). Root fix tracked in #459.',
+  "For long sessions on this host, prefer the billion-context proxy \u2014 it runs compression server-side and owns the ref coordinate space:",
+  "  npm install -g billion-context",
+  "  bili omp"
+].join("\n");
 
 // src/compress-tool.ts
 function formatK2(n) {
@@ -11158,7 +11225,7 @@ async function handleCompress(args, runtime, ctx, toolCallId) {
     }
     return s.text === r.summary ? r : { ...r, summary: s.text };
   });
-  const turnKey = lastTurnBoundaryId(entries, resolveHostSession(runtime.adapter)) ?? sid;
+  const turnKey = retryBreakerKey(ctx.sessionManager, resolveHostSession(runtime.adapter)) ?? sid;
   const snapshot = compressibleSnapshotText(turn.nudge);
   if (runtime.compressRetryCappedFor(sid, turnKey)) {
     logWarn("compress", { sid, event: "capped-reject", turnKey });
@@ -11247,6 +11314,9 @@ async function handleCompress(args, runtime, ctx, toolCallId) {
   });
   if (errors.length > 0) {
     logError("compress", { sid: ctx.sessionManager.getSessionId(), event: "errors", count: errors.length, errors: errors.slice(0, 5) });
+    if (blocksCreated === 0 && isDeclaredForkHost() && !isPiHost(ctx.sessionManager) && /does not exist|cannot be anchored|is unknown|unknown refs/i.test(errors.join(" "))) {
+      logWarn("compress", { sid: ctx.sessionManager.getSessionId(), event: "fork-ref-drift-suspected", seeIssue: "#459", errors: errors.slice(0, 3) });
+    }
   }
   if (warnings.length > 0) {
     logWarn("compress", { sid: ctx.sessionManager.getSessionId(), event: "warnings", count: warnings.length, warnings: warnings.slice(0, 5) });
@@ -11602,7 +11672,7 @@ function resolveActivePack(adapter, cwd, provider, modelId, resolver) {
 }
 var HOST_VERSION = (() => {
   try {
-    return "0.1.70";
+    return "0.1.71";
   } catch {
     return "dev";
   }
@@ -17897,6 +17967,11 @@ import * as fs6 from "fs";
 import * as os2 from "os";
 import * as path11 from "path";
 var ACP_TOOLS3 = ["compress", "decompress", "search_context", "acp_status"];
+var DELEGATE_STAND_DOWN_MESSAGE = [
+  "[billion-context-pi] pi-subagents detected \u2014 acp_delegate has been automatically disabled to avoid two overlapping sub-agent systems.",
+  "pi-subagents' agents do NOT get ACP context compression by default. Run /acp-subagents to inject compress/decompress/search_context/acp_status into its agent overrides.",
+  'To keep acp_delegate despite pi-subagents being installed, set "delegate": { "forceEnable": true } in acp.json.'
+].join("\n");
 function resolveAgentDir() {
   const envDir = process.env.PI_CODING_AGENT_DIR;
   if (envDir) {
@@ -17927,38 +18002,39 @@ function parseFrontmatterTools(content) {
   if (!name) return null;
   return tools ? { name, tools } : { name };
 }
-function findPiSubagentsInstall(agentDir, cwd) {
-  const candidates = [
-    path11.join(agentDir, "npm", "node_modules", "pi-subagents"),
-    path11.join(cwd, CONFIG_DIR_NAME2, "npm", "node_modules", "pi-subagents")
-  ];
-  const extensionRoots = [
-    path11.join(agentDir, "extensions"),
-    path11.join(cwd, CONFIG_DIR_NAME2, "extensions")
-  ];
-  for (const dir of candidates) {
-    if (fs6.existsSync(path11.join(dir, "package.json"))) return dir;
+function npmInstallAt(base) {
+  const dir = path11.join(base, "npm", "node_modules", "pi-subagents");
+  return fs6.existsSync(path11.join(dir, "package.json")) ? dir : null;
+}
+function extensionInstallUnder(root) {
+  let entries;
+  try {
+    entries = fs6.readdirSync(root, { withFileTypes: true });
+  } catch {
+    return null;
   }
-  for (const root of extensionRoots) {
-    let entries;
+  for (const entry of entries) {
+    if (!entry.isDirectory()) continue;
+    const pkgPath = path11.join(root, entry.name, "package.json");
     try {
-      entries = fs6.readdirSync(root, { withFileTypes: true });
-    } catch {
-      continue;
-    }
-    for (const entry of entries) {
-      if (!entry.isDirectory()) continue;
-      const pkgPath = path11.join(root, entry.name, "package.json");
-      try {
-        if (JSON.parse(fs6.readFileSync(pkgPath, "utf-8")).name === "pi-subagents") {
-          return path11.join(root, entry.name);
-        }
-      } catch {
+      if (JSON.parse(fs6.readFileSync(pkgPath, "utf-8")).name === "pi-subagents") {
+        return path11.join(root, entry.name);
       }
+    } catch {
     }
   }
   return null;
 }
+function findPiSubagentsInstall(agentDir, cwd) {
+  return npmInstallAt(agentDir) ?? npmInstallAt(path11.join(cwd, CONFIG_DIR_NAME2)) ?? extensionInstallUnder(path11.join(agentDir, "extensions")) ?? extensionInstallUnder(path11.join(cwd, CONFIG_DIR_NAME2, "extensions"));
+}
+function findPiSubagentsInstalls(agentDir, cwd) {
+  const pick = (paths) => paths.filter((p) => p !== null);
+  return {
+    user: pick([npmInstallAt(agentDir), extensionInstallUnder(path11.join(agentDir, "extensions"))]),
+    project: pick([npmInstallAt(path11.join(cwd, CONFIG_DIR_NAME2)), extensionInstallUnder(path11.join(cwd, CONFIG_DIR_NAME2, "extensions"))])
+  };
+}
 function discoverBuiltinAgents(installDir) {
   const agentsDir = path11.join(installDir, "agents");
   let entries;
@@ -18256,7 +18332,7 @@ async function statusReport(runtime, ctx) {
   const sentTokens = estimateTokens(coreMessages, coveredIds, imageTokens) + systemPromptTokens;
   const viewSentTokens = adjustedTokenCount(runtime.core, coreMessages, state, config, sentTokens, imageTokens, systemPromptTokens);
   const turn = runtime.core.processTurn({ messages: coreMessages, state, config, tokenCount: anchorStale ? viewSentTokens : Math.max(viewSentTokens, realUsage?.tokens ?? 0) });
-  const versionStr = "0.1.70" ? `billion-context-pi@${"0.1.70"} \xB7 pack: ${resolveSurfaceMeta(runtime.adapter, ctx?.cwd ?? process.cwd(), ctx?.model?.provider, ctx?.model?.id).pack}` : void 0;
+  const versionStr = "0.1.71" ? `billion-context-pi@${"0.1.71"} \xB7 pack: ${resolveSurfaceMeta(runtime.adapter, ctx?.cwd ?? process.cwd(), ctx?.model?.provider, ctx?.model?.id).pack}` : void 0;
   let text = buildStatusPanel({
     version: versionStr,
     tokenCount: sessionTokens,
@@ -18868,7 +18944,7 @@ async function autoInstallLatest(latest, extDirOverride) {
       "--no-fund",
       "--no-save"
     ];
-    const prevVersion = (await readPackageJson(join15(extDir, "package.json")))?.version ?? "0.1.70";
+    const prevVersion = (await readPackageJson(join15(extDir, "package.json")))?.version ?? "0.1.71";
     const { code, stderr } = await runNpmImpl(installArgs(latest), { cwd: npmDir, timeout: 6e4 });
     if (code !== 0) {
       logWarn("update", {
@@ -18886,7 +18962,7 @@ async function autoInstallLatest(latest, extDirOverride) {
     }
     const verify = await verifyInstall(npmDir, latest);
     if (!verify.ok) {
-      const rollbackTo = SEMVER_RE.test(prevVersion) ? prevVersion : "0.1.70";
+      const rollbackTo = SEMVER_RE.test(prevVersion) ? prevVersion : "0.1.71";
       logWarn("update", { event: "auto-install-verify-failed", latest, reason: verify.reason, rollbackTo });
       const rb = await runNpmImpl(installArgs(rollbackTo), { cwd: npmDir, timeout: 6e4 });
       logInfo("update", { event: "rollback", from: latest, to: rollbackTo, ok: rb.code === 0 });
@@ -18961,7 +19037,7 @@ async function checkForUpdate(autoUpdate, notify) {
     }
     const latest = await fetchLatestVersion(tag);
     if (!latest) return;
-    const current = runtimeVersion ?? "0.1.70";
+    const current = runtimeVersion ?? "0.1.71";
     const hasUpdate = isVersionNewer(latest, current);
     debug.event("update-check", {
       current,
@@ -19019,15 +19095,6 @@ async function getInstalledSpec() {
   return hostPkg?.dependencies?.[PACKAGE_NAME];
 }
 
-// src/host.ts
-function isDeclaredForkHost() {
-  const v2 = process.env.PI_ACP_FORK_HOST;
-  return v2 === "1" || v2?.toLowerCase() === "true";
-}
-function isUnsupportedHost(sm) {
-  return !isPiHost(sm) && !isDeclaredForkHost();
-}
-
 // src/proxy-detect.ts
 function isBiliProxyBaseUrl(baseUrl) {
   if (!baseUrl) return false;
@@ -19048,6 +19115,13 @@ var PROXY_STAND_DOWN_MESSAGE = [
   "If your traffic reaches the proxy via HTTPS_PROXY instead (no /bili/ prefix in the URL), export BILLION_CONTEXT_PROXY=1 before starting pi.",
   "Docs: https://github.com/ranxianglei/billion-context"
 ].join("\n");
+function nativeStandDownMessage(host) {
+  return [
+    `[billion-context-pi] billion-context native host detected (BILLION_CONTEXT_NATIVE=${host}) \u2014 the in-process native plugin proxies model API traffic and owns compression.`,
+    "ACP client-side compression has been disabled for this session to avoid double compression; nothing else to configure.",
+    "Docs: https://github.com/ranxianglei/billion-context"
+  ].join("\n");
+}
 
 // src/index.ts
 function createAcpExtension(adapter = {}) {
@@ -19061,16 +19135,18 @@ function createAcpExtension(adapter = {}) {
       return;
     }
     const runtime = createRuntime(adapter);
-    let proxyWarned = false;
+    let standDownWarned = false;
     const standDownIfProxied = (ctx) => {
-      if (!isBiliProxyBaseUrl(ctx.model?.baseUrl)) return false;
+      const nativeHost = process.env.BILLION_CONTEXT_NATIVE || void 0;
+      if (nativeHost === void 0 && !process.env.BILLION_CONTEXT_PROXY && !isBiliProxyBaseUrl(ctx.model?.baseUrl)) return false;
+      const message = nativeHost !== void 0 ? nativeStandDownMessage(nativeHost) : PROXY_STAND_DOWN_MESSAGE;
       runtime.refused = true;
-      runtime.refusalMessage = PROXY_STAND_DOWN_MESSAGE;
-      if (!proxyWarned) {
-        proxyWarned = true;
-        logWarn("host", { event: "proxy-baseurl-detected", sid: ctx.sessionManager.getSessionId(), action: "refused" });
-        if (ctx.hasUI) ctx.ui.notify(PROXY_STAND_DOWN_MESSAGE, "warning");
-        else console.error(PROXY_STAND_DOWN_MESSAGE);
+      runtime.refusalMessage = message;
+      if (!standDownWarned) {
+        standDownWarned = true;
+        logWarn("host", { event: nativeHost !== void 0 ? "native-host-detected" : "proxy-detected", sid: ctx.sessionManager.getSessionId(), action: "refused", native: nativeHost ?? null });
+        if (ctx.hasUI) ctx.ui.notify(message, "warning");
+        else console.error(message);
       }
       return true;
     };
@@ -19127,6 +19203,8 @@ function wireDelegateReadTracking(pi) {
 }
 function wireSessionLifecycle(pi, runtime, standDownIfProxied) {
   let ompWarned = false;
+  let forkWarned = false;
+  let subagentStandDownWarned = false;
   pi.on("session_start", async (_event, ctx) => {
     if (isUnsupportedHost(ctx.sessionManager)) {
       runtime.refused = true;
@@ -19139,17 +19217,29 @@ function wireSessionLifecycle(pi, runtime, standDownIfProxied) {
       }
       return;
     }
+    if (isDeclaredForkHost() && !isPiHost(ctx.sessionManager)) {
+      const sid2 = ctx.sessionManager.getSessionId();
+      logWarn("host", { event: "fork-host-admitted", sid: sid2, knownLimitation: "ref-drift", seeIssue: "#454", rootFix: "#459" });
+      if (!forkWarned) {
+        forkWarned = true;
+        if (ctx.hasUI) ctx.ui.notify(FORK_HOST_WARNING_MESSAGE, "warning");
+        else console.error(FORK_HOST_WARNING_MESSAGE);
+      }
+    }
     if (standDownIfProxied(ctx)) return;
     runtime.store.invalidate();
     runtime.clearNudgeTracking(ctx.sessionManager.getSessionId());
     runtime.throttleFor(ctx.sessionManager.getSessionId()).reset();
     runtime.clearCompressRetryTracking(ctx.sessionManager.getSessionId());
+    runtime.dropHostUsageSamples(ctx.sessionManager.getSessionId());
+    runtime.dropSizeDivergence(ctx.sessionManager.getSessionId());
     resetDelegateUsage();
     setDelegateDisplayUsage("separate");
     setDelegatePolicy(DEFAULT_DELEGATE_POLICY);
     const sid = ctx.sessionManager.getSessionId();
     const modelInfo = ctx.model;
-    logInfo("session", { event: "start", sid, cwd: ctx.cwd, debug: runtime.adapter.debug ?? null, version: true ? "0.1.70" : null, model: modelInfo?.id ?? null, modelApi: modelInfo?.api ?? null, contextWindow: modelInfo?.contextWindow ?? null });
+    logInfo("session", { event: "start", sid, cwd: ctx.cwd, debug: runtime.adapter.debug ?? null, version: true ? "0.1.71" : null, model: modelInfo?.id ?? null, modelApi: modelInfo?.api ?? null, contextWindow: modelInfo?.contextWindow ?? null });
+    let delegateStoodDown = false;
     try {
       await runtime.reloadConfig(ctx.cwd);
       const delegateCfg = resolveDelegate(runtime.adapter);
@@ -19157,9 +19247,24 @@ function wireSessionLifecycle(pi, runtime, standDownIfProxied) {
       setDelegatePolicy(delegateCfg);
       setDelegateDefaults({ thinkingLevel: delegateCfg.thinkingLevel, agents: delegateCfg.agents });
       setDelegateNotifyIfRead(delegateCfg.notifyIfRead);
+      if (delegateCfg.enabled && !delegateCfg.forceEnable) {
+        const scopes = findPiSubagentsInstalls(resolveAgentDir(), ctx.cwd ?? process.cwd());
+        if (scopes.project[0] !== void 0) {
+          delegateStoodDown = true;
+          logWarn("delegate", { event: "delegate-auto-disabled", sid, install: scopes.project[0], scope: "project", hint: "run /acp-subagents to give its agents ACP compression tools; delegate.forceEnable=true keeps acp_delegate" });
+          if (!subagentStandDownWarned) {
+            subagentStandDownWarned = true;
+            if (ctx.hasUI) ctx.ui.notify(DELEGATE_STAND_DOWN_MESSAGE, "warning");
+            else console.error(DELEGATE_STAND_DOWN_MESSAGE);
+          }
+        } else if (scopes.user[0] !== void 0) {
+          logWarn("delegate", { event: "delegate-user-scope-detected", sid, install: scopes.user[0], action: "warn-only", hint: "user-level pi-subagents does not disable acp_delegate; run /acp-subagents to give its agents ACP compression tools" });
+        }
+      }
     } catch (e) {
       logThrow("config", e, { sid, phase: "session_start" });
     }
+    runtime.delegateStoodDown = delegateStoodDown;
     try {
       runtime.setPrompts(resolvePrompts(runtime.adapter.prompts, { acknowledgeRisk: runtime.adapter.acknowledgePromptsRisk === true }));
     } catch (e) {
@@ -19167,7 +19272,7 @@ function wireSessionLifecycle(pi, runtime, standDownIfProxied) {
       runtime.setPrompts(defaultPrompts);
     }
     const delegatePolicy2 = resolveDelegate(runtime.adapter);
-    if (delegatePolicy2.enabled) {
+    if (delegatePolicy2.enabled && !runtime.delegateStoodDown) {
       pi.registerTool(makeDelegateTool(pi));
       pi.registerTool(makeDelegateWaitTool(pi));
       pi.registerTool(makeDelegateCancelTool(pi));
@@ -19192,6 +19297,8 @@ function wireSessionLifecycle(pi, runtime, standDownIfProxied) {
     runtime.dropTokenScale(sid);
     runtime.clearNudgeTracking(sid);
     runtime.clearCompressRetryTracking(sid);
+    runtime.dropHostUsageSamples(sid);
+    runtime.dropSizeDivergence(sid);
     delegateStatusWidget.dispose();
     closeLogStream();
   });
@@ -19235,7 +19342,9 @@ function wireContextTransform(pi, runtime, standDownIfProxied) {
       }
       const hostFloorActive = !usageAnchorPredatesCompression(entries);
       const realPromptTokens = realUsage?.tokens ?? 0;
-      const applyFloors = (base) => Math.max(base, hostFloorActive ? realPromptTokens : 0, armedFloor);
+      const hostUsageStable = hostFloorActive && realPromptTokens > 0 ? runtime.noteHostUsage(sid, realPromptTokens) : false;
+      const calibrate = (base) => hostUsageStable ? Math.min(base, Math.ceil(realPromptTokens * 1.2)) : base;
+      const applyFloors = (base) => Math.max(calibrate(base), hostFloorActive ? realPromptTokens : 0, armedFloor);
       let tokenCount = applyFloors(sentTokens);
       if (state.blocks.some((b2) => b2.active && b2.effectiveMessageIds.length > 0)) {
         const view = sentViewTokenCount(runtime.core, coreMessages, state, config, tokenCount, imageTokens, systemPromptTokens);
@@ -19244,12 +19353,20 @@ function wireContextTransform(pi, runtime, standDownIfProxied) {
           logInfo("turn", { sid, event: "view-recount", prelim: sentTokens, viewTokens: view.viewTokens, tokenCount });
         }
       }
+      const sizeDivergent = hostFloorActive && realPromptTokens > 0 && Math.abs(tokenCount - realPromptTokens) / realPromptTokens > 0.5;
+      if (runtime.noteSizeDivergence(sid, sizeDivergent)) {
+        logWarn("turn", { sid, event: "size-divergence", est: tokenCount, host: realPromptTokens, ratio: Number((tokenCount / realPromptTokens).toFixed(2)), stable: hostUsageStable });
+      }
       if (runtime.noteTokenScale(sid, !hostFloorActive)) {
-        state.nudge.lastNudgeShownTokens = 0;
-        state.nudge.lastPerMessageNudgeTokens = 0;
-        state.nudge.lastShownByTier = {};
+        state.nudge.lastNudgeShownTokens = state.nudge.lastNudgeShownTokens > 0 ? tokenCount : 0;
+        state.nudge.lastPerMessageNudgeTokens = state.nudge.lastPerMessageNudgeTokens > 0 ? tokenCount : 0;
+        const reanchored = {};
+        for (const [tier, shown] of Object.entries(state.nudge.lastShownByTier)) {
+          if (shown > 0) reanchored[Number(tier)] = tokenCount;
+        }
+        state.nudge.lastShownByTier = reanchored;
         runtime.clearNudgeTokenStamps(sid);
-        logInfo("growth-scale", { sid, event: "scale-flip-reset", anchorStale: !hostFloorActive });
+        logInfo("growth-scale", { sid, event: "scale-flip-reanchor", anchorStale: !hostFloorActive, tokenCount });
       }
       debug.event("context-in", {
         sid,
@@ -19330,8 +19447,9 @@ function wireContextTransform(pi, runtime, standDownIfProxied) {
       const debugOn2 = debug.enabled;
       const turnPolicy = resolveHostSession(runtime.adapter);
       const turnKey = lastTurnBoundaryId(entries, turnPolicy) ?? sid;
+      const retryTurnKey = retryBreakerKey(ctx.sessionManager, turnPolicy) ?? sid;
       const compressOutcomes = collectCompressOutcomes(entries, lastTurnBoundaryIndex(entries, turnPolicy));
-      const outcome = compressOutcomes.length > 0 ? runtime.noteCompressOutcomes(sid, turnKey, compressOutcomes) : null;
+      const outcome = compressOutcomes.length > 0 ? runtime.noteCompressOutcomes(sid, retryTurnKey, compressOutcomes) : null;
       const adaptiveGrowth = !config.modelContextLimit || config.modelContextLimit <= 0 ? config.nudge.growthFloor : Math.min(
         config.nudge.growthCap,
         Math.max(config.nudge.growthFloor, Math.round(config.modelContextLimit * config.nudge.growthRatio))
@@ -19346,7 +19464,7 @@ function wireContextTransform(pi, runtime, standDownIfProxied) {
       if (turn.nudge?.shouldInject) {
         const emergency = turn.nudge.breakdown?.emergencyOverride === 1;
         turn.nudge.compressibleRanges = viableRanges(turn.nudge.compressibleRanges);
-        const retryCapped = runtime.compressRetryCappedFor(sid, turnKey);
+        const retryCapped = runtime.compressRetryCappedFor(sid, retryTurnKey);
         const reInjectReady = shownAt === void 0 || tokenCount - shownAt >= reInjectFloor;
         const alreadyShown = retryCapped || !emergency && runtime.nudgeShownFor(sid, turnKey) && !reInjectReady;
         if (!alreadyShown) {
@@ -19414,7 +19532,7 @@ function wireSystemPrompt(pi, runtime) {
       }
       runtime.setPrompts(defaultPrompts);
     }
-    const delegate = resolveDelegate(runtime.adapter).enabled;
+    const delegate = resolveDelegate(runtime.adapter).enabled && !runtime.delegateStoodDown;
     const acp = buildAcpSystemPrompt(runtime.prompts, merged.promptSections);
     const delegateText = merged.delegatePrompt !== void 0 ? merged.delegatePrompt : ACP_DELEGATE_PROMPT;
     const prompt = delegate && delegateText !== null ? `${acp}