chore: preserve sealed model wait evidence bytes

This commit is contained in:
Agent Workbench Dev
2026-11-01 21:25:15 +08:00
parent 9df182cc8f
commit 017923aace
20 changed files with 266 additions and 261 deletions
+3
View File
@@ -60,3 +60,6 @@ docs/engineering-evals-*-evidence/*.log binary
docs/engineering-evals-*-evidence/**/*.blob binary
docs/engineering-evals-*-evidence/*.blob binary
docs/engineering-evals-*.json -text whitespace=blank-at-eol,blank-at-eof,space-before-tab,cr-at-eol
docs/model-wait-release-evidence/** -text whitespace=blank-at-eol,blank-at-eof,space-before-tab,cr-at-eol
docs/model-wait-release-identity.json -text whitespace=blank-at-eol,blank-at-eof,space-before-tab,cr-at-eol
@@ -1 +1 @@
{"success":true,"inputs":354,"compiled":22,"portableRoot":"D:\\Ai-project\\REA逆向\\Agent-Workbench\\releases\\Agent-Workbench-1.20.2-win-x64-20261101124548744"}
{"success":true,"inputs":354,"compiled":22,"portableRoot":"D:\\Ai-project\\REA逆向\\Agent-Workbench\\releases\\Agent-Workbench-1.20.2-win-x64-20261101124548744"}
@@ -1,20 +1,20 @@
> local-agent-workbench@1.20.2 build
> tsc --noEmit && vite build && node scripts/build-mobile.mjs && node scripts/build-server.mjs
vite v7.3.7 building client environment for production...
transforming...
✓ 1732 modules transformed.
rendering chunks...
computing gzip size...
dist-client/.vite/manifest.json 0.55 kB │ gzip: 0.22 kB
dist-client/index.html 0.58 kB │ gzip: 0.39 kB
dist-client/mobile.html 0.82 kB │ gzip: 0.46 kB
dist-client/assets/mobile-CNDnI_ru.css 11.92 kB │ gzip: 3.12 kB
dist-client/assets/desktop-B65VXbkr.css 147.58 kB │ gzip: 27.99 kB
dist-client/assets/mobile-DVWh54VU.js 24.31 kB │ gzip: 8.65 kB
dist-client/assets/x-Cc3T3VuL.js 225.51 kB │ gzip: 70.77 kB
dist-client/assets/desktop-DakkeVg6.js 322.94 kB │ gzip: 91.46 kB
✓ built in 1.17s
Mobile shell built: 3 allowed assets; API and project data are never cached.
Server bundle built.
> local-agent-workbench@1.20.2 build
> tsc --noEmit && vite build && node scripts/build-mobile.mjs && node scripts/build-server.mjs
vite v7.3.7 building client environment for production...
transforming...
✓ 1732 modules transformed.
rendering chunks...
computing gzip size...
dist-client/.vite/manifest.json 0.55 kB │ gzip: 0.22 kB
dist-client/index.html 0.58 kB │ gzip: 0.39 kB
dist-client/mobile.html 0.82 kB │ gzip: 0.46 kB
dist-client/assets/mobile-CNDnI_ru.css 11.92 kB │ gzip: 3.12 kB
dist-client/assets/desktop-B65VXbkr.css 147.58 kB │ gzip: 27.99 kB
dist-client/assets/mobile-DVWh54VU.js 24.31 kB │ gzip: 8.65 kB
dist-client/assets/x-Cc3T3VuL.js 225.51 kB │ gzip: 70.77 kB
dist-client/assets/desktop-DakkeVg6.js 322.94 kB │ gzip: 91.46 kB
✓ built in 1.17s
Mobile shell built: 3 allowed assets; API and project data are never cached.
Server bundle built.
@@ -1,6 +1,6 @@
PASS desktop saves seconds, preserves original pool/member and sends no model request
PASS invalid timeout rejected without saving
PASS saved values survive reload; narrow screen remains usable
PASS active request locks wait changes in both API and UI
PASS actual stalled stream shows exact idle reason, saves unknown use, never writes or retries
{"success":true,"directory":"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\model-wait-ui-a700392c-c9d3-44a5-868a-976d0e5d8d7a","checks":5,"screenshots":4}
PASS desktop saves seconds, preserves original pool/member and sends no model request
PASS invalid timeout rejected without saving
PASS saved values survive reload; narrow screen remains usable
PASS active request locks wait changes in both API and UI
PASS actual stalled stream shows exact idle reason, saves unknown use, never writes or retries
{"success":true,"directory":"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\model-wait-ui-a700392c-c9d3-44a5-868a-976d0e5d8d7a","checks":5,"screenshots":4}
@@ -1 +1 @@
{"success":false,"phase":"stage","output":".qa/model-wait-formal-b-stage.json","error":"AssertionError [ERR_ASSERTION]: sessionsSha256\n+ actual - expected\n\n+ '0ceeaa1dfd2efd10dc7e3536db689d3b17c88c4293dca4cecef118aca684bf1d'\n- '206f79d44152116ff4be90a12cb7f84ec6f8543fa854964e2763ee3f55c654f3'\n\n at compareOriginalState (file:///D:/Ai-project/REA%E9%80%86%E5%90%91/Agent-Workbench/.qa/desktop-f-state.mjs:47:116)\n at compareWaitState (file:///D:/Ai-project/REA%E9%80%86%E5%90%91/Agent-Workbench/.qa/model-wait-state.mjs:10:15)\n at file:///D:/Ai-project/REA%E9%80%86%E5%90%91/Agent-Workbench/.qa/model-wait-formal.mjs:44:115\n at process.processTicksAndRejections (node:internal/process/task_queues:104:5)"}
{"success":false,"phase":"stage","output":".qa/model-wait-formal-b-stage.json","error":"AssertionError [ERR_ASSERTION]: sessionsSha256\n+ actual - expected\n\n+ '0ceeaa1dfd2efd10dc7e3536db689d3b17c88c4293dca4cecef118aca684bf1d'\n- '206f79d44152116ff4be90a12cb7f84ec6f8543fa854964e2763ee3f55c654f3'\n\n at compareOriginalState (file:///D:/Ai-project/REA%E9%80%86%E5%90%91/Agent-Workbench/.qa/desktop-f-state.mjs:47:116)\n at compareWaitState (file:///D:/Ai-project/REA%E9%80%86%E5%90%91/Agent-Workbench/.qa/model-wait-state.mjs:10:15)\n at file:///D:/Ai-project/REA%E9%80%86%E5%90%91/Agent-Workbench/.qa/model-wait-formal.mjs:44:115\n at process.processTicksAndRejections (node:internal/process/task_queues:104:5)"}
@@ -1 +1 @@
{"success":true,"phase":"staged-resume","eventCursor":6138,"records":3207}
{"success":true,"phase":"staged-resume","eventCursor":6138,"records":3207}
@@ -1 +1 @@
{"success":true,"phase":"upgraded","output":".qa/model-wait-formal-c-upgrade.json"}
{"success":true,"phase":"upgraded","output":".qa/model-wait-formal-c-upgrade.json"}
@@ -1 +1 @@
{"success":false,"phase":"stage","output":".qa/model-wait-formal-stage.json","error":"AssertionError [ERR_ASSERTION]: Expected values to be strictly deep-equal:\n+ actual - expected\n\n+ undefined\n- {\n- modelResponseTimeoutMs: 300000,\n- modelStreamIdleTimeoutMs: 120000\n- }\n\n at compareWaitState (file:///D:/Ai-project/REA%E9%80%86%E5%90%91/Agent-Workbench/.qa/model-wait-state.mjs:7:12)\n at file:///D:/Ai-project/REA%E9%80%86%E5%90%91/Agent-Workbench/.qa/model-wait-formal.mjs:41:5\n at process.processTicksAndRejections (node:internal/process/task_queues:104:5)"}
{"success":false,"phase":"stage","output":".qa/model-wait-formal-stage.json","error":"AssertionError [ERR_ASSERTION]: Expected values to be strictly deep-equal:\n+ actual - expected\n\n+ undefined\n- {\n- modelResponseTimeoutMs: 300000,\n- modelStreamIdleTimeoutMs: 120000\n- }\n\n at compareWaitState (file:///D:/Ai-project/REA%E9%80%86%E5%90%91/Agent-Workbench/.qa/model-wait-state.mjs:7:12)\n at file:///D:/Ai-project/REA%E9%80%86%E5%90%91/Agent-Workbench/.qa/model-wait-formal.mjs:41:5\n at process.processTicksAndRejections (node:internal/process/task_queues:104:5)"}
@@ -1 +1 @@
{"success":false,"results":2,"screenshots":4,"error":"page.screenshot: Timeout 30000ms exceeded.\nCall log:\n\u001b[2m - taking page screenshot\u001b[22m\n\u001b[2m - waiting for fonts to load...\u001b[22m\n\u001b[2m - fonts loaded\u001b[22m\n\n at capture (D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\model-wait-formal-ui.mjs:28:64)\n at inspect (D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\model-wait-formal-ui.mjs:33:9)\n at async file:///D:/Ai-project/REA%E9%80%86%E5%90%91/Agent-Workbench/.qa/model-wait-formal-ui.mjs:61:109"}
{"success":false,"results":2,"screenshots":4,"error":"page.screenshot: Timeout 30000ms exceeded.\nCall log:\n\u001b[2m - taking page screenshot\u001b[22m\n\u001b[2m - waiting for fonts to load...\u001b[22m\n\u001b[2m - fonts loaded\u001b[22m\n\n at capture (D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\model-wait-formal-ui.mjs:28:64)\n at inspect (D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\model-wait-formal-ui.mjs:33:9)\n at async file:///D:/Ai-project/REA%E9%80%86%E5%90%91/Agent-Workbench/.qa/model-wait-formal-ui.mjs:61:109"}
@@ -1 +1 @@
{"success":true,"results":4,"screenshots":8}
{"success":true,"results":4,"screenshots":8}
@@ -1,37 +1,37 @@
✔ HTTP download preserves original malformed SSE bytes and artifact SHA instead of decoding invalid UTF8 (2426.5386ms)
✔ terminated SSE adapts empty optional tool arguments, retains the exact wire bytes and executes the validated batch once (2791.2168ms)
✔ reported SSE usage survives rejected required arguments "" without dispatching any tool (1584.8417ms)
✔ reported SSE usage survives rejected required arguments "{\"path\":" without dispatching any tool (1354.9623ms)
✔ reported SSE usage survives rejected required arguments "null" without dispatching any tool (1377.0252ms)
✔ a stream without DONE retains reported usage but cannot adapt or execute even an optional tool (1520.6525ms)
✔ a timeout after reasoning-only fragments creates one unknown receipt and never invents reported tokens (1597.1202ms)
✔ real Chat SSE carries UTF8 and split tool JSON; Runtime executes only after DONE, stores partial progress and original usage (2330.514ms)
✔ Responses typed SSE validates original item identity, terminal data, raw usage and tool result pair (2277.3525ms)
✔ incomplete network stream retains unknown attempt and never automatically switches pool or dispatches fragment tools (1590.889ms)
✔ stopping an active HTTP stream closes only original request and keeps partial text without executing tools (1390.7969ms)
✔ explicit streaming probe uses actual SSE despite unknown declaration and binds its own original attempt (700.3784ms)
✔ coalesced initial SSE events publish the latest partial state even when the upstream stays idle (1398.8886ms)
✔ stream timeout closes the original HTTP response, preserves redacted capture and never retries (988.5203ms)
✔ stream idle timeout preserves incomplete tool arguments, records unknown use and never fails over (1096.3096ms)
✔ SSE accepts CR-only final event boundaries and retains exact invalid UTF8 bytes without dispatch (0.7592ms)
▶ stream parser rejects truncation, finish length, malformed JSON, duplicate sequence, changed final item and oversized data
✔ reject 0 (0.1688ms)
✔ reject 1 (0.1502ms)
✔ reject 2 (0.1392ms)
✔ reject 3 (4.911ms)
✔ reject 4 (0.136ms)
✔ reject 5 (0.1465ms)
✔ stream parser rejects truncation, finish length, malformed JSON, duplicate sequence, changed final item and oversized data (6.0882ms)
✔ older configurations receive bounded wait defaults without changing the original object (1.0865ms)
✔ active stream survives sixty seconds; byte activity resets idle but cannot extend total deadline (1.0491ms)
✔ silent stream has its own accurate deadline; caller cancellation and cleanup preserve their cause (1.3579ms)
✔ waiting for headers is bounded independently of stream-idle time (0.2424ms)
✔ wait settings persist, pin new runs and leave model pool, credentials and old run snapshots intact (1271.6174ms)
ℹ tests 28
ℹ suites 0
ℹ pass 28
ℹ fail 0
ℹ cancelled 0
ℹ skipped 0
ℹ todo 0
ℹ duration_ms 24652.2961
✔ HTTP download preserves original malformed SSE bytes and artifact SHA instead of decoding invalid UTF8 (2426.5386ms)
✔ terminated SSE adapts empty optional tool arguments, retains the exact wire bytes and executes the validated batch once (2791.2168ms)
✔ reported SSE usage survives rejected required arguments "" without dispatching any tool (1584.8417ms)
✔ reported SSE usage survives rejected required arguments "{\"path\":" without dispatching any tool (1354.9623ms)
✔ reported SSE usage survives rejected required arguments "null" without dispatching any tool (1377.0252ms)
✔ a stream without DONE retains reported usage but cannot adapt or execute even an optional tool (1520.6525ms)
✔ a timeout after reasoning-only fragments creates one unknown receipt and never invents reported tokens (1597.1202ms)
✔ real Chat SSE carries UTF8 and split tool JSON; Runtime executes only after DONE, stores partial progress and original usage (2330.514ms)
✔ Responses typed SSE validates original item identity, terminal data, raw usage and tool result pair (2277.3525ms)
✔ incomplete network stream retains unknown attempt and never automatically switches pool or dispatches fragment tools (1590.889ms)
✔ stopping an active HTTP stream closes only original request and keeps partial text without executing tools (1390.7969ms)
✔ explicit streaming probe uses actual SSE despite unknown declaration and binds its own original attempt (700.3784ms)
✔ coalesced initial SSE events publish the latest partial state even when the upstream stays idle (1398.8886ms)
✔ stream timeout closes the original HTTP response, preserves redacted capture and never retries (988.5203ms)
✔ stream idle timeout preserves incomplete tool arguments, records unknown use and never fails over (1096.3096ms)
✔ SSE accepts CR-only final event boundaries and retains exact invalid UTF8 bytes without dispatch (0.7592ms)
▶ stream parser rejects truncation, finish length, malformed JSON, duplicate sequence, changed final item and oversized data
✔ reject 0 (0.1688ms)
✔ reject 1 (0.1502ms)
✔ reject 2 (0.1392ms)
✔ reject 3 (4.911ms)
✔ reject 4 (0.136ms)
✔ reject 5 (0.1465ms)
✔ stream parser rejects truncation, finish length, malformed JSON, duplicate sequence, changed final item and oversized data (6.0882ms)
✔ older configurations receive bounded wait defaults without changing the original object (1.0865ms)
✔ active stream survives sixty seconds; byte activity resets idle but cannot extend total deadline (1.0491ms)
✔ silent stream has its own accurate deadline; caller cancellation and cleanup preserve their cause (1.3579ms)
✔ waiting for headers is bounded independently of stream-idle time (0.2424ms)
✔ wait settings persist, pin new runs and leave model pool, credentials and old run snapshots intact (1271.6174ms)
ℹ tests 28
ℹ suites 0
ℹ pass 28
ℹ fail 0
ℹ cancelled 0
ℹ skipped 0
ℹ todo 0
ℹ duration_ms 24652.2961
@@ -1 +1 @@
{"success":true,"results":[{"streaming":false,"durationMs":65034,"finishReason":"stop","content":"non-stream completed"},{"streaming":true,"durationMs":65037,"finishReason":"stop","content":"开始完成"}],"requests":2,"realModelRequests":0}
{"success":true,"results":[{"streaming":false,"durationMs":65034,"finishReason":"stop","content":"non-stream completed"},{"streaming":true,"durationMs":65037,"finishReason":"stop","content":"开始完成"}],"requests":2,"realModelRequests":0}
@@ -1 +1 @@
{"success":true,"retained":{"originalRows":"exact","originalExports":"exact","originalArtifactBytes":"exact","originalModelCenter":"exact","publicSchemaAddition":"/waitSettings:300000/120000; original disk config unchanged"}}
{"success":true,"retained":{"originalRows":"exact","originalExports":"exact","originalArtifactBytes":"exact","originalModelCenter":"exact","publicSchemaAddition":"/waitSettings:300000/120000; original disk config unchanged"}}
@@ -1,70 +1,70 @@
{
"portableRoot": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\releases\\Agent-Workbench-1.20.2-win-x64-20261101124548744",
"sourceRoot": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\releases\\Agent-Workbench-1.20.2-source-20261101124548744",
"portable": {
"path": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\releases\\Agent-Workbench-1.20.2-win-x64-20261101124548744.zip",
"bytes": 627026168,
"sha256": "a5b47f3c4a02ca5794c24321b89e773a7fdd161063cee4c680a82ece657210f8"
},
"source": {
"path": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\releases\\Agent-Workbench-1.20.2-source-20261101124548744.zip",
"bytes": 627194835,
"sha256": "ec2eea17c90b38fc65a095c5b5eb25ff46a27bdf111e9808961fbf6a20165656"
},
"installer": {
"path": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\releases\\Agent-Workbench-1.20.2-win-x64-20261101124548744-Setup.exe",
"bytes": 586191896,
"sha256": "ca278ea713f61832fce93dbbb56e07e30b285850b0593a6741e78a53cfe07240",
"compiler": {
"version": "6.3.1",
"sha256": "3b22d36fc7e7c421d8caf780008ecdb42b14b3f0371b9bafe3cad64e6648d9cf"
},
"source": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\installer-build-3f1750c0-e122-4b94-9de0-56d8c961c654\\installer.iss",
"version": "1.20.2",
"candidateDirectory": "Agent-Workbench-1.20.2-win-x64-20261101124548744",
"qaFixture": null
},
"desktopRuntime": {
"version": "45.2.0",
"sha256": "010ea51bfb528c8143513439aa93fb40754403daa2023f9decd4517da71706ab",
"bytes": 249810432,
"executable": "desktop/AgentWorkbench.exe",
"scope": "Fixed original Windows Electron runtime, full license texts, isolated application loader and native bootstrap; original executable hash unchanged."
},
"nativeLauncher": {
"bytes": 12288,
"sha256": "e3d31cb7ce5009d3ca5a96d04dc16b1add0bd1f5cde7e1304461de6713d37463",
"inputs": {
"compiler": "46809206887326d2d24db1eff1f3064de972c3451abe766b49111450a5e08e00",
"source": "1c8e7d7ee5c383897bec48e9aadb8051b47c09686f74bbf34e0b6b695cd01ae5",
"icon": "f1879074e45f01c480ad9bf50b818637cc60ad56526056c886f808a4812d7b0c"
},
"compiler": "Windows .NET Framework C# compiler; source-bound cached original output, no deterministic-build claim."
},
"nodeVersion": "v24.22.1",
"reaRuntime": {
"version": "6.8.0",
"protocolVersion": "2025-03-26",
"packages": 149,
"files": 8413,
"bytes": 142395402,
"lockSha256": "687fb3fa3089a12f239dd5e87d9fa07df55a87df4935931d495a92600a1b0f6e"
},
"portableFiles": 19560,
"licenseTexts": 135,
"browserRuntime": {
"playwrightVersion": "1.64.1",
"packages": [
{
"name": "playwright",
"version": "1.64.1",
"files": 60
},
{
"name": "playwright-core",
"version": "1.64.1",
"files": 115
}
]
}
}
{
"portableRoot": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\releases\\Agent-Workbench-1.20.2-win-x64-20261101124548744",
"sourceRoot": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\releases\\Agent-Workbench-1.20.2-source-20261101124548744",
"portable": {
"path": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\releases\\Agent-Workbench-1.20.2-win-x64-20261101124548744.zip",
"bytes": 627026168,
"sha256": "a5b47f3c4a02ca5794c24321b89e773a7fdd161063cee4c680a82ece657210f8"
},
"source": {
"path": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\releases\\Agent-Workbench-1.20.2-source-20261101124548744.zip",
"bytes": 627194835,
"sha256": "ec2eea17c90b38fc65a095c5b5eb25ff46a27bdf111e9808961fbf6a20165656"
},
"installer": {
"path": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\releases\\Agent-Workbench-1.20.2-win-x64-20261101124548744-Setup.exe",
"bytes": 586191896,
"sha256": "ca278ea713f61832fce93dbbb56e07e30b285850b0593a6741e78a53cfe07240",
"compiler": {
"version": "6.3.1",
"sha256": "3b22d36fc7e7c421d8caf780008ecdb42b14b3f0371b9bafe3cad64e6648d9cf"
},
"source": "D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\installer-build-3f1750c0-e122-4b94-9de0-56d8c961c654\\installer.iss",
"version": "1.20.2",
"candidateDirectory": "Agent-Workbench-1.20.2-win-x64-20261101124548744",
"qaFixture": null
},
"desktopRuntime": {
"version": "45.2.0",
"sha256": "010ea51bfb528c8143513439aa93fb40754403daa2023f9decd4517da71706ab",
"bytes": 249810432,
"executable": "desktop/AgentWorkbench.exe",
"scope": "Fixed original Windows Electron runtime, full license texts, isolated application loader and native bootstrap; original executable hash unchanged."
},
"nativeLauncher": {
"bytes": 12288,
"sha256": "e3d31cb7ce5009d3ca5a96d04dc16b1add0bd1f5cde7e1304461de6713d37463",
"inputs": {
"compiler": "46809206887326d2d24db1eff1f3064de972c3451abe766b49111450a5e08e00",
"source": "1c8e7d7ee5c383897bec48e9aadb8051b47c09686f74bbf34e0b6b695cd01ae5",
"icon": "f1879074e45f01c480ad9bf50b818637cc60ad56526056c886f808a4812d7b0c"
},
"compiler": "Windows .NET Framework C# compiler; source-bound cached original output, no deterministic-build claim."
},
"nodeVersion": "v24.22.1",
"reaRuntime": {
"version": "6.8.0",
"protocolVersion": "2025-03-26",
"packages": 149,
"files": 8413,
"bytes": 142395402,
"lockSha256": "687fb3fa3089a12f239dd5e87d9fa07df55a87df4935931d495a92600a1b0f6e"
},
"portableFiles": 19560,
"licenseTexts": 135,
"browserRuntime": {
"playwrightVersion": "1.64.1",
"packages": [
{
"name": "playwright",
"version": "1.64.1",
"files": 60
},
{
"name": "playwright-core",
"version": "1.64.1",
"files": 115
}
]
}
}
@@ -1,6 +1,6 @@
PASS actual form saves Responses and declared stream, refresh preserves both and narrow editor remains operable
PASS separate text, nonce tool and actual SSE probes bind only their requested capability and own attempts
PASS actual pooled task excludes text-only before dispatch, executes original tool once and downloads SHA-matched original SSE
PASS desktop and actual paired gateway show same bounded stream; mobile stops original HTTP with zero fragment tool dispatch or replay
PASS truncated output keeps explicit unknown ledger, partial evidence and no automatic failover or incomplete tool execution
EVIDENCE D:\Ai-project\REA逆向\Agent-Workbench\.qa\model-protocols-ui-12baf4c3-6ef1-4afe-b4fc-dd990e44223a
PASS actual form saves Responses and declared stream, refresh preserves both and narrow editor remains operable
PASS separate text, nonce tool and actual SSE probes bind only their requested capability and own attempts
PASS actual pooled task excludes text-only before dispatch, executes original tool once and downloads SHA-matched original SSE
PASS desktop and actual paired gateway show same bounded stream; mobile stops original HTTP with zero fragment tool dispatch or replay
PASS truncated output keeps explicit unknown ledger, partial evidence and no automatic failover or incomplete tool execution
EVIDENCE D:\Ai-project\REA逆向\Agent-Workbench\.qa\model-protocols-ui-12baf4c3-6ef1-4afe-b4fc-dd990e44223a
@@ -1,70 +1,70 @@
✔ a question about the previous workbench error receives local error context and current change states (2572.0986ms)
✔ a transient model failure after an approved command retries only the model request (3613.6709ms)
✔ exhausted model retries preserve tool results, and explicit resume adds no new user turn or replay (3626.3978ms)
✔ stopping during model backoff cancels further requests and clears the running request state (1781.9182ms)
✔ HTTP authentication errors are classified, redacted and never automatically retried (21.0006ms)
✔ temporary HTTP service errors retry and recover with recorded attempt counts (20.4214ms)
✔ invalid and oversized protocol responses fail once with a useful response-format diagnosis (3.9159ms)
✔ HTTP rate limits expose their source and obey the service retry interval before recovery (1016.1704ms)
✔ Windows commands preserve Chinese through a PowerShell pipeline into Node (1072.6146ms)
✔ Windows shell syntax errors stay readable and native command exit codes are retained (1457.3246ms)
✔ review and completed states do not turn a generic continue click into a new model request (888.6061ms)
✔ old session data receives a truthful historical connection diagnosis without rewriting history (564.2336ms)
✔ model registry keeps original credentials/endpoint intact, rejects stale edits, and forgets only volatile keys on restart (1470.5084ms)
✔ config import previews exact revisions, keeps secrets volatile, syncs only when enabled and preserves manual edits on conflicts (1171.4104ms)
✔ real HTTP pool fails over 429 to another model and never repeats completed tools; attempts seal actual request bytes and Run identity (4000.4454ms)
✔ pool rotation and weights select members on new Runs; disabled and missing credentials are excluded without requests (5179.6987ms)
✔ 401 blocks the credential revision; updating the key restores availability without leaking error echoes (1754.9103ms)
✔ queued request cancellation and in-flight disable release only owned leases and do not dispatch another HTTP request (1277.136ms)
✔ compiled model-center HTTP API persists public profiles, blocks forged origins, and local operations never invoke a model (2529.498ms)
✔ model registry disk failures restore routing, pool profiles and credential references without partial updates (1011.5669ms)
✔ model selection and saved configuration commit together and failures preserve the original mode and endpoint (1006.0569ms)
✔ configuration imports roll back the entire batch and retain one usable preview after validation or commit failure (674.2603ms)
✔ saved engine budgets verify all Chat/Responses route variants and reject a mismatched output allowance (1090.6984ms)
✔ owned CC Switch SQLite imports the selected TOML Responses wire_api and leaves unknown clients untouched (1548.632ms)
✔ Responses executes one original tool through Runtime, carries encrypted Items, reports raw usage and seals exact wire bytes (2755.3635ms)
✔ mixed-protocol pool keeps both actual variants and changes route on 429 without re-executing completed tools (3770.6098ms)
▶ Responses incomplete, failed, missing call identity and unsupported hosted outputs never dispatch a tool
✔ Reject {"body":{"id":"resp_owned","status":"incomplete","output":[{"type":"function_call","id":"fc_partial" (1949.4119ms)
✔ Reject {"body":{"id":"resp_owned","status":"failed","output":[{"type":"function_call","id":"fc_failed","cal (2082.742ms)
✔ Reject {"body":{"id":"resp_owned","status":"completed","output":[{"type":"function_call","id":"fc_bad","cal (2016.589ms)
✔ Reject {"body":{"id":"resp_owned","status":"completed","output":[{"type":"computer_call","id":"not-supporte (1919.9253ms)
✔ Responses incomplete, failed, missing call identity and unsupported hosted outputs never dispatch a tool (7969.4088ms)
▶ invalid JSON in a nonstream Chat or Responses tool group prevents every tool and pool replay
✔ chat-completions (1824.526ms)
✔ responses (2252.4703ms)
✔ invalid JSON in a nonstream Chat or Responses tool group prevents every tool and pool replay (4077.3534ms)
✔ wire adapter preserves tool IDs but never passes Responses-private Items to Chat or another endpoint (0.1564ms)
✔ connection JSON import accepts explicit Responses protocol and continues to reject unknown protocols without reading client sessions (862.7606ms)
✔ capability routing skips text-only members before acquiring a lease and records exact refusal in the original Run (2017.5982ms)
✔ all incompatible members reject with no network request, no lease, preserved Task and explicit non-dispatched ledger (1570.3869ms)
✔ explicit synthetic tool probe verifies only actual matching call, never executes it and loses verification on revision changes (1059.7389ms)
✔ plain text probe cannot verify tools; wrong tool nonce and client supplied evidence cannot claim verified support (1192.9938ms)
✔ a disabled connection invalidates an in-flight probe even when toggle mutates its original registry object (1022.9602ms)
✔ a stop from the event loop interrupts immediate model replies before later tools (1620.311ms)
✔ a stop after the first tool preserves a complete interrupted multi-call protocol group (2388.9177ms)
✔ the compatible HTTP loop preserves call IDs, validates errors and waits before executing commands (6014.3933ms)
✔ denial prevents command execution and produces a paired denied result (2269.3286ms)
✔ stopping during approval interrupts pending calls without executing or replaying them (2111.8268ms)
✔ stopping a running fixture command terminates its own process tree and retains uncertain state (2361.3446ms)
✔ command output decodes split UTF-8 and excludes the workbench key from child environment (1288.7983ms)
✔ model-list testing uses a tentative key without changing the saved connection and redacts service errors (12.439ms)
✔ truncated or repeated model calls fail before execution; tool input validation rejects unknown capabilities (4150.4139ms)
✔ request projection keeps exact tool pairs and complete recent user instructions while preserving disk history (1713.3195ms)
✔ operational questions are diagnostic while explicit repair requests remain execution instructions (1.0512ms)
✔ file delivery requires a positive instruction, not saved evidence or a prohibited write (1.2058ms)
✔ an unsubstantiated text claim cannot complete a requested file delivery (1601.8128ms)
✔ HTML without tests ends at reviewed file delivery and never offers automatic rechecking (2770.1918ms)
✔ diagnostic answers retain the unfinished task, limits and check result, and enforce read-only tools (4569.8906ms)
✔ review continuation is a system action and adds no testing demand or fake user instruction (3182.5911ms)
✔ a provider cannot execute tests against an explicit no-test task even if it ignores advertised tools (1482.0105ms)
✔ a budget pause becomes visible only after its input snapshot can be sealed with the original Run (1270.4707ms)
✔ alternating duplicate file reads and listing files still pauses without treating new pages as duplicates (11266.4412ms)
ℹ tests 60
ℹ suites 0
ℹ pass 60
ℹ fail 0
ℹ cancelled 0
ℹ skipped 0
ℹ todo 0
ℹ duration_ms 65317.1144
✔ a question about the previous workbench error receives local error context and current change states (2572.0986ms)
✔ a transient model failure after an approved command retries only the model request (3613.6709ms)
✔ exhausted model retries preserve tool results, and explicit resume adds no new user turn or replay (3626.3978ms)
✔ stopping during model backoff cancels further requests and clears the running request state (1781.9182ms)
✔ HTTP authentication errors are classified, redacted and never automatically retried (21.0006ms)
✔ temporary HTTP service errors retry and recover with recorded attempt counts (20.4214ms)
✔ invalid and oversized protocol responses fail once with a useful response-format diagnosis (3.9159ms)
✔ HTTP rate limits expose their source and obey the service retry interval before recovery (1016.1704ms)
✔ Windows commands preserve Chinese through a PowerShell pipeline into Node (1072.6146ms)
✔ Windows shell syntax errors stay readable and native command exit codes are retained (1457.3246ms)
✔ review and completed states do not turn a generic continue click into a new model request (888.6061ms)
✔ old session data receives a truthful historical connection diagnosis without rewriting history (564.2336ms)
✔ model registry keeps original credentials/endpoint intact, rejects stale edits, and forgets only volatile keys on restart (1470.5084ms)
✔ config import previews exact revisions, keeps secrets volatile, syncs only when enabled and preserves manual edits on conflicts (1171.4104ms)
✔ real HTTP pool fails over 429 to another model and never repeats completed tools; attempts seal actual request bytes and Run identity (4000.4454ms)
✔ pool rotation and weights select members on new Runs; disabled and missing credentials are excluded without requests (5179.6987ms)
✔ 401 blocks the credential revision; updating the key restores availability without leaking error echoes (1754.9103ms)
✔ queued request cancellation and in-flight disable release only owned leases and do not dispatch another HTTP request (1277.136ms)
✔ compiled model-center HTTP API persists public profiles, blocks forged origins, and local operations never invoke a model (2529.498ms)
✔ model registry disk failures restore routing, pool profiles and credential references without partial updates (1011.5669ms)
✔ model selection and saved configuration commit together and failures preserve the original mode and endpoint (1006.0569ms)
✔ configuration imports roll back the entire batch and retain one usable preview after validation or commit failure (674.2603ms)
✔ saved engine budgets verify all Chat/Responses route variants and reject a mismatched output allowance (1090.6984ms)
✔ owned CC Switch SQLite imports the selected TOML Responses wire_api and leaves unknown clients untouched (1548.632ms)
✔ Responses executes one original tool through Runtime, carries encrypted Items, reports raw usage and seals exact wire bytes (2755.3635ms)
✔ mixed-protocol pool keeps both actual variants and changes route on 429 without re-executing completed tools (3770.6098ms)
▶ Responses incomplete, failed, missing call identity and unsupported hosted outputs never dispatch a tool
✔ Reject {"body":{"id":"resp_owned","status":"incomplete","output":[{"type":"function_call","id":"fc_partial" (1949.4119ms)
✔ Reject {"body":{"id":"resp_owned","status":"failed","output":[{"type":"function_call","id":"fc_failed","cal (2082.742ms)
✔ Reject {"body":{"id":"resp_owned","status":"completed","output":[{"type":"function_call","id":"fc_bad","cal (2016.589ms)
✔ Reject {"body":{"id":"resp_owned","status":"completed","output":[{"type":"computer_call","id":"not-supporte (1919.9253ms)
✔ Responses incomplete, failed, missing call identity and unsupported hosted outputs never dispatch a tool (7969.4088ms)
▶ invalid JSON in a nonstream Chat or Responses tool group prevents every tool and pool replay
✔ chat-completions (1824.526ms)
✔ responses (2252.4703ms)
✔ invalid JSON in a nonstream Chat or Responses tool group prevents every tool and pool replay (4077.3534ms)
✔ wire adapter preserves tool IDs but never passes Responses-private Items to Chat or another endpoint (0.1564ms)
✔ connection JSON import accepts explicit Responses protocol and continues to reject unknown protocols without reading client sessions (862.7606ms)
✔ capability routing skips text-only members before acquiring a lease and records exact refusal in the original Run (2017.5982ms)
✔ all incompatible members reject with no network request, no lease, preserved Task and explicit non-dispatched ledger (1570.3869ms)
✔ explicit synthetic tool probe verifies only actual matching call, never executes it and loses verification on revision changes (1059.7389ms)
✔ plain text probe cannot verify tools; wrong tool nonce and client supplied evidence cannot claim verified support (1192.9938ms)
✔ a disabled connection invalidates an in-flight probe even when toggle mutates its original registry object (1022.9602ms)
✔ a stop from the event loop interrupts immediate model replies before later tools (1620.311ms)
✔ a stop after the first tool preserves a complete interrupted multi-call protocol group (2388.9177ms)
✔ the compatible HTTP loop preserves call IDs, validates errors and waits before executing commands (6014.3933ms)
✔ denial prevents command execution and produces a paired denied result (2269.3286ms)
✔ stopping during approval interrupts pending calls without executing or replaying them (2111.8268ms)
✔ stopping a running fixture command terminates its own process tree and retains uncertain state (2361.3446ms)
✔ command output decodes split UTF-8 and excludes the workbench key from child environment (1288.7983ms)
✔ model-list testing uses a tentative key without changing the saved connection and redacts service errors (12.439ms)
✔ truncated or repeated model calls fail before execution; tool input validation rejects unknown capabilities (4150.4139ms)
✔ request projection keeps exact tool pairs and complete recent user instructions while preserving disk history (1713.3195ms)
✔ operational questions are diagnostic while explicit repair requests remain execution instructions (1.0512ms)
✔ file delivery requires a positive instruction, not saved evidence or a prohibited write (1.2058ms)
✔ an unsubstantiated text claim cannot complete a requested file delivery (1601.8128ms)
✔ HTML without tests ends at reviewed file delivery and never offers automatic rechecking (2770.1918ms)
✔ diagnostic answers retain the unfinished task, limits and check result, and enforce read-only tools (4569.8906ms)
✔ review continuation is a system action and adds no testing demand or fake user instruction (3182.5911ms)
✔ a provider cannot execute tests against an explicit no-test task even if it ignores advertised tools (1482.0105ms)
✔ a budget pause becomes visible only after its input snapshot can be sealed with the original Run (1270.4707ms)
✔ alternating duplicate file reads and listing files still pauses without treating new pages as duplicates (11266.4412ms)
ℹ tests 60
ℹ suites 0
ℹ pass 60
ℹ fail 0
ℹ cancelled 0
ℹ skipped 0
ℹ todo 0
ℹ duration_ms 65317.1144
@@ -1,6 +1,6 @@
PASS desktop saves seconds, preserves original pool/member and sends no model request
PASS invalid timeout rejected without saving
PASS saved values survive reload; narrow screen remains usable
PASS active request locks wait changes in both API and UI
PASS actual stalled stream shows exact idle reason, saves unknown use, never writes or retries
{"success":true,"directory":"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\model-wait-ui-79af8e3a-14f6-4114-a28f-580430d7d8e8","checks":5,"screenshots":4}
PASS desktop saves seconds, preserves original pool/member and sends no model request
PASS invalid timeout rejected without saving
PASS saved values survive reload; narrow screen remains usable
PASS active request locks wait changes in both API and UI
PASS actual stalled stream shows exact idle reason, saves unknown use, never writes or retries
{"success":true,"directory":"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\model-wait-ui-79af8e3a-14f6-4114-a28f-580430d7d8e8","checks":5,"screenshots":4}
@@ -1,6 +1,6 @@
PASS desktop saves seconds, preserves original pool/member and sends no model request
PASS invalid timeout rejected without saving
PASS saved values survive reload; narrow screen remains usable
PASS active request locks wait changes in both API and UI
PASS actual stalled stream shows exact idle reason, saves unknown use, never writes or retries
{"success":true,"directory":"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\model-wait-ui-6050b00d-fd6d-4e3f-a6e8-e703f6a8bb8c","checks":5,"screenshots":4}
PASS desktop saves seconds, preserves original pool/member and sends no model request
PASS invalid timeout rejected without saving
PASS saved values survive reload; narrow screen remains usable
PASS active request locks wait changes in both API and UI
PASS actual stalled stream shows exact idle reason, saves unknown use, never writes or retries
{"success":true,"directory":"D:\\Ai-project\\REA逆向\\Agent-Workbench\\.qa\\model-wait-ui-6050b00d-fd6d-4e3f-a6e8-e703f6a8bb8c","checks":5,"screenshots":4}
@@ -1,31 +1,31 @@
✔ JSON migration survives process exit before commit and retains exact legacy backups (4222.5377ms)
✔ process exit after_state rolls back state, event and outbox together (2616.4939ms)
✔ process exit after_event rolls back state, event and outbox together (2814.9776ms)
✔ crash recovery at accepted retains execution identity and never replays side effects (2317.0945ms)
✔ crash recovery at queued retains execution identity and never replays side effects (2731.1714ms)
✔ crash recovery at approval_wait retains execution identity and never replays side effects (3499.4361ms)
✔ crash recovery at approved retains execution identity and never replays side effects (2630.7769ms)
✔ crash recovery at effect_unrecorded retains execution identity and never replays side effects (2793.8295ms)
✔ crash recovery at result_committed retains execution identity and never replays side effects (2711.2371ms)
✔ crash recovery at before_next_model retains execution identity and never replays side effects (2910.4549ms)
✔ state and event failures publish no success and preserve the last committed memory snapshot (585.7129ms)
✔ tool result artifacts and event cursors retain project ownership across restart (1303.6506ms)
✔ token budgets pause before tools execute and continuation keeps original task and accumulated usage (1493.6537ms)
✔ a time budget interrupts an unresponsive model and persists a resumable run (989.5623ms)
✔ a corrupted large result artifact cannot become a recovered success (911.5817ms)
✔ schema1 connections remain readable without inventing observed capabilities or migrating original state (1304.3561ms)
✔ protocol, unsupported capabilities and stream choice persist; client claims cannot forge verified evidence (1244.037ms)
✔ ability evidence belongs to exact connection and credential revision; unknown and unsupported reject required routing (0.2737ms)
✔ invalid declarations and inconsistent stream settings reject before registry, revision and credentials mutate (1325.4096ms)
✔ hot upgrade adopts a retained virtual data root without reopening SQLite or changing rows, Provider or routing (1903.3432ms)
✔ failed adoption restores the original virtual data-root name and original open database (955.6613ms)
✔ hot upgrade refreshes retained model and importer methods while preserving credential and routing objects (1340.7943ms)
✔ hot upgrade refuses active probes and restores retained model methods if installation fails (854.4396ms)
ℹ tests 23
ℹ suites 0
ℹ pass 23
ℹ fail 0
ℹ cancelled 0
ℹ skipped 0
ℹ todo 0
ℹ duration_ms 34661.5798
✔ JSON migration survives process exit before commit and retains exact legacy backups (4222.5377ms)
✔ process exit after_state rolls back state, event and outbox together (2616.4939ms)
✔ process exit after_event rolls back state, event and outbox together (2814.9776ms)
✔ crash recovery at accepted retains execution identity and never replays side effects (2317.0945ms)
✔ crash recovery at queued retains execution identity and never replays side effects (2731.1714ms)
✔ crash recovery at approval_wait retains execution identity and never replays side effects (3499.4361ms)
✔ crash recovery at approved retains execution identity and never replays side effects (2630.7769ms)
✔ crash recovery at effect_unrecorded retains execution identity and never replays side effects (2793.8295ms)
✔ crash recovery at result_committed retains execution identity and never replays side effects (2711.2371ms)
✔ crash recovery at before_next_model retains execution identity and never replays side effects (2910.4549ms)
✔ state and event failures publish no success and preserve the last committed memory snapshot (585.7129ms)
✔ tool result artifacts and event cursors retain project ownership across restart (1303.6506ms)
✔ token budgets pause before tools execute and continuation keeps original task and accumulated usage (1493.6537ms)
✔ a time budget interrupts an unresponsive model and persists a resumable run (989.5623ms)
✔ a corrupted large result artifact cannot become a recovered success (911.5817ms)
✔ schema1 connections remain readable without inventing observed capabilities or migrating original state (1304.3561ms)
✔ protocol, unsupported capabilities and stream choice persist; client claims cannot forge verified evidence (1244.037ms)
✔ ability evidence belongs to exact connection and credential revision; unknown and unsupported reject required routing (0.2737ms)
✔ invalid declarations and inconsistent stream settings reject before registry, revision and credentials mutate (1325.4096ms)
✔ hot upgrade adopts a retained virtual data root without reopening SQLite or changing rows, Provider or routing (1903.3432ms)
✔ failed adoption restores the original virtual data-root name and original open database (955.6613ms)
✔ hot upgrade refreshes retained model and importer methods while preserving credential and routing objects (1340.7943ms)
✔ hot upgrade refuses active probes and restores retained model methods if installation fails (854.4396ms)
ℹ tests 23
ℹ suites 0
ℹ pass 23
ℹ fail 0
ℹ cancelled 0
ℹ skipped 0
ℹ todo 0
ℹ duration_ms 34661.5798
+2
View File
@@ -2,6 +2,8 @@
## 1.20.2 本地交付完成
最终Git字节核对补齐新证据目录的-text属性:保留验收日志原CRLF字节,避免后续检出时SHA与封存清单不符。按范围重新暂存原字节并增加逐个Git blob核对,产品源码和已安装发行均不变。首份ACCEPTANCE仅校验工作区证据;最终FINAL收据额外绑定提交内原证据字节,前份保留。
1.20.2已完成模型响应等待修复与保留状态部署:单次响应默认300秒、流式无数据120秒,模型中心可调整。111项专项/关联回归、真实本机JSON/SSE各65秒、五组等待UI/五组协议UI及五组随包等待UI通过;实际旧版到新版升级及正式Edge/Electron两视口回看通过。正式PID36552/Provider、17项目/31全部会话/3207记录/6138事件和1761个工件文件原字节保持,仅公开模型中心增加计算出的等待默认值;磁盘配置不回写。本轮验收发起真实模型请求0。
正式预检首轮辅助比较器错误要求旧版也已含新waitSettings字段,在安装前停止;修正为仅新增字段出现时做严格归一化,原失败保留,新b轮完成完整Setup暂存与独立候选核对,但因安装期间又新增原模型重试,原会话摘要变化被拒绝;新c轮沿用同一已校验候选,使用新鲜完整基线/备份后升级通过。用户自行新增的原超时记录一并保持。本次未主动续跑用户任务。代码提交3b650275ffe6984a1147a39ba1c22dd3a6859a61;正式UI首轮Edge两视口通过,Electron首窗口尚在导航时立即reload引发ERR_ABORTED;原失败和四张截图保留。b轮补等就绪后遇隐藏窗口截图超时;c轮沿用上一版已验证的可见窗口方式通过,复用已完成Edge截图,不修改产品或重复安装。源码/归档/安装SHA与最终去敏证据见docs/model-wait-release-identity.json。封存后另做本地文档Git收尾,不改不可变发行。