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

D
Docker
B
Blog RSS Feed
Microsoft Security Blog
Microsoft Security Blog
Y
Y Combinator Blog
N
Netflix TechBlog - Medium
M
MIT News - Artificial intelligence
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
B
Blog
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
C
Check Point Blog
The GitHub Blog
The GitHub Blog
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
P
Proofpoint News Feed
Martin Fowler
Martin Fowler
大猫的无限游戏
大猫的无限游戏
GbyAI
GbyAI
博客园_首页
A
About on SuperTechFans
Blog — PlanetScale
Blog — PlanetScale
人人都是产品经理
人人都是产品经理
T
Tailwind CSS Blog
aimingoo的专栏
aimingoo的专栏
T
The Blog of Author Tim Ferriss
The Cloudflare Blog

Recent Commits to openclaw:main

test: merge chat side-result checks · openclaw/openclaw@ddd2c2a test: merge cron history checks · openclaw/openclaw@f7eb746 test: merge responsive navigation shell checks · openclaw/openclaw@c2e4b47 docs(changelog): add codex oauth fixes · openclaw/openclaw@628e6cd test: merge navigation routing cases · openclaw/openclaw@5d8cecb Tests: mock channel registry bundled fallback · openclaw/openclaw@2b08233 Secrets: avoid broad web search discovery for single plugin config · openclaw/openclaw@a464f59 test: merge config view browser checks · openclaw/openclaw@20cf511 fix(status): align oauth health with runtime · openclaw/openclaw@eed7116 feat: add macOS screen snapshots for monitor preview (#67954) thanks … · openclaw/openclaw@f377db1 fix: report shared auth scopes in hello-ok (#67810) thanks @BunsDev · openclaw/openclaw@0b6c39b Auto-reply: avoid eager bundled route fallback · openclaw/openclaw@3ea1bf4 Tests: narrow session binding contract setup · openclaw/openclaw@54e4e16 fix(macOS): enable undo/redo in webchat composer text input (#34962) · openclaw/openclaw@00951dc Tests: speed up channel setup promotion · openclaw/openclaw@82b529a Docs: refresh agent instructions · openclaw/openclaw@5775fe2 fix(auth): serialize OAuth refresh across agents to fix #26322 (#67876) · openclaw/openclaw@8e79080 test: allow ollama public surface boundary test · openclaw/openclaw@7d4f1a6 Docs: add test performance guardrails · openclaw/openclaw@89706d3 Tests: restore context-engine usage proof · openclaw/openclaw@e4c4f95 Tests: slim context engine runtime coverage · openclaw/openclaw@74c198f ci: retry failed custom checkouts · openclaw/openclaw@0ee5baf test: trim duplicate provider auth onboarding cases · openclaw/openclaw@1ffc02e matrix: fix sessions_spawn --thread subagent session spawning (#67643) · openclaw/openclaw@1ce2596 test: reduce auth choice fixture churn · openclaw/openclaw@857b9cd test: mock health status config boundaries · openclaw/openclaw@9d5ab4a test: mock onboard config io boundary · openclaw/openclaw@299694d test: mock legacy state plugin boundaries · openclaw/openclaw@2713089 test: mock channel install boundaries · openclaw/openclaw@b945248 test: mock doctor preview channel boundaries · openclaw/openclaw@b1a3ad4
fix(agents): detect incomplete tool-use turns with pre-to...
amknight · 2026-05-03 · via Recent Commits to openclaw:main

@@ -26,6 +26,7 @@ import {

2626

resolveEmptyResponseRetryInstruction,

2727

resolvePlanningOnlyRetryLimit,

2828

resolvePlanningOnlyRetryInstruction,

29+

isIncompleteTerminalAssistantTurn,

2930

resolveIncompleteTurnPayloadText,

3031

resolveReasoningOnlyRetryInstruction,

3132

STRICT_AGENTIC_BLOCKED_TEXT,

@@ -995,6 +996,136 @@ describe("runEmbeddedPiAgent incomplete-turn safety", () => {

995996

).toBe("abandoned");

996997

});

997998999+

it("flags tool-use stop reason as incomplete even when pre-tool text exists (#76477)", () => {

1000+

expect(

1001+

isIncompleteTerminalAssistantTurn({

1002+

hasAssistantVisibleText: true,

1003+

lastAssistant: { stopReason: "toolUse" },

1004+

}),

1005+

).toBe(true);

1006+

expect(

1007+

isIncompleteTerminalAssistantTurn({

1008+

hasAssistantVisibleText: false,

1009+

lastAssistant: { stopReason: "toolUse" },

1010+

}),

1011+

).toBe(true);

1012+

expect(

1013+

isIncompleteTerminalAssistantTurn({

1014+

hasAssistantVisibleText: true,

1015+

lastAssistant: { stopReason: "end_turn" },

1016+

}),

1017+

).toBe(false);

1018+

});

1019+1020+

it("detects tool-use terminal turn with pre-tool text as incomplete (#76477)", () => {

1021+

// When the last assistant message ended with stopReason=toolUse, pre-tool

1022+

// text alone must not suppress the incomplete-turn guard. The model

1023+

// expected to continue after tool results but the post-tool response was

1024+

// never produced.

1025+

const incompleteTurnText = resolveIncompleteTurnPayloadText({

1026+

payloadCount: 1,

1027+

aborted: false,

1028+

timedOut: false,

1029+

attempt: makeAttemptResult({

1030+

assistantTexts: ["Initial analysis of the codebase..."],

1031+

toolMetas: [{ toolName: "read", meta: "path=src/index.ts" }],

1032+

lastAssistant: {

1033+

role: "assistant",

1034+

stopReason: "toolUse",

1035+

provider: "anthropic",

1036+

model: "sonnet-4.6",

1037+

content: [

1038+

{ type: "text", text: "Initial analysis of the codebase..." },

1039+

{ type: "tool_use", id: "tool_1", name: "read", input: { path: "src/index.ts" } },

1040+

],

1041+

} as unknown as EmbeddedRunAttemptResult["lastAssistant"],

1042+

}),

1043+

});

1044+1045+

expect(incompleteTurnText).not.toBeNull();

1046+

expect(incompleteTurnText).toContain("couldn't generate a response");

1047+

});

1048+1049+

it("surfaces tool-use terminal with pre-tool text and side effects as replay-unsafe (#76477)", () => {

1050+

const incompleteTurnText = resolveIncompleteTurnPayloadText({

1051+

payloadCount: 1,

1052+

aborted: false,

1053+

timedOut: false,

1054+

attempt: makeAttemptResult({

1055+

assistantTexts: ["Let me update the file..."],

1056+

toolMetas: [{ toolName: "write" }],

1057+

lastAssistant: {

1058+

role: "assistant",

1059+

stopReason: "toolUse",

1060+

provider: "openai",

1061+

model: "gpt-5.4",

1062+

content: [

1063+

{ type: "text", text: "Let me update the file..." },

1064+

{ type: "tool_use", id: "tool_1", name: "write", input: {} },

1065+

],

1066+

} as unknown as EmbeddedRunAttemptResult["lastAssistant"],

1067+

}),

1068+

});

1069+1070+

expect(incompleteTurnText).toContain("verify before retrying");

1071+

});

1072+1073+

it("does not flag a completed tool-use turn with end_turn as incomplete (#76477)", () => {

1074+

// When the model successfully produces post-tool text, lastAssistant has

1075+

// stopReason=end_turn. The incomplete-turn guard should not fire.

1076+

const incompleteTurnText = resolveIncompleteTurnPayloadText({

1077+

payloadCount: 2,

1078+

aborted: false,

1079+

timedOut: false,

1080+

attempt: makeAttemptResult({

1081+

assistantTexts: ["Initial analysis...", "Here is the final answer."],

1082+

toolMetas: [{ toolName: "read" }],

1083+

lastAssistant: {

1084+

role: "assistant",

1085+

stopReason: "end_turn",

1086+

provider: "anthropic",

1087+

model: "sonnet-4.6",

1088+

content: [{ type: "text", text: "Here is the final answer." }],

1089+

} as unknown as EmbeddedRunAttemptResult["lastAssistant"],

1090+

}),

1091+

});

1092+1093+

expect(incompleteTurnText).toBeNull();

1094+

});

1095+1096+

it("surfaces an error for tool-use terminal turn with pre-tool text via runEmbeddedPiAgent (#76477)", async () => {

1097+

mockedClassifyFailoverReason.mockReturnValue(null);

1098+

mockedRunEmbeddedAttempt.mockResolvedValueOnce(

1099+

makeAttemptResult({

1100+

assistantTexts: ["Initial analysis of the issue..."],

1101+

toolMetas: [{ toolName: "read", meta: "path=src/index.ts" }],

1102+

lastAssistant: {

1103+

stopReason: "toolUse",

1104+

provider: "anthropic",

1105+

model: "sonnet-4.6",

1106+

content: [

1107+

{ type: "text", text: "Initial analysis of the issue..." },

1108+

{ type: "tool_use", id: "tool_1", name: "read", input: { path: "src/index.ts" } },

1109+

],

1110+

} as unknown as EmbeddedRunAttemptResult["lastAssistant"],

1111+

}),

1112+

);

1113+1114+

const result = await runEmbeddedPiAgent({

1115+

...overflowBaseRunParams,

1116+

provider: "anthropic",

1117+

model: "sonnet-4.6",

1118+

runId: "run-tool-use-dropped-final-text",

1119+

});

1120+1121+

expect(mockedRunEmbeddedAttempt).toHaveBeenCalledTimes(1);

1122+

expect(result.payloads?.[0]?.isError).toBe(true);

1123+

expect(result.payloads?.[0]?.text).toContain("couldn't generate a response");

1124+

expect(mockedLog.warn).toHaveBeenCalledWith(

1125+

expect.stringContaining("incomplete turn detected"),

1126+

);

1127+

});

1128+9981129

it("treats missing replay metadata as replay-invalid", () => {

9991130

const attempt = makeAttemptResult();

10001131

delete (attempt as Partial<EmbeddedRunAttemptResult>).replayMetadata;