@livx.cc/agentx 0.96.15 → 0.96.17

package/dist/{Agent-DRe91tAy.d.ts → Agent-DdhD1pGw.d.ts}

@@ -240,6 +240,20 @@ declare class AgentOptions {
         tool: string;
         args: any;
     }) => string | Promise<string>;
+    /** Browser-run adaptive trimming. Browser MCP tools (`mcp__*browser*`) return huge per-call results
+     *  (full DOM text + screenshots) that accumulate per step and blow the cumulative `maxTokens` budget
+     *  kill-switch mid-browse. The fix attacks PER-STEP TOKEN GROWTH, and engages ONLY once a browser tool
+     *  is ACTUALLY invoked (not on mere config presence — so normal coding is untouched):
+     *   - `resultBytes`: tighter `maxToolResultBytes` applied to BROWSER tool results (each oversized DOM is
+     *     cropped/spilled to a small stub → small fresh-token tail per step).
+     *   - `keepOutputs`: keep only the most-recent N tool-result bodies verbatim ONCE a browser tool has run
+     *     (older huge DOMs collapse to one-line stubs → small cached prefix → small 0.1×cacheRead/step).
+     *  Explicit `keepToolOutputs`/`maxToolResultBytes` are NOT overridden below their browser values — the
+     *  tighter of (user, browser) wins, never loosening a user's cap. Unset = off (no adaptive behavior). */
+    browserTrim?: {
+        resultBytes?: number;
+        keepOutputs?: number;
+    };
     /** VFS dir(s) of skills (`<dir>/<id>/SKILL.md`). If set: inject a catalog + add the `Skill` tool. Multiple dirs are merged (first wins on name collisions). */
     skillsDir?: string | string[];
     /** VFS dir(s) of slash-command templates (`<dir>/<name>.md`). If set: inject a catalog + add the `SlashCommand` tool. Multiple dirs are merged (first wins). */
@@ -326,6 +340,7 @@ declare class Agent {
     private systemPromptCache;
     private started;
     private parkedMs;
+    private browserActive;
     /** Time a human-blocking await (a permission/plan prompt) and bank it in `parkedMs` so idle prompt
      *  time never trips the wall-clock kill-switch. The agent did no work while parked on the user. */
     private park;

package/dist/cli.d.ts

@@ -1,5 +1,5 @@
 #!/usr/bin/env bun
-import { H as Hooks, h as RunResult, R as ReasoningEffort, A as Agent } from './Agent-DRe91tAy.js';
+import { H as Hooks, h as RunResult, R as ReasoningEffort, A as Agent } from './Agent-DdhD1pGw.js';
 import { IFilesystem } from '@livx.cc/wcli/core';
 import { M as Message, c as ContentPart, e as MessageContent } from './tools-DtpN8Agv.js';
 

package/dist/cli.js

@@ -3141,6 +3141,9 @@ function reasoningToChatFragment(model, effort) {
 
 // src/Agent.ts
 var log4 = forComponent("Agent");
+function isBrowserTool(name) {
+  return name.startsWith("mcp__") && /browser/i.test(name.slice(5).split("__")[0]);
+}
 function isAbortError(err2) {
   const e = err2;
   const blob = `${e?.message ?? ""} ${e?.name ?? ""} ${e?.code ?? ""} ${e?.cause?.name ?? ""}`;
@@ -3188,6 +3191,17 @@ var AgentOptions = class {
    *  and returns the (cropped) string to put in context — e.g. spill to scratch and return a recoverable,
    *  paginated stub. Called only when a result exceeds `maxToolResultBytes`. */
   capToolResult;
+  /** Browser-run adaptive trimming. Browser MCP tools (`mcp__*browser*`) return huge per-call results
+   *  (full DOM text + screenshots) that accumulate per step and blow the cumulative `maxTokens` budget
+   *  kill-switch mid-browse. The fix attacks PER-STEP TOKEN GROWTH, and engages ONLY once a browser tool
+   *  is ACTUALLY invoked (not on mere config presence — so normal coding is untouched):
+   *   - `resultBytes`: tighter `maxToolResultBytes` applied to BROWSER tool results (each oversized DOM is
+   *     cropped/spilled to a small stub → small fresh-token tail per step).
+   *   - `keepOutputs`: keep only the most-recent N tool-result bodies verbatim ONCE a browser tool has run
+   *     (older huge DOMs collapse to one-line stubs → small cached prefix → small 0.1×cacheRead/step).
+   *  Explicit `keepToolOutputs`/`maxToolResultBytes` are NOT overridden below their browser values — the
+   *  tighter of (user, browser) wins, never loosening a user's cap. Unset = off (no adaptive behavior). */
+  browserTrim;
   /** VFS dir(s) of skills (`<dir>/<id>/SKILL.md`). If set: inject a catalog + add the `Skill` tool. Multiple dirs are merged (first wins on name collisions). */
   skillsDir;
   /** VFS dir(s) of slash-command templates (`<dir>/<name>.md`). If set: inject a catalog + add the `SlashCommand` tool. Multiple dirs are merged (first wins). */
@@ -3272,6 +3286,8 @@ var Agent = class _Agent {
   // session-start lifecycle hook fires once per conversation
   parkedMs = 0;
   // cumulative time blocked on the HUMAN (permission/plan prompts) — excluded from the timeout
+  browserActive = false;
+  // flips true once a browser MCP tool is actually invoked → adaptive trimming engages (browserTrim)
   /** Time a human-blocking await (a permission/plan prompt) and bank it in `parkedMs` so idle prompt
    *  time never trips the wall-clock kill-switch. The agent did no work while parked on the user. */
   async park(p) {
@@ -3680,7 +3696,11 @@ var Agent = class _Agent {
     }
     if (!threw) result = await this.maybeAutoTest(tc.function.name, result);
     if (images?.length && !result) result = `[${images.length} image${images.length > 1 ? "s" : ""} attached]`;
-    const cap = this.options.maxToolResultBytes ?? 0;
+    const browser = isBrowserTool(tc.function.name);
+    if (browser && this.options.browserTrim) this.browserActive = true;
+    const browserBytes = browser ? this.options.browserTrim?.resultBytes : void 0;
+    const baseCap = this.options.maxToolResultBytes ?? 0;
+    const cap = browserBytes ? baseCap > 0 ? Math.min(baseCap, browserBytes) : browserBytes : baseCap;
     if (!threw && cap > 0 && result.length > cap) {
       const info = { tool: tc.function.name, args };
       result = this.options.capToolResult ? await this.options.capToolResult(result, info) : cropResult(result, cap);
@@ -3724,7 +3744,9 @@ ${out}`;
     let out = null;
     if (o.compaction?.maxMessages && m.length > o.compaction.maxMessages) out = compact(m, o.compaction.maxMessages);
     else if (o.maxContextMessages && m.length > o.maxContextMessages) out = dropOldest(m, o.maxContextMessages);
-    if (o.keepToolOutputs) out = stubOldToolResults(out ?? m, o.keepToolOutputs);
+    const browserKeep = this.browserActive ? o.browserTrim?.keepOutputs : void 0;
+    const keep = browserKeep != null ? o.keepToolOutputs ? Math.min(o.keepToolOutputs, browserKeep) : browserKeep : o.keepToolOutputs;
+    if (keep) out = stubOldToolResults(out ?? m, keep);
     if (o.maxContextTokens) {
       const pre = (out ?? m).length;
       out = fitTokenBudget(out ?? m, o.maxContextTokens);
@@ -5918,6 +5940,9 @@ var SonioxSTTOptions = class {
   /** Client-side endpoint: finalized text + no new tokens for this long = utterance (don't wait for
    *  Soniox's semantic <end>, which adds 0.5-1.5s — the difference between ping-pong and lag). */
   silenceEndpointMs = 500;
+  /** No-audio watchdog: if the mic source stops delivering chunks for this long, capture is dead →
+   *  fire onFatal + stop (else Soniox idle-timeouts and reconnect-loops forever). 0 = disable. */
+  noAudioTimeoutMs = 1e4;
 };
 var SonioxSTT = class {
   options;
@@ -5931,6 +5956,15 @@ var SonioxSTT = class {
   /** mic energy (RMS) per chunk — drives the energy-based heuristic barge-in tier */
   onLevel = () => {
   };
+  /** Unrecoverable: the mic source stopped delivering audio (Soniox starves → idle-timeout reconnect
+   *  loop). The host tears voice down instead of spinning forever. */
+  onFatal = () => {
+  };
+  lastChunkAt = 0;
+  // timestamp of the most recent mic chunk (0 = none yet)
+  startedChunksAt = 0;
+  // when capture started (grace before the first chunk)
+  noAudioTimer = null;
   finalText = "";
   partialText = "";
   lastChangeAt = 0;
@@ -5982,7 +6016,22 @@ var SonioxSTT = class {
       this.onUtterance(combined, now2());
     }, 120);
     this.endpointTimer.unref?.();
+    this.startedChunksAt = now2();
+    const noAudioMs = this.options.noAudioTimeoutMs;
+    if (noAudioMs > 0) {
+      this.noAudioTimer = setInterval(() => {
+        if (this.stopped) return;
+        const ref = this.lastChunkAt || this.startedChunksAt;
+        if (now2() - ref > noAudioMs) {
+          log10.error(`stt: no mic audio for >${Math.round(noAudioMs / 1e3)}s \u2014 capture device stopped delivering`);
+          this.onFatal("microphone stopped delivering audio (try a different input device, e.g. AirPods, or check System Settings \u2192 Sound \u2192 Input)");
+          this.stop();
+        }
+      }, Math.max(250, Math.min(2e3, noAudioMs / 4)));
+      this.noAudioTimer.unref?.();
+    }
     await this.options.source.start((chunk) => {
+      this.lastChunkAt = now2();
       let sum = 0;
       const view = new DataView(chunk.buffer, chunk.byteOffset, chunk.byteLength);
       for (let i = 0; i + 1 < chunk.byteLength; i += 2) {
@@ -6024,6 +6073,7 @@ var SonioxSTT = class {
   stop() {
     this.stopped = true;
     if (this.endpointTimer) clearInterval(this.endpointTimer);
+    if (this.noAudioTimer) clearInterval(this.noAudioTimer);
     this.options.source?.stop();
     if (this.ws) this.ws.onclose = null;
     this.ws?.close();
@@ -7022,7 +7072,8 @@ The filesystem root '/' is the real machine root \u2014 you have full filesystem
     ...o.maxToolCalls != null ? { maxToolCalls: o.maxToolCalls } : {},
     // Context-v2: by default, note-take over a long session (stub tool outputs older than the last 8).
     keepToolOutputs: o.keepToolOutputs ?? 8,
-    ...o.maxContextTokens != null ? { maxContextTokens: o.maxContextTokens } : {}
+    ...o.maxContextTokens != null ? { maxContextTokens: o.maxContextTokens } : {},
+    ...o.browserTrim ? { browserTrim: o.browserTrim } : {}
   });
 }
 function summarizeCall(name, args) {
@@ -7216,22 +7267,46 @@ var AecDuplexAudio = class {
     this.bin = bin;
   }
   bin;
-  aec = true;
+  /** Mutable: starts true (VPIO/AEC). Flips false if we fall back to non-VPIO capture (heuristic tier). */
+  _aec = true;
+  get aec() {
+    return this._aec;
+  }
   onFatal;
   proc = null;
   stopped = false;
   micDenied = false;
+  noVpio = false;
+  // currently running the non-VPIO fallback
+  triedFallback = false;
+  // one-shot guard
+  gotChunk = false;
+  // any mic audio since (re)spawn?
+  onChunk = () => {
+  };
+  fallbackTimer = null;
   bytesWritten = 0;
   startedAt = 0;
   // --- AudioSource ---
   start(onChunk) {
-    this.proc = spawn2(this.bin, [], { stdio: ["pipe", "pipe", "pipe"] });
+    this.onChunk = onChunk;
+    this.spawnHelper();
+  }
+  /** (Re)spawn the helper. On the first spawn, arm a fast watchdog: if VPIO delivers NO audio within
+   *  ~2.5s, the VP input tap is dead on this machine (seen on macOS 26.5.x) — respawn once with
+   *  MIC_NO_VPIO=1 (plain capture, heuristic echo) so the mic actually works instead of starving STT. */
+  spawnHelper() {
+    const env = this.noVpio ? { ...process.env, MIC_NO_VPIO: "1" } : process.env;
+    this.proc = spawn2(this.bin, [], { stdio: ["pipe", "pipe", "pipe"], env });
     this.proc.stdin.on("error", () => {
     });
     this.proc.on("exit", (c) => {
       if (c && !this.stopped) log16.error(`aec duplex audio exited (${c}) \u2014 check mic permission / MIC_AEC=0`);
     });
-    this.proc.stdout.on("data", (chunk) => onChunk(chunk));
+    this.proc.stdout.on("data", (chunk) => {
+      this.gotChunk = true;
+      this.onChunk(chunk);
+    });
     this.proc.stderr.on("data", (d) => {
       for (const ln of String(d).split("\n")) {
         const s = ln.trim();
@@ -7245,9 +7320,21 @@ var AecDuplexAudio = class {
         } else log16.debug(`mic-aec: ${s}`);
       }
     });
-  }
-  stop() {
-    this.stopped = true;
+    if (!this.noVpio && !this.triedFallback) {
+      this.fallbackTimer = setTimeout(() => {
+        if (this.stopped || this.gotChunk) return;
+        this.triedFallback = true;
+        this.noVpio = true;
+        this._aec = false;
+        log16.warn("mic-aec: VPIO delivered no audio in 2.5s \u2014 falling back to non-VPIO capture (heuristic echo; headphones recommended)");
+        this.killProc();
+        this.spawnHelper();
+      }, 2500);
+      this.fallbackTimer.unref?.();
+    }
+  }
+  /** Kill the current child WITHOUT marking the whole source stopped (used for the fallback respawn). */
+  killProc() {
     const p = this.proc;
     this.proc = null;
     if (!p) return;
@@ -7259,6 +7346,11 @@ var AecDuplexAudio = class {
       }
     }, 500).unref?.();
   }
+  stop() {
+    this.stopped = true;
+    if (this.fallbackTimer) clearTimeout(this.fallbackTimer);
+    this.killProc();
+  }
   // --- AudioSink (frame writer; same played/drain byte-math as the ffplay Player) ---
   frame(payload) {
     const stdin = this.proc?.stdin;
@@ -7360,10 +7452,11 @@ var VoiceIO = class extends VoiceEngine {
     });
     this.duplexSource = duplex;
   }
-  /** Host hook for an unrecoverable audio-source failure (e.g. mic permission denied). Only the duplex
-   *  AEC source can hit it; a no-op otherwise. */
+  /** Host hook for an unrecoverable audio failure — mic permission denied (duplex source) or no mic
+   *  audio at all (STT watchdog). Routed to whichever can detect it. */
   set onFatal(fn) {
     if (this.duplexSource) this.duplexSource.onFatal = fn;
+    if (this.stt) this.stt.onFatal = fn;
   }
   /** ready = keys present (AEC vs heuristic is decided at start()) */
   static available(env = process.env) {
@@ -10878,12 +10971,7 @@ function makeAskResolver(cwd) {
     }
   };
 }
-function hasBrowserMcp(cfg, extraTools) {
-  const isBrowser = (s) => /browser/i.test(s);
-  if (Object.entries(cfg.mcpServers ?? {}).some(([n, c]) => isBrowser(n) && !c?.disabled)) return true;
-  return extraTools.some((t) => t.name.startsWith("mcp__") && isBrowser(t.name.slice(5).split("__")[0]));
-}
-var BROWSER_MAX_CONTEXT_TOKENS = 12e4;
+var BROWSER_TRIM = { resultBytes: 8e3, keepOutputs: 2 };
 function optsFor(args, ai, cfg = {}, extraTools = []) {
   const perm = resolvePermMode(args, canPrompt);
   if (perm.notice) err(dim(`  \u26A0 ${perm.notice}
@@ -10922,9 +11010,10 @@ function optsFor(args, ai, cfg = {}, extraTools = []) {
     maxRepeats: cfg.maxRepeats,
     maxToolCalls: cfg.maxToolCalls,
     keepToolOutputs: cfg.keepToolOutputs,
-    // Token-aware backstop: explicit config wins; else auto-enable for browser MCP runs (huge DOM/screenshot
-    // results otherwise accumulate past the maxTokens budget kill-switch mid-browse). Off for normal sessions.
-    maxContextTokens: cfg.maxContextTokens ?? (hasBrowserMcp(cfg, extraTools) ? BROWSER_MAX_CONTEXT_TOKENS : void 0),
+    maxContextTokens: cfg.maxContextTokens,
+    // Adaptive browser trimming (engages on actual browser tool USE, inside the Agent — see BROWSER_TRIM).
+    // Config can tune the values; null-coalesced so a config 0/value still wins over the default.
+    browserTrim: cfg.browserTrim ?? BROWSER_TRIM,
     learnFromMistakes: cfg.learnFromMistakes,
     // Forwarded to cursor/* delegations for environment parity (chat-model providers ignore it).
     // Raw config (pre-OAuth): unresolved-oauth http servers are skipped by the cursor mapper.
@@ -10946,9 +11035,6 @@ async function makeAgent(args, ai, cfg, extraTools = []) {
   if (args.harden && !virtual) err(dim(`  \u26E8 hardened shell: writes confined to cwd+tmp${args.hardenNet ? "" : ", network blocked"} (sandbox-exec/bwrap)
 `));
   const opts = optsFor(args, ai, cfg, extraTools);
-  if (opts.maxContextTokens === BROWSER_MAX_CONTEXT_TOKENS && cfg.maxContextTokens == null)
-    err(dim(`  \u229E browser MCP detected \u2014 auto-capping sent context at ~${BROWSER_MAX_CONTEXT_TOKENS / 1e3}k tok (set maxContextTokens in config to override)
-`));
   const agent = await buildAgent(opts);
   const display = displayHooks(agent.options.fs, { flush: agent.options.host?.flushText });
   agent.options.hooks = cfg.hooks ? composeHooks(display, hooksFromConfig(cfg.hooks)) : display;