mirrors/pikasTech-agentrun
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

fix: use idle timeout for codex turns

Browse Source
This commit is contained in:
AgentRun Codex
2026-06-11 12:52:46 +08:00
parent dafcfc7ce4
commit f49695a821
11 changed files with 115 additions and 72 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/selftest/cases/30-codex-stdio.ts
+10 -12
View File
@@ -182,7 +182,7 @@ const selfTest: SelfTestCase = async (context) => {
assert.equal(liveResult.terminalStatus, "completed", "slow live tool event turn should complete");
await runInterruptBeforeTurnStartResponseCase({ client, managerUrl: server.baseUrl, context });
await runHardTimeoutDuringToolProgressCase({ client, managerUrl: server.baseUrl, context });
await runToolProgressRefreshesIdleTimeoutCase({ client, managerUrl: server.baseUrl, context });
const noisy = await createRunWithCommand(client, context, "hello noisy reasoning", "selftest-noisy-reasoning-events", 15_000);
const noisyResult = await runOnce({ managerUrl: server.baseUrl, runId: noisy.runId, codexCommand: context.fakeCodexCommand, codexArgs: context.fakeCodexArgs, codexHome: context.codexHome, env: { CODEX_HOME: context.codexHome, AGENTRUN_FAKE_CODEX_MODE: "noisy-reasoning-events" }, oneShot: true }) as JsonRecord;
@@ -245,7 +245,7 @@ const selfTest: SelfTestCase = async (context) => {
await runSessionStorageSubdirCase({ client, managerUrl: server.baseUrl, context });
await runSessionStorageNoSecretLeakCase({ client, managerUrl: server.baseUrl, context });
return { name: "codex-stdio", tests: ["runner-lease-heartbeat", "runner-lease-conflict-recovery", "codex-stdio-fake-turn", "codex-stdio-k8s-sandbox-override", "codex-stdio-projected-writable-home", "codex-stdio-deepseek-profile-fake-turn", "codex-stdio-dsflash-go-profile-fake-turn", "codex-stdio-dsflash-go-config-metadata", "codex-stdio-minimax-m3-profile-fake-turn", "codex-stdio-deepseek-missing-secret-no-fallback", "codex-stdio-minimax-m3-missing-secret-no-fallback", "codex-stdio-config-model-authoritative", "codex-stdio-explicit-model-forwarded", "codex-stdio-final-agent-message-only", "codex-stdio-web-search-progress", "codex-stdio-stale-thread-resume-failed", "codex-stdio-live-tool-events", "codex-stdio-interrupt-before-turn-start-response", "codex-stdio-hard-timeout-during-tool-progress", "codex-stdio-noisy-reasoning-suppression", "codex-stdio-missing-turn-result", "codex-stdio-provider-auth-failed", "codex-stdio-provider-rate-limited", "codex-stdio-provider-invalid-tool-call", "codex-stdio-provider-compact-unsupported", "codex-stdio-provider-stream-disconnected", "codex-stdio-provider-503-rpc-error", "codex-stdio-provider-503-terminal", "codex-stdio-provider-unavailable", "codex-stdio-provider-503-retry-event", "codex-stdio-provider-refused-retry-recovered", "codex-stdio-invalid-json", "codex-stdio-timeout", "codex-stdio-idle-timeout-progress-refresh", "codex-stdio-command-failure-keeps-run-open", "codex-stdio-secret-unavailable", "codex-stdio-spawn-failure"] };
return { name: "codex-stdio", tests: ["runner-lease-heartbeat", "runner-lease-conflict-recovery", "codex-stdio-fake-turn", "codex-stdio-k8s-sandbox-override", "codex-stdio-projected-writable-home", "codex-stdio-deepseek-profile-fake-turn", "codex-stdio-dsflash-go-profile-fake-turn", "codex-stdio-dsflash-go-config-metadata", "codex-stdio-minimax-m3-profile-fake-turn", "codex-stdio-deepseek-missing-secret-no-fallback", "codex-stdio-minimax-m3-missing-secret-no-fallback", "codex-stdio-config-model-authoritative", "codex-stdio-explicit-model-forwarded", "codex-stdio-final-agent-message-only", "codex-stdio-web-search-progress", "codex-stdio-stale-thread-resume-failed", "codex-stdio-live-tool-events", "codex-stdio-interrupt-before-turn-start-response", "codex-stdio-tool-progress-refreshes-idle-timeout", "codex-stdio-noisy-reasoning-suppression", "codex-stdio-missing-turn-result", "codex-stdio-provider-auth-failed", "codex-stdio-provider-rate-limited", "codex-stdio-provider-invalid-tool-call", "codex-stdio-provider-compact-unsupported", "codex-stdio-provider-stream-disconnected", "codex-stdio-provider-503-rpc-error", "codex-stdio-provider-503-terminal", "codex-stdio-provider-unavailable", "codex-stdio-provider-503-retry-event", "codex-stdio-provider-refused-retry-recovered", "codex-stdio-invalid-json", "codex-stdio-timeout", "codex-stdio-idle-timeout-progress-refresh", "codex-stdio-command-failure-keeps-run-open", "codex-stdio-secret-unavailable", "codex-stdio-spawn-failure"] };
} finally {
await new Promise<void>((resolve) => server.server.close(() => resolve()));
}
@@ -350,26 +350,24 @@ async function runInterruptBeforeTurnStartResponseCase(options: { client: Manage
assertNoSecretLeak({ result, events });
}
async function runHardTimeoutDuringToolProgressCase(options: { client: ManagerClient; managerUrl: string; context: SelfTestContext }): Promise<void> {
const item = await createRunWithCommand(options.client, options.context, "hard timeout during tool progress", "selftest-hard-timeout-tool-progress", 120);
async function runToolProgressRefreshesIdleTimeoutCase(options: { client: ManagerClient; managerUrl: string; context: SelfTestContext }): Promise<void> {
const item = await createRunWithCommand(options.client, options.context, "tool progress refreshes idle timeout", "selftest-tool-progress-refreshes-idle", 120);
const result = await runOnce({
managerUrl: options.managerUrl,
runId: item.runId,
codexCommand: options.context.fakeCodexCommand,
codexArgs: options.context.fakeCodexArgs,
codexHome: options.context.codexHome,
env: { CODEX_HOME: options.context.codexHome, AGENTRUN_FAKE_CODEX_MODE: "hard-timeout-tool-progress" },
env: { CODEX_HOME: options.context.codexHome, AGENTRUN_FAKE_CODEX_MODE: "tool-progress-refreshes-idle" },
oneShot: true,
}) as JsonRecord;
assert.equal(result.terminalStatus, "failed", "hard timeout should fail even while the tool keeps producing progress");
assert.equal(result.failureKind, "backend-timeout");
assert.equal(result.terminalStatus, "completed", "tool progress should refresh idle timeout until terminal completion");
assert.equal(result.failureKind, null);
const events = await options.client.get(`/api/v1/runs/${item.runId}/events?afterSeq=0&limit=100`) as { items?: Array<{ type: string; payload: unknown }> };
assert.ok(events.items?.some((event) => event.type === "command_output" && String(eventPayload(event).text ?? "").includes("progress")), "progress output should be visible before hard timeout");
assert.ok(events.items?.some((event) => event.type === "error" && eventPayload(event).phase === "turn:hard-timeout"), "hard timeout should be recorded as an error event");
assert.ok(events.items?.some((event) => event.type === "backend_status" && eventPayload(event).phase === "turn-interrupt-requested"), "timeout should request backend interrupt before process teardown");
assert.ok(events.items?.some((event) => event.type === "backend_status" && eventPayload(event).phase === "turn/interrupt:completed"), "timeout interrupt result should be visible");
assert.ok(events.items?.some((event) => event.type === "command_output" && String(eventPayload(event).text ?? "").includes("progress 3")), "progress output should stay visible while idle timeout is refreshed");
assert.equal(events.items?.some((event) => event.type === "error" && eventPayload(event).failureKind === "backend-timeout"), false, "progressing tool output must not fail on wall-clock elapsed time");
const command = await options.client.get(`/api/v1/runs/${item.runId}/commands/${item.commandId}`) as { state?: string };
assert.equal(command.state, "failed", "hard timed out command should be failed");
assert.equal(command.state, "completed", "command should complete after progress-delayed terminal status");
assertNoSecretLeak({ result, events });
}