@nowcrew/daemon 0.5.29 → 0.5.30

package/README.md

@@ -75,6 +75,10 @@ The server selects the machine and sends a validated `execution:start` containin
 - requested permission and channel/thread identifiers;
 - reporting flags for final text, activity, and console streams.
 
+An Agent bound to a computer has a hard machine constraint: if that computer is offline or incompatible,
+dispatch reports unavailable instead of running on another compatible daemon. An unbound Agent may use
+the global compatible-machine pool.
+
 The daemon intersects requested permission with local policy, checks advertised resource limits, prepares
 the local workspace/environment, and launches only a built-in runtime adapter (`claude`, `codex`, or
 `kimi`). It does not decide collaboration rules, scheduled output policy, fallback delivery, or thread

package/dist/runtimes/codex-app-server-runner.js

@@ -4,6 +4,7 @@ import { parseArgs } from "node:util";
 import { pathToFileURL } from "node:url";
 import spawn from "cross-spawn";
 import { z } from "zod";
+import { signalSupervisorTree } from "../execution-supervisor.js";
 import { startFirstProgressWatchdog } from "./progress-watchdog.js";
 const RunnerInputSchema = z.object({
     systemPrompt: z.string().min(1),
@@ -18,6 +19,22 @@ const RunnerInputSchema = z.object({
 const RPC_TIMEOUT_MS = 30_000;
 const INITIALIZE_RPC_TIMEOUT_MS = 60_000;
 const ERROR_MESSAGE_CAP = 2_000;
+// LocalExecutor retains the final 2,000 stderr characters; leave room for stage and error lines.
+const STDERR_TAIL_CAP = 1_200;
+const STDERR_LINE_CAPTURE_CAP = 1_200;
+const STDERR_LINE_OMITTED = "[stderr line omitted: exceeded capture limit]\n";
+const MAX_INITIALIZE_ATTEMPTS = 2;
+const PROCESS_TREE_STOP_TIMEOUT_MS = 1_000;
+class CodexRpcTimeoutError extends Error {
+    method;
+    timeoutMs;
+    constructor(method, timeoutMs) {
+        super(`Codex app-server ${method} timed out after ${timeoutMs}ms`);
+        this.method = method;
+        this.timeoutMs = timeoutMs;
+        this.name = "CodexRpcTimeoutError";
+    }
+}
 export function codexRpcTimeoutMs(method) {
     return method === "initialize" ? INITIALIZE_RPC_TIMEOUT_MS : RPC_TIMEOUT_MS;
 }
@@ -29,6 +46,87 @@ function safeErrorMessage(error, secrets) {
     }
     return message.slice(0, ERROR_MESSAGE_CAP);
 }
+function sensitiveEnvironmentValues(env) {
+    return Object.entries(env)
+        .filter(([name, value]) => value !== undefined
+        && /(?:api.?key|token|secret|password|credential|authorization)/i.test(name))
+        .map(([, value]) => value);
+}
+function redactionSecrets(input, env) {
+    const values = [input.systemPrompt, input.wakePrompt, ...sensitiveEnvironmentValues(env)];
+    return [...new Set(values.flatMap((value) => [
+            value,
+            ...value.split(/\r?\n/).map((line) => line.trim()),
+        ]).filter(Boolean))];
+}
+export function redactCodexStderr(text, secrets) {
+    let redacted = text;
+    for (const secret of [...secrets].sort((left, right) => right.length - left.length)) {
+        if (secret)
+            redacted = redacted.replaceAll(secret, "[redacted]");
+    }
+    return redacted
+        .replace(/(Bearer\s+)[^\s"']+/gi, "$1[redacted]")
+        .replace(/\bsk-[A-Za-z0-9_-]{6,}\b/g, "[redacted]")
+        .replace(/((?:api[_-]?key|access[_-]?token|auth(?:orization)?|password|secret)\s*[=:]\s*)[^\s"']+/gi, "$1[redacted]");
+}
+class StderrCapture {
+    secrets;
+    tail = "";
+    pendingLine = "";
+    omittingLine = false;
+    constructor(child, secrets) {
+        this.secrets = secrets;
+        child.stderr?.setEncoding("utf8");
+        child.stderr?.on("data", (chunk) => this.append(chunk));
+        child.stderr?.once("end", () => this.flush());
+    }
+    redactedTail() {
+        const pending = this.omittingLine
+            ? STDERR_LINE_OMITTED
+            : redactCodexStderr(this.pendingLine, this.secrets);
+        return `${this.tail}${pending}`.slice(-STDERR_TAIL_CAP).trim();
+    }
+    append(chunk) {
+        let remaining = chunk;
+        while (remaining.length > 0) {
+            const newline = remaining.indexOf("\n");
+            const segment = newline < 0 ? remaining : remaining.slice(0, newline);
+            if (!this.omittingLine) {
+                if (this.pendingLine.length + segment.length <= STDERR_LINE_CAPTURE_CAP) {
+                    this.pendingLine += segment;
+                }
+                else {
+                    this.pendingLine = "";
+                    this.omittingLine = true;
+                }
+            }
+            if (newline < 0)
+                return;
+            this.emit(this.omittingLine ? STDERR_LINE_OMITTED : `${this.pendingLine}\n`);
+            this.pendingLine = "";
+            this.omittingLine = false;
+            remaining = remaining.slice(newline + 1);
+        }
+    }
+    flush() {
+        if (this.pendingLine.length === 0 && !this.omittingLine)
+            return;
+        this.emit(this.omittingLine ? STDERR_LINE_OMITTED : this.pendingLine);
+        this.pendingLine = "";
+        this.omittingLine = false;
+    }
+    emit(text) {
+        const redacted = redactCodexStderr(text, this.secrets);
+        process.stderr.write(redacted);
+        this.tail = `${this.tail}${redacted}`.slice(-STDERR_TAIL_CAP);
+    }
+}
+function logStage(stage, status, attempt, startedAt, detail) {
+    const suffix = detail === undefined ? "" : ` ${detail}`;
+    process.stderr.write(`[codex-app-server] stage=${stage} status=${status} attempt=${attempt}`
+        + ` elapsed_ms=${Date.now() - startedAt}${suffix}\n`);
+}
 function jsonLine(event) {
     if (process.stdout.write(`${JSON.stringify(event)}\n`))
         return Promise.resolve();
@@ -117,7 +215,7 @@ class CodexRpcClient {
         return new Promise((resolve, reject) => {
             const timer = setTimeout(() => {
                 this.pending.delete(id);
-                reject(new Error(`Codex app-server ${method} timed out after ${timeoutMs}ms`));
+                reject(new CodexRpcTimeoutError(method, timeoutMs));
             }, timeoutMs);
             this.pending.set(id, { resolve, reject, timer });
             this.write({ jsonrpc: "2.0", id, method, params });
@@ -184,31 +282,70 @@ class CodexRpcClient {
         this.pending.clear();
     }
 }
-async function stopChild(child) {
-    if (child.exitCode !== null || child.signalCode !== null)
+function processTreeAlive(child) {
+    if (process.platform === "win32" || child.pid === undefined) {
+        return child.exitCode === null && child.signalCode === null;
+    }
+    try {
+        process.kill(-child.pid, 0);
+        return true;
+    }
+    catch (error) {
+        return error.code === "EPERM";
+    }
+}
+async function signalProcessTree(child, signal) {
+    if (child.pid !== undefined) {
+        try {
+            await signalSupervisorTree(child.pid, signal);
+            return;
+        }
+        catch (error) {
+            if (error.code !== "ESRCH")
+                throw error;
+        }
+    }
+    if (child.exitCode === null && child.signalCode === null)
+        child.kill(signal);
+}
+async function waitForProcessTreeExit(child, timeoutMs) {
+    const deadline = Date.now() + timeoutMs;
+    while (processTreeAlive(child) && Date.now() < deadline) {
+        await new Promise((resolve) => setTimeout(resolve, 25));
+    }
+    return !processTreeAlive(child);
+}
+async function stopChildTree(child) {
+    if (!processTreeAlive(child))
+        return;
+    await signalProcessTree(child, "SIGTERM");
+    if (await waitForProcessTreeExit(child, PROCESS_TREE_STOP_TIMEOUT_MS))
         return;
-    const closed = once(child, "close").then(() => undefined);
-    child.kill("SIGTERM");
-    let timer;
-    const graceful = await Promise.race([
-        closed.then(() => true),
-        new Promise((resolve) => { timer = setTimeout(() => resolve(false), 1_000); }),
-    ]);
-    if (timer !== undefined)
-        clearTimeout(timer);
-    if (!graceful && child.exitCode === null && child.signalCode === null) {
-        child.kill("SIGKILL");
-        await closed;
+    await signalProcessTree(child, "SIGKILL");
+    if (!await waitForProcessTreeExit(child, PROCESS_TREE_STOP_TIMEOUT_MS)) {
+        throw new Error(`Codex app-server process tree ${child.pid ?? "unknown"} did not exit`);
     }
 }
-export async function runCodexAppServer(bin) {
-    const input = await readRunnerInput();
+async function runCodexAppServerAttempt(bin, input, attempt, initializeTimeoutMs) {
+    const attemptStartedAt = Date.now();
     const child = spawn(bin, ["app-server", "--listen", "stdio://"], {
         cwd: process.cwd(),
         env: process.env,
         stdio: ["pipe", "pipe", "pipe"],
+        detached: process.platform !== "win32",
+    });
+    const secrets = redactionSecrets(input, process.env);
+    const stderr = new StderrCapture(child, secrets);
+    logStage("spawn", "start", attempt, attemptStartedAt, `pid=${child.pid ?? "unknown"}`);
+    let didSpawn = false;
+    child.once("spawn", () => {
+        didSpawn = true;
+        logStage("spawn", "ok", attempt, attemptStartedAt, `pid=${child.pid ?? "unknown"}`);
+    });
+    child.once("error", () => {
+        if (!didSpawn)
+            logStage("spawn", "error", attempt, attemptStartedAt);
     });
-    child.stderr?.pipe(process.stderr, { end: false });
     let threadId = null;
     let announcedThreadId = null;
     let turnId = null;
@@ -246,6 +383,11 @@ export async function runCodexAppServer(bin) {
                 process.stderr.write(`Codex turn error: ${detail.slice(0, ERROR_MESSAGE_CAP)}\n`);
         }
     }, completionReject);
+    let treeStopPromise = null;
+    const ensureChildTreeStopped = () => {
+        treeStopPromise ??= stopChildTree(child);
+        return treeStopPromise;
+    };
     let cancelling = false;
     const cancel = async () => {
         if (cancelling)
@@ -254,17 +396,22 @@ export async function runCodexAppServer(bin) {
         if (threadId !== null && turnId !== null) {
             await rpc.request("turn/interrupt", { threadId, turnId }, 5_000).catch(() => undefined);
         }
-        await stopChild(child);
+        await ensureChildTreeStopped();
     };
     const onSignal = () => { void cancel().finally(() => process.exit(130)); };
     process.once("SIGTERM", onSignal);
     process.once("SIGINT", onSignal);
+    let activeStage = "initialize";
+    let stageStartedAt = Date.now();
     try {
         await rpc.request("initialize", {
             clientInfo: { name: "nowcrew-daemon", version: "1" },
             capabilities: { experimentalApi: true, requestAttestation: false },
-        });
+        }, initializeTimeoutMs);
+        logStage(activeStage, "ok", attempt, stageStartedAt);
         rpc.notify("initialized");
+        activeStage = input.resume ? "thread_resume" : "thread_start";
+        stageStartedAt = Date.now();
         const threadParams = {
             cwd: process.cwd(),
             approvalPolicy: "never",
@@ -275,6 +422,7 @@ export async function runCodexAppServer(bin) {
         const thread = input.resume && input.sessionId !== undefined
             ? await rpc.request("thread/resume", { threadId: input.sessionId, ...threadParams })
             : await rpc.request("thread/start", threadParams);
+        logStage(activeStage, "ok", attempt, stageStartedAt);
         threadId = thread.thread.id;
         if (announcedThreadId !== threadId) {
             announcedThreadId = threadId;
@@ -284,6 +432,8 @@ export async function runCodexAppServer(bin) {
             completionReject(new Error("Codex produced no semantic progress within the startup window"));
             void cancel();
         });
+        activeStage = "turn_start";
+        stageStartedAt = Date.now();
         const started = await rpc.request("turn/start", {
             threadId,
             input: [
@@ -294,8 +444,12 @@ export async function runCodexAppServer(bin) {
                 ? {}
                 : { effort: input.reasoning }),
         });
+        logStage(activeStage, "ok", attempt, stageStartedAt);
         turnId = started.turn.id;
+        activeStage = "turn_complete";
+        stageStartedAt = Date.now();
         const completed = await completion;
+        logStage(activeStage, "ok", attempt, stageStartedAt);
         firstProgress.stop();
         const usage = lastUsage?.last;
         await jsonLine({
@@ -309,33 +463,76 @@ export async function runCodexAppServer(bin) {
             }),
         });
         if (completed.turn?.status === "completed")
-            return 0;
+            return { code: 0, initializeTimedOut: false };
         const detail = completed.turn?.error?.message ?? `turn status ${completed.turn?.status ?? "unknown"}`;
         process.stderr.write(`Codex turn failed: ${detail.slice(0, ERROR_MESSAGE_CAP)}\n`);
-        return completed.turn?.status === "interrupted" ? 130 : 1;
+        return {
+            code: completed.turn?.status === "interrupted" ? 130 : 1,
+            initializeTimedOut: false,
+        };
     }
     catch (error) {
-        process.stderr.write(`Codex app-server execution failed: ${safeErrorMessage(error, [input.systemPrompt, input.wakePrompt])}\n`);
-        return 1;
+        const initializeTimedOut = error instanceof CodexRpcTimeoutError && error.method === "initialize";
+        if (initializeTimedOut) {
+            const tail = stderr.redactedTail();
+            logStage(activeStage, "timeout", attempt, stageStartedAt, `stderr_tail=${JSON.stringify(tail || "[empty]")}`);
+        }
+        else {
+            logStage(activeStage, "error", attempt, stageStartedAt);
+        }
+        process.stderr.write(`Codex app-server execution failed: ${safeErrorMessage(error, secrets)}\n`);
+        return { code: 1, initializeTimedOut };
     }
     finally {
         firstProgress.stop();
         process.off("SIGTERM", onSignal);
         process.off("SIGINT", onSignal);
-        await stopChild(child);
+        const cleanupStartedAt = Date.now();
+        try {
+            await ensureChildTreeStopped();
+            logStage("cleanup", "ok", attempt, cleanupStartedAt);
+        }
+        catch (error) {
+            logStage("cleanup", "error", attempt, cleanupStartedAt);
+            throw error;
+        }
+    }
+}
+export async function runCodexAppServer(bin, options = {}) {
+    const input = await readRunnerInput();
+    const initializeTimeoutMs = Math.min(options.initializeTimeoutMs ?? INITIALIZE_RPC_TIMEOUT_MS, INITIALIZE_RPC_TIMEOUT_MS);
+    for (let attempt = 1; attempt <= MAX_INITIALIZE_ATTEMPTS; attempt += 1) {
+        const result = await runCodexAppServerAttempt(bin, input, attempt, initializeTimeoutMs);
+        if (!result.initializeTimedOut || attempt === MAX_INITIALIZE_ATTEMPTS)
+            return result.code;
+        process.stderr.write(`[codex-app-server] stage=retry status=start attempt=${attempt + 1} elapsed_ms=0 reason=initialize_timeout\n`);
     }
+    return 1;
 }
-function binFromArgv(argv) {
+function configFromArgv(argv) {
     const { values } = parseArgs({
         args: [...argv],
-        options: { bin: { type: "string" } },
+        options: {
+            bin: { type: "string" },
+            "initialize-timeout-ms": { type: "string" },
+        },
     });
     if (!values.bin)
         throw new Error("--bin is required");
-    return values.bin;
+    const rawTimeout = values["initialize-timeout-ms"];
+    if (rawTimeout === undefined)
+        return { bin: values.bin };
+    const initializeTimeoutMs = Number(rawTimeout);
+    if (!Number.isInteger(initializeTimeoutMs) || initializeTimeoutMs <= 0) {
+        throw new Error("--initialize-timeout-ms must be a positive integer");
+    }
+    return { bin: values.bin, initializeTimeoutMs };
 }
 if (process.argv[1] !== undefined && import.meta.url === pathToFileURL(process.argv[1]).href) {
-    runCodexAppServer(binFromArgv(process.argv.slice(2)))
+    const config = configFromArgv(process.argv.slice(2));
+    runCodexAppServer(config.bin, config.initializeTimeoutMs === undefined
+        ? {}
+        : { initializeTimeoutMs: config.initializeTimeoutMs })
         .then((code) => { process.exitCode = code; })
         .catch((error) => {
         process.stderr.write(`Codex app-server runner failed: ${safeErrorMessage(error, [])}\n`);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@nowcrew/daemon",
-  "version": "0.5.29",
+  "version": "0.5.30",
   "type": "module",
   "description": "crew daemon — 运行在用户机器:拉起/管理 agent 进程,注入 crew CLI,归一化 runtime 事件",
   "license": "Apache-2.0",