惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Microsoft Azure Blog
Microsoft Azure Blog
有赞技术团队
有赞技术团队
Threat Intelligence Blog | Flashpoint
Threat Intelligence Blog | Flashpoint
cs.CV updates on arXiv.org
cs.CV updates on arXiv.org
F
Fox-IT International blog
Recorded Future
Recorded Future
T
ThreatConnect
T
The Exploit Database - CXSecurity.com
SecWiki News
SecWiki News
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
人人都是产品经理
人人都是产品经理
T
Tenable Blog
L
LINUX DO - 最新话题
博客园_首页
Hugging Face - Blog
Hugging Face - Blog
罗磊的独立博客
博客园 - 司徒正美
The Hacker News
The Hacker News
博客园 - 聂微东
CTFtime.org: upcoming CTF events
CTFtime.org: upcoming CTF events
Scott Helme
Scott Helme
博客园 - 【当耐特】
O
OpenAI News
Schneier on Security
Schneier on Security
Latest news
Latest news
S
Security @ Cisco Blogs
S
Secure Thoughts
F
Full Disclosure
L
Lohrmann on Cybersecurity
S
SegmentFault 最新的问题
T
Tor Project blog
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
量子位
小众软件
小众软件
T
Threat Research - Cisco Blogs
Simon Willison's Weblog
Simon Willison's Weblog
IT之家
IT之家
大猫的无限游戏
大猫的无限游戏
N
News and Events Feed by Topic
E
Exploit-DB.com RSS Feed
J
Java Code Geeks
Last Week in AI
Last Week in AI
酷 壳 – CoolShell
酷 壳 – CoolShell
Application and Cybersecurity Blog
Application and Cybersecurity Blog
S
Schneier on Security
Cisco Talos Blog
Cisco Talos Blog
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
P
Proofpoint News Feed
Recent Commits to openclaw:main
Recent Commits to openclaw:main
雷峰网
雷峰网

Recent Commits to openclaw:main

perf(utils): preserve message identity in stripInlineDirectiveTagsFro… fix(status): show configured cost for aws-sdk models (#85619) fix(microsoft-foundry): DeepSeek V4 models incorrectly use openai-com… · openclaw/openclaw@ec43acb fix(skills): show empty state notice in config wizard (#85032) test(codex): avoid forced-tool allowlist flake fix: avoid gateway startup event-loop stalls test(codex): type forced-tool request mock · openclaw/openclaw@f6ab188 test(codex): avoid forced-tool turn flake · openclaw/openclaw@8d1ab83 test(codex): avoid startup cleanup socket flake fix(gateway): pin relative state dir at startup · openclaw/openclaw@2e5be0c fix(whatsapp): persist inbound delivery in plugin state (#85506) · openclaw/openclaw@b47bace test(codex): make sandbox cleanup proof deterministic fix(cron): route topic targets through channel plugins · openclaw/openclaw@9175491 fix(agents): simplify subagent completion handoff ci(release): isolate npm publish concurrency · openclaw/openclaw@9c26b87 ci(release): allow beta publish after npm preflight · openclaw/openclaw@0e37263 fix(release): allow large beta smoke run lists ci(release): retry child workflow polling · openclaw/openclaw@c689f71 ci(release): poll child workflows through actions api · openclaw/openclaw@e5dab55 fix(bootstrap): guard bootstrap name checks against undefined names (… · openclaw/openclaw@25fa46b fix(cli): waitForever must keep the event loop alive (#85694) · openclaw/openclaw@eca9645 fix(cli-output): ignore cumulative usage from result events in stream… · openclaw/openclaw@84229d9 fix(secrets): show irreversible warning after interactive apply confi… · openclaw/openclaw@bb52b54 fix(agents/harness): pass CLI runtime aliases through to PI in select… · openclaw/openclaw@2c3b7ea docs: add bug-sweep changelog entries fix(release): run npm shims on Windows · openclaw/openclaw@1f32a48 fix(ci): repair crabbox hydrate replay (#85706) · openclaw/openclaw@0d7d99b feat(auth): support named model login profiles fix(gateway): restore WebChat image understanding routing · openclaw/openclaw@55a0c9b fix(anthropic): migrate 1M context to GA handling fix(scripts): resolve Crabbox shims on Windows · openclaw/openclaw@5c535df test(ci): harden installer smoke coverage · openclaw/openclaw@68bcd4e fix(status): bound deep docker audit probes (#85476) · openclaw/openclaw@f7c05dc docs: expand meeting notes docs · openclaw/openclaw@a7e0fa0 fix(cli): validate tasks audit limit (#84901) · openclaw/openclaw@44d470f fix(twitch): preserve newer message handler during cleanup (#85425) · openclaw/openclaw@71ddc01 fix(ci): require live docker credentials by resource · openclaw/openclaw@1e21121 fix(diagnostics): drop snake case otel ids (#72645) · openclaw/openclaw@e0bafc5 fix(ci): clear signal and docs guard blockers (#85693) · openclaw/openclaw@3a1d4dd fix(ci): require factory auth for droid live docker · openclaw/openclaw@cc6c372 fix: honor disabled synthetic auth lookup · openclaw/openclaw@a4a1abb fix(scripts): repair live docker auth shellcheck · openclaw/openclaw@4e34ac4 fix(docker): avoid printing gateway token · openclaw/openclaw@5db773f fix(cli): keep logs follow on live gateway state · openclaw/openclaw@6e3b318 fix(e2e): prefer x64 MinGit on Windows · openclaw/openclaw@15d9134 feat: add meeting notes plugin docs: update changelog for memory artifacts (#85060) (thanks @brokema… · openclaw/openclaw@9e55383 fix(memory-lancedb): expose public memory artifacts · openclaw/openclaw@aac1abe fix(memory): preserve sidecar capability hooks · openclaw/openclaw@e6288ca fix(e2e): scrub Windows update config on PowerShell 5.1 · openclaw/openclaw@6657b49 docs: absorb documentation PR sweep · openclaw/openclaw@2c536a8 fix(agents): stabilize Linux fallback tests · openclaw/openclaw@6b04170 fix(codex): preserve native web search action metadata (#85378) · openclaw/openclaw@bcf756c test: refresh Codex prompt snapshots · openclaw/openclaw@492d656 docs: absorb hook and subagent guidance PRs fix(agents): audit tool policy blocks (#85673) · openclaw/openclaw@09dd051 ci: fix plugin npm bundled dependency install · openclaw/openclaw@d485464 feat(diagnostics): classify skill and tool usage (#80370) docs(skills): clarify control ui recording proof (#85568) · openclaw/openclaw@0b476b9 test(agents): repair main failure fixtures · openclaw/openclaw@c29967b feat(diagnostics): trace gateway secret preparation (#83019) · openclaw/openclaw@4f0c902 fix(diagnostics): harden observability exports and smokes (#85371) · openclaw/openclaw@7f05be0 fix(stepfun): drop stale auth choice metadata · openclaw/openclaw@0b2ab6c test(e2e): sample kitchen sink RSS on Windows · openclaw/openclaw@73c1e37 test(plugins): fail gauntlet on load diagnostics fix(build): preserve tsdown heap floor · openclaw/openclaw@9ff1a43 fix(tools): honor config apiKey in media tool preflight (#85570) · openclaw/openclaw@31c269f fix(e2e): support macOS script wrappers fix(ci): scope changed shrinkwrap checks · openclaw/openclaw@743fd4c chore(ui): refresh fa control ui locale chore(ui): refresh nl control ui locale · openclaw/openclaw@908464b chore(ui): refresh vi control ui locale · openclaw/openclaw@62b75f4 chore(ui): refresh th control ui locale · openclaw/openclaw@fc4ba31 chore(ui): refresh id control ui locale · openclaw/openclaw@5b1bdd1 chore(ui): refresh pl control ui locale · openclaw/openclaw@534d4b1 chore(ui): refresh uk control ui locale · openclaw/openclaw@055c3bd chore(ui): refresh ar control ui locale · openclaw/openclaw@89c5a68 chore(ui): refresh it control ui locale · openclaw/openclaw@44ca805 chore(ui): refresh tr control ui locale chore(ui): refresh fr control ui locale · openclaw/openclaw@2240b0e chore(ui): refresh ko control ui locale · openclaw/openclaw@5fa250b chore(ui): refresh ja-JP control ui locale · openclaw/openclaw@f4ea401 chore(ui): refresh es control ui locale · openclaw/openclaw@751dde0 chore(ui): refresh de control ui locale · openclaw/openclaw@72a9b5b chore(ui): refresh pt-BR control ui locale · openclaw/openclaw@501b6e0 chore(ui): refresh zh-CN control ui locale chore(ui): refresh zh-TW control ui locale · openclaw/openclaw@88dee79 Add Slack approval QA checkpoints (#85141) fix(agents): mirror internal ui message tool replies (#85564) perf(whatsapp): narrow runtime setter entry (#85589) · openclaw/openclaw@463929d fix: smooth chat focus mode layout · openclaw/openclaw@bb5abef Fix Telegram missing harness spool poison (#85605) · openclaw/openclaw@b745082 fix(session): surface previous-transcript archive failures on /new ro… · openclaw/openclaw@679a46d Policy: add agent workspace conformance checks (#85096) · openclaw/openclaw@a94f344 fix(installer): fail failed Windows git builds · openclaw/openclaw@2edd6e2 fix(codex): honor node exec policy for native surfaces (#85534) · openclaw/openclaw@e0405ec fix(qa-lab): stabilize codex runtime parity fixtures fix(github): preserve sufficient proof against negative relabel (#85567) · openclaw/openclaw@6b52dff docs: add changelog for context pressure preflight · openclaw/openclaw@5ca734f Fix context pressure preflight for tool-heavy sessions (#85541) · openclaw/openclaw@c08400e
fix(agents): add openai-responses family to non-visible turn retry gu… · openclaw/openclaw@49e9c3e
SebTardif · 2026-05-23 · via Recent Commits to openclaw:main
Original file line numberDiff line numberDiff line change

@@ -1594,6 +1594,54 @@ describe("runEmbeddedPiAgent incomplete-turn safety", () => {

15941594

expect(retryInstruction).toBeNull();

15951595

});

15961596
1597+

it("retries empty openai-codex-responses turns with non-zero output tokens (#85364)", () => {

1598+

const retryInstruction = resolveEmptyResponseRetryInstruction({

1599+

provider: "openai-codex",

1600+

modelId: "gpt-5.5",

1601+

modelApi: "openai-codex-responses",

1602+

payloadCount: 0,

1603+

aborted: false,

1604+

timedOut: false,

1605+

attempt: makeAttemptResult({

1606+

assistantTexts: [],

1607+

lastAssistant: {

1608+

role: "assistant",

1609+

stopReason: "stop",

1610+

provider: "openai-codex",

1611+

model: "gpt-5.5",

1612+

content: [],

1613+

usage: { input: 24794, output: 111, cacheRead: 4608, totalTokens: 29513 },

1614+

} as unknown as EmbeddedRunAttemptResult["lastAssistant"],

1615+

}),

1616+

});

1617+
1618+

expect(retryInstruction).toBe(EMPTY_RESPONSE_RETRY_INSTRUCTION);

1619+

});

1620+
1621+

it("retries empty openai-responses turns without visible text", () => {

1622+

const retryInstruction = resolveEmptyResponseRetryInstruction({

1623+

provider: "openai",

1624+

modelId: "gpt-5.5",

1625+

modelApi: "openai-responses",

1626+

payloadCount: 0,

1627+

aborted: false,

1628+

timedOut: false,

1629+

attempt: makeAttemptResult({

1630+

assistantTexts: [],

1631+

lastAssistant: {

1632+

role: "assistant",

1633+

stopReason: "stop",

1634+

provider: "openai",

1635+

model: "gpt-5.5",

1636+

content: [],

1637+

usage: { input: 5000, output: 200, totalTokens: 5200 },

1638+

} as unknown as EmbeddedRunAttemptResult["lastAssistant"],

1639+

}),

1640+

});

1641+
1642+

expect(retryInstruction).toBe(EMPTY_RESPONSE_RETRY_INSTRUCTION);

1643+

});

1644+
15971645

it("retries generic empty OpenAI-compatible turns from custom endpoints", () => {

15981646

const retryInstruction = resolveEmptyResponseRetryInstruction({

15991647

provider: "llama-cpp-local",

Original file line numberDiff line numberDiff line change

@@ -133,6 +133,19 @@ const GEMINI_INCOMPLETE_TURN_MODEL_ID_PATTERN = /^gemini(?:[.-]|$)/;

133133

// Ollama native `/api/chat` can finish with only thinking/internal blocks when

134134

// constrained, but it should not inherit the stricter planning-only/ack prompts.

135135

const OLLAMA_INCOMPLETE_TURN_PROVIDER_ID_PATTERN = /^ollama(?:-|$)/;

136+

// Model APIs eligible for the non-visible turn retry guard. OpenAI Responses

137+

// family can produce reasoning-only turns where usage.output > 0 but no visible

138+

// text is emitted; without the guard these pass through as successful. (#85364)

139+

const RETRY_GUARD_MODEL_APIS = new Set([

140+

"openai-completions",

141+

"anthropic-messages",

142+

"bedrock-converse-stream",

143+

"openai-responses",

144+

"openai-codex-responses",

145+

"azure-openai-responses",

146+

"openclaw-openai-responses-transport",

147+

"openclaw-azure-openai-responses-transport",

148+

]);

136149

const DEFAULT_PLANNING_ONLY_RETRY_LIMIT = 1;

137150

const STRICT_AGENTIC_PLANNING_ONLY_RETRY_LIMIT = 2;

138151

// Allow one immediate continuation plus one follow-up continuation before

@@ -627,11 +640,7 @@ function shouldApplyNonVisibleTurnRetryGuard(params: {

627640

if (shouldApplyPlanningOnlyRetryGuard(params)) {

628641

return true;

629642

}

630-

if (

631-

normalizeLowercaseStringOrEmpty(params.modelApi ?? "") === "openai-completions" ||

632-

normalizeLowercaseStringOrEmpty(params.modelApi ?? "") === "anthropic-messages" ||

633-

normalizeLowercaseStringOrEmpty(params.modelApi ?? "") === "bedrock-converse-stream"

634-

) {

643+

if (RETRY_GUARD_MODEL_APIS.has(normalizeLowercaseStringOrEmpty(params.modelApi ?? ""))) {

635644

return true;

636645

}

637646

// Non-visible final turns are narrower than planning-only turns: there is no