omnius 1.0.921 → 1.0.922
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/index.js +2 -2
- package/dist/library.js +588 -228
- package/dist/update-worker-runtime.js +9 -0
- package/docs/DISCOVERY.json +36 -0
- package/docs/DISCOVERY.md +1 -0
- package/docs/work-orders/evidence/telegram-context-retention-live-20261104.json +1024 -0
- package/npm-shrinkwrap.json +3 -2
- package/package.json +1 -1
|
@@ -0,0 +1,1024 @@
|
|
|
1
|
+
{
|
|
2
|
+
"schema": "omnius.telegram-context-retention-verification.v1",
|
|
3
|
+
"verifiedAt": "2026-11-05T00:59:16.629099+00:00",
|
|
4
|
+
"issues": [
|
|
5
|
+
10,
|
|
6
|
+
11,
|
|
7
|
+
12,
|
|
8
|
+
13,
|
|
9
|
+
14,
|
|
10
|
+
15
|
|
11
|
+
],
|
|
12
|
+
"baselineCommit": "d2ae0cb070ff6799321862cb323e16c71456ef99",
|
|
13
|
+
"changes": {
|
|
14
|
+
"10": {
|
|
15
|
+
"fix": "Retain exact, revision-bound read ranges in the protected source working set; use complete request capacity for deterministic range demotion; preserve CRLF and trailing whitespace; retire inline bodies only after verified complete extraction.",
|
|
16
|
+
"tests": [
|
|
17
|
+
"file-working-set-regression.test.ts",
|
|
18
|
+
"evidenceLedger.test.ts",
|
|
19
|
+
"canonical-source-body-runner.test.ts",
|
|
20
|
+
"multi-section-edit-runner-regression.test.ts"
|
|
21
|
+
]
|
|
22
|
+
},
|
|
23
|
+
"11": {
|
|
24
|
+
"fix": "Make raw-history admission non-destructive, clear stale pressure on later successful admission, retain protected canonical signals, protect the final frame, record every admission phase, and flag missing or empty task frames. Size automatic output reservation to the actual window.",
|
|
25
|
+
"tests": [
|
|
26
|
+
"task-frame-admission-retention.test.ts",
|
|
27
|
+
"primary-output-reservation.test.ts",
|
|
28
|
+
"agenticRunner-canonical-production.test.ts",
|
|
29
|
+
"multi-section-edit-runner-regression.test.ts"
|
|
30
|
+
]
|
|
31
|
+
},
|
|
32
|
+
"12": {
|
|
33
|
+
"fix": "Repeat file reads return actual source text. No discovery stub points at a frame that may be absent; cached reads remain range- and revision-bound.",
|
|
34
|
+
"tests": [
|
|
35
|
+
"file-working-set-regression.test.ts",
|
|
36
|
+
"file-read-output-streamlining.test.ts"
|
|
37
|
+
]
|
|
38
|
+
},
|
|
39
|
+
"13": {
|
|
40
|
+
"fix": "Preserve complete shell output when physical headroom fits it. Under pressure, bound individual lines before folding, preserve line-number prefixes, name omitted line intervals, and retain all ordinary lines in short log packets around an embedded asset.",
|
|
41
|
+
"tests": [
|
|
42
|
+
"output-preview.test.ts",
|
|
43
|
+
"multi-section-edit-runner-regression.test.ts"
|
|
44
|
+
]
|
|
45
|
+
},
|
|
46
|
+
"14": {
|
|
47
|
+
"fix": "Project Telegram passive/retained, health and world-state blocks wherever they appear; retain the exact prior request prefix; carry new exact source bodies verbatim; qualify raw/receipt and batch transitions with host argument/artifact custody; trace each fallback. Optional auxiliary calls defer without backend dispatch; required task review/extraction records an explicit cache handoff and subsequent chain restart.",
|
|
48
|
+
"tests": [
|
|
49
|
+
"append-only-outbound.test.ts",
|
|
50
|
+
"auxiliary-output-budget.test.ts",
|
|
51
|
+
"multi-section-edit-runner-regression.test.ts"
|
|
52
|
+
]
|
|
53
|
+
},
|
|
54
|
+
"15": {
|
|
55
|
+
"fix": "Repair the causal chain in order: frame admission, exact file working set and repeat reads, line-aware previews, then append-only and auxiliary cache behavior. The 32-turn runner regression covers a 54,161-byte/424-line file, embedded base64, three successive reads, atomic three-section edit, genuine Telegram history, retained/health/world blocks, 26 follow-up probes, exact source visibility, and immutable previous-request prefixes.",
|
|
56
|
+
"tests": [
|
|
57
|
+
"multi-section-edit-runner-regression.test.ts",
|
|
58
|
+
"scripts/live-multi-section-edit-harness.mjs"
|
|
59
|
+
]
|
|
60
|
+
}
|
|
61
|
+
},
|
|
62
|
+
"live": {
|
|
63
|
+
"endpoint": "http://127.0.0.1:11434",
|
|
64
|
+
"model": "robit/qwen3.8-27b-obliterated-e03:27b",
|
|
65
|
+
"gpuUuid": "GPU-3538758b-6ba7-fa1b-a80c-e315a606c3f0",
|
|
66
|
+
"gpuDevice": "GPU-3538758b-6ba7-fa1b-a80c-e315a606c3f0, NVIDIA A100 80GB PCIe, 8.0",
|
|
67
|
+
"contextTokens": 262144,
|
|
68
|
+
"runtimeSha256": "3b82d3f9a4691d6d72a9eb6a9f51cd4572bd436d8d9f4694490285d75d22d292",
|
|
69
|
+
"startedAt": "2026-11-05T00:53:37.976Z",
|
|
70
|
+
"finishedAt": "2026-11-05T00:54:59.333Z",
|
|
71
|
+
"mockedInferenceReplies": 0,
|
|
72
|
+
"mockedToolResults": 0,
|
|
73
|
+
"fixture": {
|
|
74
|
+
"bytes": 54161,
|
|
75
|
+
"lines": 424,
|
|
76
|
+
"sha256": "f8da87f63ef9cfc3d047a8c0a8feb675ef771870755ce48db4f147ca0b3dae09",
|
|
77
|
+
"expectedSha256": "8c224cbe5841b78c7c13fc53a3982338e6a48b1aae538a2cf6dd8a1cc6bd2f01"
|
|
78
|
+
},
|
|
79
|
+
"checks": {
|
|
80
|
+
"exactEdits": true,
|
|
81
|
+
"verified": true,
|
|
82
|
+
"threeReadsOrFewer": true,
|
|
83
|
+
"readCount": 3,
|
|
84
|
+
"turnsToFirstEdit": 4,
|
|
85
|
+
"everyRequestHasFrame": true,
|
|
86
|
+
"retainedThreeChunksAtEdit": true,
|
|
87
|
+
"optionalAuxiliaryDidNotEvict": true,
|
|
88
|
+
"allLaterRequestsExtend": true,
|
|
89
|
+
"noRepeatedToolCalls": true
|
|
90
|
+
},
|
|
91
|
+
"result": {
|
|
92
|
+
"completed": true,
|
|
93
|
+
"turns": 6,
|
|
94
|
+
"summary": "Edited index.html in three places: replaced the parallax-controls div with a details dropdown, changed display from flex to block (gap: 4px preserved), and replaced enableCamera's preview-only return with return [{x:0,y:0,z:0}]; while keeping the function async. The WASM_BASE64 asset and all other bytes were preserved. verify_fixture confirmed all three edits and byte preservation."
|
|
95
|
+
},
|
|
96
|
+
"optionalAuxiliary": {
|
|
97
|
+
"deferred": true,
|
|
98
|
+
"networkCalls": 0,
|
|
99
|
+
"reason": "primary_prompt_cache_reserved"
|
|
100
|
+
},
|
|
101
|
+
"requests": [
|
|
102
|
+
{
|
|
103
|
+
"turn": 1,
|
|
104
|
+
"logicalRequestId": "multi-section-live-1793840019244-1:turn:1:primary",
|
|
105
|
+
"messageCount": 8,
|
|
106
|
+
"requestChars": 24603,
|
|
107
|
+
"framePresent": true,
|
|
108
|
+
"exactPrefixOfPrevious": null,
|
|
109
|
+
"visibleRanges": [
|
|
110
|
+
false,
|
|
111
|
+
false,
|
|
112
|
+
false
|
|
113
|
+
]
|
|
114
|
+
},
|
|
115
|
+
{
|
|
116
|
+
"turn": 2,
|
|
117
|
+
"logicalRequestId": "multi-section-live-1793840019244-1:turn:2:primary",
|
|
118
|
+
"messageCount": 12,
|
|
119
|
+
"requestChars": 95444,
|
|
120
|
+
"framePresent": true,
|
|
121
|
+
"exactPrefixOfPrevious": true,
|
|
122
|
+
"visibleRanges": [
|
|
123
|
+
true,
|
|
124
|
+
false,
|
|
125
|
+
false
|
|
126
|
+
]
|
|
127
|
+
},
|
|
128
|
+
{
|
|
129
|
+
"turn": 3,
|
|
130
|
+
"logicalRequestId": "multi-section-live-1793840019244-1:turn:3:primary",
|
|
131
|
+
"messageCount": 16,
|
|
132
|
+
"requestChars": 168654,
|
|
133
|
+
"framePresent": true,
|
|
134
|
+
"exactPrefixOfPrevious": true,
|
|
135
|
+
"visibleRanges": [
|
|
136
|
+
true,
|
|
137
|
+
true,
|
|
138
|
+
false
|
|
139
|
+
]
|
|
140
|
+
},
|
|
141
|
+
{
|
|
142
|
+
"turn": 4,
|
|
143
|
+
"logicalRequestId": "multi-section-live-1793840019244-1:turn:4:primary",
|
|
144
|
+
"messageCount": 19,
|
|
145
|
+
"requestChars": 186535,
|
|
146
|
+
"framePresent": true,
|
|
147
|
+
"exactPrefixOfPrevious": true,
|
|
148
|
+
"visibleRanges": [
|
|
149
|
+
true,
|
|
150
|
+
true,
|
|
151
|
+
true
|
|
152
|
+
]
|
|
153
|
+
},
|
|
154
|
+
{
|
|
155
|
+
"turn": 5,
|
|
156
|
+
"logicalRequestId": "multi-section-live-1793840019244-1:turn:5:primary",
|
|
157
|
+
"messageCount": 22,
|
|
158
|
+
"requestChars": 240119,
|
|
159
|
+
"framePresent": true,
|
|
160
|
+
"exactPrefixOfPrevious": true,
|
|
161
|
+
"visibleRanges": [
|
|
162
|
+
true,
|
|
163
|
+
true,
|
|
164
|
+
true
|
|
165
|
+
]
|
|
166
|
+
},
|
|
167
|
+
{
|
|
168
|
+
"turn": 6,
|
|
169
|
+
"logicalRequestId": "multi-section-live-1793840019244-1:turn:6:primary",
|
|
170
|
+
"messageCount": 25,
|
|
171
|
+
"requestChars": 244684,
|
|
172
|
+
"framePresent": true,
|
|
173
|
+
"exactPrefixOfPrevious": true,
|
|
174
|
+
"visibleRanges": [
|
|
175
|
+
true,
|
|
176
|
+
true,
|
|
177
|
+
true
|
|
178
|
+
]
|
|
179
|
+
}
|
|
180
|
+
],
|
|
181
|
+
"inferenceMeasurements": [
|
|
182
|
+
{
|
|
183
|
+
"logicalRequestId": "multi-section-live-1793840019244-1:turn:1:primary",
|
|
184
|
+
"status": 200,
|
|
185
|
+
"model": "robit/qwen3.8-27b-obliterated-e03:27b",
|
|
186
|
+
"promptEvalCount": 6829,
|
|
187
|
+
"promptEvalDurationMs": 5149.782,
|
|
188
|
+
"evalCount": 64,
|
|
189
|
+
"headerMs": 7064,
|
|
190
|
+
"durationMs": 7087,
|
|
191
|
+
"requestSha256": "c1a4d97b4eb863aec35ca69620591e12e60c1495ceb8ff0c8a3fe3b63e4b0e7c",
|
|
192
|
+
"promptTokenDelta": 6829
|
|
193
|
+
},
|
|
194
|
+
{
|
|
195
|
+
"logicalRequestId": "multi-section-live-1793840019244-1:turn:2:primary",
|
|
196
|
+
"status": 200,
|
|
197
|
+
"model": "robit/qwen3.8-27b-obliterated-e03:27b",
|
|
198
|
+
"promptEvalCount": 26725,
|
|
199
|
+
"promptEvalDurationMs": 15902.183,
|
|
200
|
+
"evalCount": 66,
|
|
201
|
+
"headerMs": 17734,
|
|
202
|
+
"durationMs": 17760,
|
|
203
|
+
"requestSha256": "2d8048649955f102f3c10f3f049405369ce9a9f34d16d70b3a36ca4337730cd8",
|
|
204
|
+
"promptTokenDelta": 19896
|
|
205
|
+
},
|
|
206
|
+
{
|
|
207
|
+
"logicalRequestId": "multi-section-live-1793840019244-1:turn:3:primary",
|
|
208
|
+
"status": 200,
|
|
209
|
+
"model": "robit/qwen3.8-27b-obliterated-e03:27b",
|
|
210
|
+
"promptEvalCount": 42034,
|
|
211
|
+
"promptEvalDurationMs": 13573.853,
|
|
212
|
+
"evalCount": 66,
|
|
213
|
+
"headerMs": 15573,
|
|
214
|
+
"durationMs": 15604,
|
|
215
|
+
"requestSha256": "409fe30736f89d80fa87cd271eb491fd2ffb4031cf987026b95e5a9ddc313d1b",
|
|
216
|
+
"promptTokenDelta": 15309
|
|
217
|
+
},
|
|
218
|
+
{
|
|
219
|
+
"logicalRequestId": "multi-section-live-1793840019244-1:turn:4:primary",
|
|
220
|
+
"status": 200,
|
|
221
|
+
"model": "robit/qwen3.8-27b-obliterated-e03:27b",
|
|
222
|
+
"promptEvalCount": 46172,
|
|
223
|
+
"promptEvalDurationMs": 3884.151,
|
|
224
|
+
"evalCount": 196,
|
|
225
|
+
"headerMs": 9603,
|
|
226
|
+
"durationMs": 9634,
|
|
227
|
+
"requestSha256": "3085141f741b19adfaca212429edc941d5d92f8316e0869491cb71ccadd3de45",
|
|
228
|
+
"promptTokenDelta": 4138
|
|
229
|
+
},
|
|
230
|
+
{
|
|
231
|
+
"logicalRequestId": "multi-section-live-1793840019244-1:turn:5:primary",
|
|
232
|
+
"status": 200,
|
|
233
|
+
"model": "robit/qwen3.8-27b-obliterated-e03:27b",
|
|
234
|
+
"promptEvalCount": 59433,
|
|
235
|
+
"promptEvalDurationMs": 13043.922,
|
|
236
|
+
"evalCount": 15,
|
|
237
|
+
"headerMs": 13803,
|
|
238
|
+
"durationMs": 13961,
|
|
239
|
+
"requestSha256": "856b34a48394f6f521c5eb7c6d996c6c64063bd3773e306787104eb849381e6c",
|
|
240
|
+
"promptTokenDelta": 13261
|
|
241
|
+
},
|
|
242
|
+
{
|
|
243
|
+
"logicalRequestId": "multi-section-live-1793840019244-1:turn:6:primary",
|
|
244
|
+
"status": 200,
|
|
245
|
+
"model": "robit/qwen3.8-27b-obliterated-e03:27b",
|
|
246
|
+
"promptEvalCount": 60691,
|
|
247
|
+
"promptEvalDurationMs": 1351.973,
|
|
248
|
+
"evalCount": 150,
|
|
249
|
+
"headerMs": 6213,
|
|
250
|
+
"durationMs": 6242,
|
|
251
|
+
"requestSha256": "36c3cc544458774ee0d4ea16aed94851dbce3eeefe069c64231d741992f98973",
|
|
252
|
+
"promptTokenDelta": 1258
|
|
253
|
+
}
|
|
254
|
+
],
|
|
255
|
+
"appendOnlyDecisions": [
|
|
256
|
+
{
|
|
257
|
+
"logicalRequestId": "multi-section-live-1793840019244-1:turn:1:primary",
|
|
258
|
+
"mode": "full",
|
|
259
|
+
"reason": "new_chain"
|
|
260
|
+
},
|
|
261
|
+
{
|
|
262
|
+
"logicalRequestId": "multi-section-live-1793840019244-1:turn:2:primary",
|
|
263
|
+
"mode": "extended",
|
|
264
|
+
"reason": "extended_with_update"
|
|
265
|
+
},
|
|
266
|
+
{
|
|
267
|
+
"logicalRequestId": "multi-section-live-1793840019244-1:turn:3:primary",
|
|
268
|
+
"mode": "extended",
|
|
269
|
+
"reason": "extended_with_update"
|
|
270
|
+
},
|
|
271
|
+
{
|
|
272
|
+
"logicalRequestId": "multi-section-live-1793840019244-1:turn:4:primary",
|
|
273
|
+
"mode": "extended",
|
|
274
|
+
"reason": "extended_with_update"
|
|
275
|
+
},
|
|
276
|
+
{
|
|
277
|
+
"logicalRequestId": "multi-section-live-1793840019244-1:turn:5:primary",
|
|
278
|
+
"mode": "extended",
|
|
279
|
+
"reason": "extended_with_update"
|
|
280
|
+
},
|
|
281
|
+
{
|
|
282
|
+
"logicalRequestId": "multi-section-live-1793840019244-1:turn:6:primary",
|
|
283
|
+
"mode": "extended",
|
|
284
|
+
"reason": "extended_with_update"
|
|
285
|
+
}
|
|
286
|
+
],
|
|
287
|
+
"rawReportSha256": "562ea9414c5de37f80013871b328adb57acffa7c29a086bb2d619db48f149f09"
|
|
288
|
+
},
|
|
289
|
+
"liveShellPreview": {
|
|
290
|
+
"passed": true,
|
|
291
|
+
"endpoint": "http://127.0.0.1:11434",
|
|
292
|
+
"model": "robit/qwen3.8-27b-obliterated-e03:27b",
|
|
293
|
+
"gpuUuid": "GPU-3538758b-6ba7-fa1b-a80c-e315a606c3f0",
|
|
294
|
+
"mockedInferenceReplies": 0,
|
|
295
|
+
"mockedToolResults": 0,
|
|
296
|
+
"rawChars": 9061,
|
|
297
|
+
"previewChars": 3472,
|
|
298
|
+
"observedPairs": [
|
|
299
|
+
{
|
|
300
|
+
"line": 1,
|
|
301
|
+
"marker": "MARKER_01"
|
|
302
|
+
},
|
|
303
|
+
{
|
|
304
|
+
"line": 2,
|
|
305
|
+
"marker": "MARKER_02"
|
|
306
|
+
},
|
|
307
|
+
{
|
|
308
|
+
"line": 3,
|
|
309
|
+
"marker": "MARKER_03"
|
|
310
|
+
},
|
|
311
|
+
{
|
|
312
|
+
"line": 4,
|
|
313
|
+
"marker": "MARKER_04"
|
|
314
|
+
},
|
|
315
|
+
{
|
|
316
|
+
"line": 5,
|
|
317
|
+
"marker": "MARKER_05"
|
|
318
|
+
},
|
|
319
|
+
{
|
|
320
|
+
"line": 6,
|
|
321
|
+
"marker": "MARKER_06"
|
|
322
|
+
},
|
|
323
|
+
{
|
|
324
|
+
"line": 7,
|
|
325
|
+
"marker": "MARKER_07"
|
|
326
|
+
},
|
|
327
|
+
{
|
|
328
|
+
"line": 8,
|
|
329
|
+
"marker": "MARKER_08"
|
|
330
|
+
},
|
|
331
|
+
{
|
|
332
|
+
"line": 9,
|
|
333
|
+
"marker": "MARKER_09"
|
|
334
|
+
},
|
|
335
|
+
{
|
|
336
|
+
"line": 10,
|
|
337
|
+
"marker": "MARKER_10"
|
|
338
|
+
},
|
|
339
|
+
{
|
|
340
|
+
"line": 11,
|
|
341
|
+
"marker": "MARKER_11"
|
|
342
|
+
},
|
|
343
|
+
{
|
|
344
|
+
"line": 12,
|
|
345
|
+
"marker": "MARKER_12"
|
|
346
|
+
},
|
|
347
|
+
{
|
|
348
|
+
"line": 13,
|
|
349
|
+
"marker": "MARKER_13"
|
|
350
|
+
},
|
|
351
|
+
{
|
|
352
|
+
"line": 14,
|
|
353
|
+
"marker": "MARKER_14"
|
|
354
|
+
},
|
|
355
|
+
{
|
|
356
|
+
"line": 15,
|
|
357
|
+
"marker": "MARKER_15"
|
|
358
|
+
},
|
|
359
|
+
{
|
|
360
|
+
"line": 16,
|
|
361
|
+
"marker": "MARKER_16"
|
|
362
|
+
},
|
|
363
|
+
{
|
|
364
|
+
"line": 17,
|
|
365
|
+
"marker": "MARKER_17"
|
|
366
|
+
},
|
|
367
|
+
{
|
|
368
|
+
"line": 18,
|
|
369
|
+
"marker": "MARKER_18"
|
|
370
|
+
},
|
|
371
|
+
{
|
|
372
|
+
"line": 19,
|
|
373
|
+
"marker": "MARKER_19"
|
|
374
|
+
},
|
|
375
|
+
{
|
|
376
|
+
"line": 20,
|
|
377
|
+
"marker": "MARKER_20"
|
|
378
|
+
},
|
|
379
|
+
{
|
|
380
|
+
"line": 21,
|
|
381
|
+
"marker": "MARKER_21"
|
|
382
|
+
},
|
|
383
|
+
{
|
|
384
|
+
"line": 22,
|
|
385
|
+
"marker": "MARKER_22"
|
|
386
|
+
},
|
|
387
|
+
{
|
|
388
|
+
"line": 23,
|
|
389
|
+
"marker": "MARKER_23"
|
|
390
|
+
},
|
|
391
|
+
{
|
|
392
|
+
"line": 24,
|
|
393
|
+
"marker": "MARKER_24"
|
|
394
|
+
},
|
|
395
|
+
{
|
|
396
|
+
"line": 25,
|
|
397
|
+
"marker": "MARKER_25"
|
|
398
|
+
},
|
|
399
|
+
{
|
|
400
|
+
"line": 26,
|
|
401
|
+
"marker": "MARKER_26"
|
|
402
|
+
}
|
|
403
|
+
],
|
|
404
|
+
"usage": {
|
|
405
|
+
"totalTokens": 1734,
|
|
406
|
+
"promptTokens": 1395,
|
|
407
|
+
"completionTokens": 339
|
|
408
|
+
},
|
|
409
|
+
"scope": "Real grep output and production pressure-bounded preview, followed by real model extraction of all 26 line/marker pairs.",
|
|
410
|
+
"rawReportSha256": "2f7115bd2ea1b3668fc1d7aae660c8e6081c01f20d54c281b4c5f6f968289bb5"
|
|
411
|
+
},
|
|
412
|
+
"checks": {
|
|
413
|
+
"workspaceBuild": "passed: pnpm -r build",
|
|
414
|
+
"focusedOrchestrator": {
|
|
415
|
+
"files": 23,
|
|
416
|
+
"passed": 277,
|
|
417
|
+
"skipped": 9,
|
|
418
|
+
"logSha256": "198e3c18dc29289d7009577a8ac08db85ed037fdba6875b0bdc8e154b843c627"
|
|
419
|
+
},
|
|
420
|
+
"continuation": {
|
|
421
|
+
"files": 1,
|
|
422
|
+
"passed": 31,
|
|
423
|
+
"logSha256": "b0ff5aaf7d8d8955a5ea1071dba238d8507467306834d3aa7b81059fbafc1616"
|
|
424
|
+
},
|
|
425
|
+
"execution": {
|
|
426
|
+
"files": 5,
|
|
427
|
+
"passed": 58
|
|
428
|
+
},
|
|
429
|
+
"cacheAuxiliarySteering": {
|
|
430
|
+
"files": 3,
|
|
431
|
+
"passed": 37,
|
|
432
|
+
"logSha256": "7f41c0b7f7720e20a440ab70568646d27e3cb5f727fae29ce0519c26ba86a878"
|
|
433
|
+
},
|
|
434
|
+
"shellAuthority": {
|
|
435
|
+
"files": 1,
|
|
436
|
+
"passed": 28,
|
|
437
|
+
"logSha256": "71ef011de94de25e40eb9b81f48dfa6ee4dd4fa6d51a9c063050dce832889890"
|
|
438
|
+
},
|
|
439
|
+
"worksetOptimization": {
|
|
440
|
+
"files": 6,
|
|
441
|
+
"passed": 42,
|
|
442
|
+
"includes32TurnRegression": true
|
|
443
|
+
},
|
|
444
|
+
"longLoopTiming": {
|
|
445
|
+
"passed": 9,
|
|
446
|
+
"minMs": 24420,
|
|
447
|
+
"maxMs": 28443,
|
|
448
|
+
"unchangedDefaultTimeoutMs": 30000,
|
|
449
|
+
"diagnosticGuardMs": 60000,
|
|
450
|
+
"timeoutEdits": false,
|
|
451
|
+
"logSha256": "f26eb669b220a9235a0dd836ad8364de16a3debf8bca360ed96de378545c02f5"
|
|
452
|
+
},
|
|
453
|
+
"overlappingSuites": true,
|
|
454
|
+
"buildLogSha256": "be3d9baf3aa63add506fbf3cb398d0bd1abcd687be0accfa3ec5e2d09a749bfd"
|
|
455
|
+
},
|
|
456
|
+
"broadSweep": {
|
|
457
|
+
"snapshot": "Mixed snapshot while repairs were in progress; not a final green full-suite run.",
|
|
458
|
+
"files": {
|
|
459
|
+
"passed": 452,
|
|
460
|
+
"failed": 47,
|
|
461
|
+
"skipped": 10
|
|
462
|
+
},
|
|
463
|
+
"tests": {
|
|
464
|
+
"passed": 6705,
|
|
465
|
+
"failed": 161,
|
|
466
|
+
"skipped": 163
|
|
467
|
+
},
|
|
468
|
+
"logSha256": "fcca939cc2caf4352b8b4bfab2d46eb6c379d9c82a056f85ca7556b494e77aec",
|
|
469
|
+
"independentlyReproducedBaselineFailures": 131,
|
|
470
|
+
"classificationCounts": {
|
|
471
|
+
"baseline_reproduced": 131,
|
|
472
|
+
"focused_new_or_repaired": 14,
|
|
473
|
+
"unclassified": 0,
|
|
474
|
+
"baseline_pass_verified_or_intentionally_updated": 16
|
|
475
|
+
},
|
|
476
|
+
"baselineAuditSha256": "57599d35893c5d850689dd4bce4129a0bf34ce8e85b9571f60b74e09f4d8b093",
|
|
477
|
+
"finalClassification": "All 161 mixed-sweep failures classified: 131 reproduced on baseline, 16 resolved or intentional expectation/timing checks, and 14 new/repaired focused cases. No unresolved functional regression after final checks.",
|
|
478
|
+
"baselineFailureCases": [
|
|
479
|
+
{
|
|
480
|
+
"file": "tests/activity-diagnostic-visibility.test.ts",
|
|
481
|
+
"name": "diagnostic-only activity producers > expires stale failure records quietly while preserving current failure evidence"
|
|
482
|
+
},
|
|
483
|
+
{
|
|
484
|
+
"file": "tests/admission-tool-intent.test.ts",
|
|
485
|
+
"name": "selected schemas across final context admission > never restores tools into the exact pending steering packet"
|
|
486
|
+
},
|
|
487
|
+
{
|
|
488
|
+
"file": "tests/agent-operating-contract.test.ts",
|
|
489
|
+
"name": "runtime capability contract > changes revision with semantics and names the most relevant deferred equivalent directly"
|
|
490
|
+
},
|
|
491
|
+
{
|
|
492
|
+
"file": "tests/agenticRunner-context-recovery.test.ts",
|
|
493
|
+
"name": "AgenticRunner context-overflow recovery > learns a provider ceiling, retries exactly one changed request, and never calls it unavailable"
|
|
494
|
+
},
|
|
495
|
+
{
|
|
496
|
+
"file": "tests/agenticRunner-control-observation.test.ts",
|
|
497
|
+
"name": "executed observations versus host control refusals > does not count an unknown-tool host refusal as an executed observation"
|
|
498
|
+
},
|
|
499
|
+
{
|
|
500
|
+
"file": "tests/agenticRunner-debug-artifacts.test.ts",
|
|
501
|
+
"name": "AgenticRunner debug artifacts > uses an immutable execution run ID for all artifacts while retaining session identity"
|
|
502
|
+
},
|
|
503
|
+
{
|
|
504
|
+
"file": "tests/agenticRunner-debug-artifacts.test.ts",
|
|
505
|
+
"name": "AgenticRunner debug artifacts > mirrors active leaf todos into deterministic workboard cards"
|
|
506
|
+
},
|
|
507
|
+
{
|
|
508
|
+
"file": "tests/agenticRunner-native-provider-staging.test.ts",
|
|
509
|
+
"name": "native result staging after historical failed attempts > preserves failed write and plateau evidence while staging a current reviewed mutation result for actual provider acknowledgement"
|
|
510
|
+
},
|
|
511
|
+
{
|
|
512
|
+
"file": "tests/agenticRunner-native-provider-staging.test.ts",
|
|
513
|
+
"name": "native result staging after historical failed attempts > withholds canonical staging after candidate changes"
|
|
514
|
+
},
|
|
515
|
+
{
|
|
516
|
+
"file": "tests/agenticRunner-native-provider-staging.test.ts",
|
|
517
|
+
"name": "native result staging after historical failed attempts > withholds canonical staging after physical bytes changes"
|
|
518
|
+
},
|
|
519
|
+
{
|
|
520
|
+
"file": "tests/agenticRunner-native-provider-staging.test.ts",
|
|
521
|
+
"name": "native result staging after historical failed attempts > withholds canonical staging after missing review changes"
|
|
522
|
+
},
|
|
523
|
+
{
|
|
524
|
+
"file": "tests/agenticRunner-native-provider-staging.test.ts",
|
|
525
|
+
"name": "native result staging after historical failed attempts > withholds canonical staging after failed review changes"
|
|
526
|
+
},
|
|
527
|
+
{
|
|
528
|
+
"file": "tests/agenticRunner-native-provider-staging.test.ts",
|
|
529
|
+
"name": "native result staging after historical failed attempts > withholds canonical staging after native amendment changes"
|
|
530
|
+
},
|
|
531
|
+
{
|
|
532
|
+
"file": "tests/agenticRunner-native-provider-staging.test.ts",
|
|
533
|
+
"name": "native result staging after historical failed attempts > withholds canonical staging after independent verifier changes"
|
|
534
|
+
},
|
|
535
|
+
{
|
|
536
|
+
"file": "tests/agenticRunner-native-provider-staging.test.ts",
|
|
537
|
+
"name": "native result staging after historical failed attempts > withholds canonical staging after pending operation changes"
|
|
538
|
+
},
|
|
539
|
+
{
|
|
540
|
+
"file": "tests/agenticRunner-native-provider-staging.test.ts",
|
|
541
|
+
"name": "native result staging after historical failed attempts > withholds canonical staging after foreign run changes"
|
|
542
|
+
},
|
|
543
|
+
{
|
|
544
|
+
"file": "tests/agenticRunner-native-provider-staging.test.ts",
|
|
545
|
+
"name": "native result staging after historical failed attempts > withholds canonical staging after foreign epoch changes"
|
|
546
|
+
},
|
|
547
|
+
{
|
|
548
|
+
"file": "tests/agenticRunner-native-provider-staging.test.ts",
|
|
549
|
+
"name": "native result staging after historical failed attempts > withholds canonical staging after corrupt snapshot changes"
|
|
550
|
+
},
|
|
551
|
+
{
|
|
552
|
+
"file": "tests/agenticRunner-pdf-source-review.test.ts",
|
|
553
|
+
"name": "real native PDF shared source-review evidence > keeps independent task_complete source review authoritative for 'current' evidence (direct=false)"
|
|
554
|
+
},
|
|
555
|
+
{
|
|
556
|
+
"file": "tests/agenticRunner-pdf-source-review.test.ts",
|
|
557
|
+
"name": "real native PDF shared source-review evidence > keeps independent task_complete source review authoritative for 'current' evidence (direct=true)"
|
|
558
|
+
},
|
|
559
|
+
{
|
|
560
|
+
"file": "tests/agenticRunner-pdf-source-review.test.ts",
|
|
561
|
+
"name": "real native PDF shared source-review evidence > keeps independent task_complete source review authoritative for 'changed during review' evidence (direct=false)"
|
|
562
|
+
},
|
|
563
|
+
{
|
|
564
|
+
"file": "tests/agenticRunner-pdf-source-review.test.ts",
|
|
565
|
+
"name": "real native PDF shared source-review evidence > keeps independent task_complete source review authoritative for 'changed during review' evidence (direct=true)"
|
|
566
|
+
},
|
|
567
|
+
{
|
|
568
|
+
"file": "tests/agenticRunner-pdf-source-review.test.ts",
|
|
569
|
+
"name": "real native PDF shared source-review evidence > keeps independent task_complete source review authoritative for 'unsupported' evidence (direct=false)"
|
|
570
|
+
},
|
|
571
|
+
{
|
|
572
|
+
"file": "tests/agenticRunner-pdf-source-review.test.ts",
|
|
573
|
+
"name": "real native PDF shared source-review evidence > keeps independent task_complete source review authoritative for 'unsupported' evidence (direct=true)"
|
|
574
|
+
},
|
|
575
|
+
{
|
|
576
|
+
"file": "tests/agenticRunner-requirement-lifecycle.test.ts",
|
|
577
|
+
"name": "production requirement/plan/completion lifecycle > dispatches the current requirements and selected plan purpose and durably accepts only a genuine current review"
|
|
578
|
+
},
|
|
579
|
+
{
|
|
580
|
+
"file": "tests/agenticRunner-requirement-lifecycle.test.ts",
|
|
581
|
+
"name": "production requirement/plan/completion lifecycle > an omitted requirement keeps the result open despite successful task_complete adapter"
|
|
582
|
+
},
|
|
583
|
+
{
|
|
584
|
+
"file": "tests/agenticRunner-requirement-lifecycle.test.ts",
|
|
585
|
+
"name": "production requirement/plan/completion lifecycle > reviews ordinary text before the final canonical acceptance check"
|
|
586
|
+
},
|
|
587
|
+
{
|
|
588
|
+
"file": "tests/agenticRunner-requirement-lifecycle.test.ts",
|
|
589
|
+
"name": "production requirement/plan/completion lifecycle > selects an initially proposed unit and links actual mutation outcomes without inventing acceptance"
|
|
590
|
+
},
|
|
591
|
+
{
|
|
592
|
+
"file": "tests/agenticRunner-requirement-lifecycle.test.ts",
|
|
593
|
+
"name": "production requirement/plan/completion lifecycle > rejects a nominally passing review when actual owned file contents change while it runs"
|
|
594
|
+
},
|
|
595
|
+
{
|
|
596
|
+
"file": "tests/agenticRunner-requirement-lifecycle.test.ts",
|
|
597
|
+
"name": "production requirement/plan/completion lifecycle > exposes a real unrelated mutation to the reviewer and keeps the requested target obligation open"
|
|
598
|
+
},
|
|
599
|
+
{
|
|
600
|
+
"file": "tests/agenticRunner.test.ts",
|
|
601
|
+
"name": "AgenticRunner > keeps the first admitted file_read canonical at the model boundary"
|
|
602
|
+
},
|
|
603
|
+
{
|
|
604
|
+
"file": "tests/agenticRunner.test.ts",
|
|
605
|
+
"name": "AgenticRunner > sends and records the exact canonical request after primary admission"
|
|
606
|
+
},
|
|
607
|
+
{
|
|
608
|
+
"file": "tests/agenticRunner.test.ts",
|
|
609
|
+
"name": "AgenticRunner > preserves canonical transport parity in the brute-force path"
|
|
610
|
+
},
|
|
611
|
+
{
|
|
612
|
+
"file": "tests/agenticRunner.test.ts",
|
|
613
|
+
"name": "AgenticRunner > does not hard-gate repeated no-op memory_write calls by default"
|
|
614
|
+
},
|
|
615
|
+
{
|
|
616
|
+
"file": "tests/agenticRunner.test.ts",
|
|
617
|
+
"name": "AgenticRunner > binds telegram-public tool calls to the host-owned elevation surface"
|
|
618
|
+
},
|
|
619
|
+
{
|
|
620
|
+
"file": "tests/agenticRunner.test.ts",
|
|
621
|
+
"name": "AgenticRunner > binds telegram-admin tool calls to the host-owned elevation surface"
|
|
622
|
+
},
|
|
623
|
+
{
|
|
624
|
+
"file": "tests/agenticRunner.test.ts",
|
|
625
|
+
"name": "AgenticRunner > tool_search exposes custom tool quality and docs metadata"
|
|
626
|
+
},
|
|
627
|
+
{
|
|
628
|
+
"file": "tests/agenticRunner.test.ts",
|
|
629
|
+
"name": "AgenticRunner > supplies compact creation ground truth to the next main request"
|
|
630
|
+
},
|
|
631
|
+
{
|
|
632
|
+
"file": "tests/agenticRunner.test.ts",
|
|
633
|
+
"name": "AgenticRunner > stops a medium-model empty-response loop without accepting completion"
|
|
634
|
+
},
|
|
635
|
+
{
|
|
636
|
+
"file": "tests/assistant-text-emission.test.ts",
|
|
637
|
+
"name": "model-authored assistant text at real response boundaries > publishes each answer once in 'brute' with stream=true"
|
|
638
|
+
},
|
|
639
|
+
{
|
|
640
|
+
"file": "tests/assistant-text-emission.test.ts",
|
|
641
|
+
"name": "model-authored assistant text at real response boundaries > publishes each answer once in 'brute' with stream=false"
|
|
642
|
+
},
|
|
643
|
+
{
|
|
644
|
+
"file": "tests/configured-endpoint-inference.test.ts",
|
|
645
|
+
"name": "owner-selected exact inference endpoint > preserves existing ordinary pool/auth/redirect behavior when the strict host hint is absent"
|
|
646
|
+
},
|
|
647
|
+
{
|
|
648
|
+
"file": "tests/conversation-report-runner.test.ts",
|
|
649
|
+
"name": "conversation evidence through production loops, brute=false > retains independent typed outcomes, multiple=true"
|
|
650
|
+
},
|
|
651
|
+
{
|
|
652
|
+
"file": "tests/conversation-report-runner.test.ts",
|
|
653
|
+
"name": "conversation evidence through production loops, brute=true > retains independent typed outcomes, multiple=true"
|
|
654
|
+
},
|
|
655
|
+
{
|
|
656
|
+
"file": "tests/conversation-report-runner.test.ts",
|
|
657
|
+
"name": "retains report-only evidence for a user-facing Telegram conversation sub-agent"
|
|
658
|
+
},
|
|
659
|
+
{
|
|
660
|
+
"file": "tests/conversation-report-runner.test.ts",
|
|
661
|
+
"name": "rejects an old in-flight callback after a production task-scope replacement"
|
|
662
|
+
},
|
|
663
|
+
{
|
|
664
|
+
"file": "tests/conversation-report-runner.test.ts",
|
|
665
|
+
"name": "does not let a passed child command assert aggregate delivery or mutation success"
|
|
666
|
+
},
|
|
667
|
+
{
|
|
668
|
+
"file": "tests/conversation-report-runner.test.ts",
|
|
669
|
+
"name": "keeps known aggregate failure when every attempted child command passed"
|
|
670
|
+
},
|
|
671
|
+
{
|
|
672
|
+
"file": "tests/deliveryCoverage-runner.test.ts",
|
|
673
|
+
"name": "authoritative completion coverage > holds a completed todo until its verifier runs after the final mutation"
|
|
674
|
+
},
|
|
675
|
+
{
|
|
676
|
+
"file": "tests/evidenceBranch-runner.test.ts",
|
|
677
|
+
"name": "AgenticRunner branch extraction > honors an explicit full read for a bounded 9KB source instead of replaying a branch wrapper"
|
|
678
|
+
},
|
|
679
|
+
{
|
|
680
|
+
"file": "tests/evidenceBranch-runner.test.ts",
|
|
681
|
+
"name": "AgenticRunner branch extraction > keeps a small whole-file read canonical when the real request has headroom"
|
|
682
|
+
},
|
|
683
|
+
{
|
|
684
|
+
"file": "tests/evidenceBranch-runner.test.ts",
|
|
685
|
+
"name": "AgenticRunner branch extraction > keeps a creative mutation available when an automatically routed branch is partial"
|
|
686
|
+
},
|
|
687
|
+
{
|
|
688
|
+
"file": "tests/evidenceBranch-runner.test.ts",
|
|
689
|
+
"name": "AgenticRunner branch extraction > catches a huge explicit range and never exposes its raw sentinel to the main loop"
|
|
690
|
+
},
|
|
691
|
+
{
|
|
692
|
+
"file": "tests/evidenceBranch-runner.test.ts",
|
|
693
|
+
"name": "AgenticRunner branch extraction > automatically runs one bounded inference frame for a multi-megabyte read and delivers only verified anchors"
|
|
694
|
+
},
|
|
695
|
+
{
|
|
696
|
+
"file": "tests/evidenceBranch-runner.test.ts",
|
|
697
|
+
"name": "AgenticRunner branch extraction > guards the original virtual/runtime-authored payload before generic output triage"
|
|
698
|
+
},
|
|
699
|
+
{
|
|
700
|
+
"file": "tests/evidenceBranch-runner.test.ts",
|
|
701
|
+
"name": "AgenticRunner branch extraction > applies the same pre-dispatch branch contract to streamed tool calls"
|
|
702
|
+
},
|
|
703
|
+
{
|
|
704
|
+
"file": "tests/evidenceBranch-runner.test.ts",
|
|
705
|
+
"name": "AgenticRunner branch extraction > does not turn repeated branch-backed reads into synthetic tool failures"
|
|
706
|
+
},
|
|
707
|
+
{
|
|
708
|
+
"file": "tests/feature-workflow-runner.test.ts",
|
|
709
|
+
"name": "native feature workflow in the real runner loops > retains exact workflow authority after actual parent context reduction, brute=false"
|
|
710
|
+
},
|
|
711
|
+
{
|
|
712
|
+
"file": "tests/feature-workflow-runner.test.ts",
|
|
713
|
+
"name": "native feature workflow in the real runner loops > retains exact workflow authority after actual parent context reduction, brute=true"
|
|
714
|
+
},
|
|
715
|
+
{
|
|
716
|
+
"file": "tests/feature-workflow-runner.test.ts",
|
|
717
|
+
"name": "native feature workflow in the real runner loops > binds typed continue steering without resurrecting the old workflow"
|
|
718
|
+
},
|
|
719
|
+
{
|
|
720
|
+
"file": "tests/feature-workflow-runner.test.ts",
|
|
721
|
+
"name": "native feature workflow in the real runner loops > binds typed replace steering without resurrecting the old workflow"
|
|
722
|
+
},
|
|
723
|
+
{
|
|
724
|
+
"file": "tests/feature-workflow-runner.test.ts",
|
|
725
|
+
"name": "native feature workflow in the real runner loops > binds typed replace_during_child steering without resurrecting the old workflow"
|
|
726
|
+
},
|
|
727
|
+
{
|
|
728
|
+
"file": "tests/git-progress.test.ts",
|
|
729
|
+
"name": "git progress tracking > exposes Git progress through the runner context renderers"
|
|
730
|
+
},
|
|
731
|
+
{
|
|
732
|
+
"file": "tests/git-progress.test.ts",
|
|
733
|
+
"name": "git progress tracking > omits the repository snapshot outside Git without hiding tracked state"
|
|
734
|
+
},
|
|
735
|
+
{
|
|
736
|
+
"file": "tests/grounded-recovery-progress.test.ts",
|
|
737
|
+
"name": "grounded progress and recovery advice at production boundaries > keeps a long novel read-only source run productive without todo or write pressure"
|
|
738
|
+
},
|
|
739
|
+
{
|
|
740
|
+
"file": "tests/grounded-recovery-progress.test.ts",
|
|
741
|
+
"name": "grounded progress and recovery advice at production boundaries > still reviews a stale/no-op window and gives SSMA current typed facts"
|
|
742
|
+
},
|
|
743
|
+
{
|
|
744
|
+
"file": "tests/grounded-recovery-progress.test.ts",
|
|
745
|
+
"name": "grounded progress and recovery advice at production boundaries > reviews a genuine later no-delta cadence even after one accepted source observation"
|
|
746
|
+
},
|
|
747
|
+
{
|
|
748
|
+
"file": "tests/grounded-recovery-progress.test.ts",
|
|
749
|
+
"name": "grounded progress and recovery advice at production boundaries > does not let retained progress samples from a replaced objective mask a new stale window"
|
|
750
|
+
},
|
|
751
|
+
{
|
|
752
|
+
"file": "tests/grounded-recovery-progress.test.ts",
|
|
753
|
+
"name": "grounded progress and recovery advice at production boundaries > rejects delayed PFV advice after same-owner source progress and retains the new facts on wire"
|
|
754
|
+
},
|
|
755
|
+
{
|
|
756
|
+
"file": "tests/grounded-recovery-progress.test.ts",
|
|
757
|
+
"name": "grounded progress and recovery advice at production boundaries > applies a current PFV result once, then retires it on semantic preparation"
|
|
758
|
+
},
|
|
759
|
+
{
|
|
760
|
+
"file": "tests/grounded-recovery-progress.test.ts",
|
|
761
|
+
"name": "grounded progress and recovery advice at production boundaries > rejects a delayed SSMA result when a legitimate new source advances the same task"
|
|
762
|
+
},
|
|
763
|
+
{
|
|
764
|
+
"file": "tests/grounded-recovery-progress.test.ts",
|
|
765
|
+
"name": "grounded progress and recovery advice at production boundaries > shows all inherited source coverage to the first grounding request with an empty new call log"
|
|
766
|
+
},
|
|
767
|
+
{
|
|
768
|
+
"file": "tests/grounded-recovery-progress.test.ts",
|
|
769
|
+
"name": "grounded progress and recovery advice at production boundaries > orders grounding independently from PFV advice while semantic progress invalidates both"
|
|
770
|
+
},
|
|
771
|
+
{
|
|
772
|
+
"file": "tests/grounded-recovery-progress.test.ts",
|
|
773
|
+
"name": "grounded progress and recovery advice at production boundaries > grounds retained observations before the first new call and rejects an assessment that finishes after new evidence"
|
|
774
|
+
},
|
|
775
|
+
{
|
|
776
|
+
"file": "tests/inference-compaction-runner.test.ts",
|
|
777
|
+
"name": "inference-driven runner compaction > uses the isolated analyst and retains exact source instead of a re-read pointer"
|
|
778
|
+
},
|
|
779
|
+
{
|
|
780
|
+
"file": "tests/inference-compaction-runner.test.ts",
|
|
781
|
+
"name": "inference-driven runner compaction > does not reinfer an unchanged compaction state"
|
|
782
|
+
},
|
|
783
|
+
{
|
|
784
|
+
"file": "tests/inference-network-registry.test.ts",
|
|
785
|
+
"name": "production inference network inventory > covers every production construction site for a network backend"
|
|
786
|
+
},
|
|
787
|
+
{
|
|
788
|
+
"file": "tests/inference-network-registry.test.ts",
|
|
789
|
+
"name": "production inference network inventory > keeps every documented entry point attached to a real source symbol"
|
|
790
|
+
},
|
|
791
|
+
{
|
|
792
|
+
"file": "tests/integration-mem-gaps.test.ts",
|
|
793
|
+
"name": "REG-61: sliding-window first-edit / sustained-edit nudge > keeps REG-61 detail in telemetry and exposes compact progress state"
|
|
794
|
+
},
|
|
795
|
+
{
|
|
796
|
+
"file": "tests/integration-mem-gaps.test.ts",
|
|
797
|
+
"name": "REG-61: sliding-window first-edit / sustained-edit nudge > sends compact controller state to the backend rather than a legacy REG-61 prompt"
|
|
798
|
+
},
|
|
799
|
+
{
|
|
800
|
+
"file": "tests/integration-mem-gaps.test.ts",
|
|
801
|
+
"name": "REG-61: sliding-window first-edit / sustained-edit nudge > BFC-61: REG-61 fires inside brute-force re-engagement cycles"
|
|
802
|
+
},
|
|
803
|
+
{
|
|
804
|
+
"file": "tests/integration-mem-gaps.test.ts",
|
|
805
|
+
"name": "REG-61: sliding-window first-edit / sustained-edit nudge > BFC-61: brute-force REG-61 fires include cycle label in status content"
|
|
806
|
+
},
|
|
807
|
+
{
|
|
808
|
+
"file": "tests/integration-mem-gaps.test.ts",
|
|
809
|
+
"name": "REG-61: sliding-window first-edit / sustained-edit nudge > BFC-61: stale main-loop cooldown does not block brute-force REG-61"
|
|
810
|
+
},
|
|
811
|
+
{
|
|
812
|
+
"file": "tests/integration-reg34-39.test.ts",
|
|
813
|
+
"name": "todo evidence reconciliation > accepts justified completed todos without executing their advisory verifier hints"
|
|
814
|
+
},
|
|
815
|
+
{
|
|
816
|
+
"file": "tests/integration-reg34-39.test.ts",
|
|
817
|
+
"name": "todo evidence reconciliation > records an absent declared artifact as feedback without withholding task_complete"
|
|
818
|
+
},
|
|
819
|
+
{
|
|
820
|
+
"file": "tests/large-model-controller-context.test.ts",
|
|
821
|
+
"name": "large-model controller context > keeps a compacted call's action and observed result visible on later turns"
|
|
822
|
+
},
|
|
823
|
+
{
|
|
824
|
+
"file": "tests/loop-recovery-scope.test.ts",
|
|
825
|
+
"name": "runtime loop recovery keeps the accepted objective scope > uses runtime guidance after repeated calls without injecting a new human or task epoch"
|
|
826
|
+
},
|
|
827
|
+
{
|
|
828
|
+
"file": "tests/native-web-html-evidence.test.ts",
|
|
829
|
+
"name": "puts the deep figure-to-asset source claim into the actual next primary without another fetch or model call"
|
|
830
|
+
},
|
|
831
|
+
{
|
|
832
|
+
"file": "tests/native-web-source-capacity.test.ts",
|
|
833
|
+
"name": "native web source presentation uses actual context capacity > still runs explicitly requested immutable-artifact extraction"
|
|
834
|
+
},
|
|
835
|
+
{
|
|
836
|
+
"file": "tests/native-web-source-durability.test.ts",
|
|
837
|
+
"name": "complete native web response before main context > commits the native body before the actual runner dispatches the next primary packet"
|
|
838
|
+
},
|
|
839
|
+
{
|
|
840
|
+
"file": "tests/native-web-source-durability.test.ts",
|
|
841
|
+
"name": "complete native web response before main context > archives the complete response before a pending extraction enters context, then reopens and searches it after restart"
|
|
842
|
+
},
|
|
843
|
+
{
|
|
844
|
+
"file": "tests/non-retryable-backend-reengagement.test.ts",
|
|
845
|
+
"name": "terminal broker failures across rescue cycles > does not re-engage after model_exceeds_gpu_capacity in the context-retry phase"
|
|
846
|
+
},
|
|
847
|
+
{
|
|
848
|
+
"file": "tests/non-retryable-backend-reengagement.test.ts",
|
|
849
|
+
"name": "terminal broker failures across rescue cycles > does not re-engage after logical_request_conflict in the context-retry phase"
|
|
850
|
+
},
|
|
851
|
+
{
|
|
852
|
+
"file": "tests/objective-loop-recovery-state.test.ts",
|
|
853
|
+
"name": "current objective loop recovery through production runner dispatch > delivers accepted reads and current recovery through primary dispatch with output ceiling 5000"
|
|
854
|
+
},
|
|
855
|
+
{
|
|
856
|
+
"file": "tests/objective-loop-recovery-state.test.ts",
|
|
857
|
+
"name": "current objective loop recovery through production runner dispatch > delivers accepted reads and current recovery through primary dispatch with output ceiling 262144"
|
|
858
|
+
},
|
|
859
|
+
{
|
|
860
|
+
"file": "tests/objective-loop-recovery-state.test.ts",
|
|
861
|
+
"name": "current objective loop recovery through production runner dispatch > does not project a cached no-op write or unsupported model prose as a source observation"
|
|
862
|
+
},
|
|
863
|
+
{
|
|
864
|
+
"file": "tests/objective-loop-recovery-state.test.ts",
|
|
865
|
+
"name": "current objective loop recovery through production runner dispatch > delivers the same fresh objective facts through primary and brute-force dispatch with output ceiling 5000"
|
|
866
|
+
},
|
|
867
|
+
{
|
|
868
|
+
"file": "tests/objective-loop-recovery-state.test.ts",
|
|
869
|
+
"name": "current objective loop recovery through production runner dispatch > delivers the same fresh objective facts through primary and brute-force dispatch with output ceiling 262144"
|
|
870
|
+
},
|
|
871
|
+
{
|
|
872
|
+
"file": "tests/objective-loop-recovery-state.test.ts",
|
|
873
|
+
"name": "current objective loop recovery through production runner dispatch > recovers a failed local alias within the unchanged parent goal"
|
|
874
|
+
},
|
|
875
|
+
{
|
|
876
|
+
"file": "tests/objective-loop-recovery-state.test.ts",
|
|
877
|
+
"name": "current objective loop recovery through production runner dispatch > keeps the native target visible when an observed source still requires verification"
|
|
878
|
+
},
|
|
879
|
+
{
|
|
880
|
+
"file": "tests/objective-loop-recovery-state.test.ts",
|
|
881
|
+
"name": "current objective loop recovery through production runner dispatch > omits source targets from a scope read-batch binding at actual dispatch"
|
|
882
|
+
},
|
|
883
|
+
{
|
|
884
|
+
"file": "tests/objective-loop-recovery-state.test.ts",
|
|
885
|
+
"name": "current objective loop recovery through production runner dispatch > omits source targets from a epoch read-batch binding at actual dispatch"
|
|
886
|
+
},
|
|
887
|
+
{
|
|
888
|
+
"file": "tests/objective-loop-recovery-state.test.ts",
|
|
889
|
+
"name": "current objective loop recovery through production runner dispatch > omits source targets from a ref read-batch binding at actual dispatch"
|
|
890
|
+
},
|
|
891
|
+
{
|
|
892
|
+
"file": "tests/objective-loop-recovery-state.test.ts",
|
|
893
|
+
"name": "current objective loop recovery through production runner dispatch > omits source targets from a hash read-batch binding at actual dispatch"
|
|
894
|
+
},
|
|
895
|
+
{
|
|
896
|
+
"file": "tests/objective-loop-recovery-state.test.ts",
|
|
897
|
+
"name": "current objective loop recovery through production runner dispatch > omits source targets from a expired read-batch binding at actual dispatch"
|
|
898
|
+
},
|
|
899
|
+
{
|
|
900
|
+
"file": "tests/objective-loop-recovery-state.test.ts",
|
|
901
|
+
"name": "current objective loop recovery through production runner dispatch > omits source targets from a revision read-batch binding at actual dispatch"
|
|
902
|
+
},
|
|
903
|
+
{
|
|
904
|
+
"file": "tests/objective-loop-recovery-state.test.ts",
|
|
905
|
+
"name": "current objective loop recovery through production runner dispatch > removes exact legacy host loop envelopes at dispatch while retaining identical quoted source and policy bytes"
|
|
906
|
+
},
|
|
907
|
+
{
|
|
908
|
+
"file": "tests/qwen9b-evidence-harness.test.ts",
|
|
909
|
+
"name": "qwen-9B evidence harness (small tier, conversational surface) > does not use phrase matching to manufacture a provenance hold"
|
|
910
|
+
},
|
|
911
|
+
{
|
|
912
|
+
"file": "tests/resident-memory-progress.test.ts",
|
|
913
|
+
"name": "external effect progress at the real runner stuck detector > does not treat fresh delivery receipts as zero-file-mutation stasis or task completion"
|
|
914
|
+
},
|
|
915
|
+
{
|
|
916
|
+
"file": "tests/resident-memory-progress.test.ts",
|
|
917
|
+
"name": "external effect progress at the real runner stuck detector > does not treat fresh external receipts as zero-file-mutation stasis or task completion"
|
|
918
|
+
},
|
|
919
|
+
{
|
|
920
|
+
"file": "tests/resident-memory-progress.test.ts",
|
|
921
|
+
"name": "external effect progress at the real runner stuck detector > retains genuine stalled observation detection"
|
|
922
|
+
},
|
|
923
|
+
{
|
|
924
|
+
"file": "tests/resident-memory-progress.test.ts",
|
|
925
|
+
"name": "external effect progress at the real runner stuck detector > retains genuine stalled repeated detection"
|
|
926
|
+
},
|
|
927
|
+
{
|
|
928
|
+
"file": "tests/resident-memory-progress.test.ts",
|
|
929
|
+
"name": "external effect progress at the real runner stuck detector > retains genuine stalled failed detection"
|
|
930
|
+
},
|
|
931
|
+
{
|
|
932
|
+
"file": "tests/resident-memory-progress.test.ts",
|
|
933
|
+
"name": "external effect progress at the real runner stuck detector > retains genuine stalled repeated-external detection"
|
|
934
|
+
},
|
|
935
|
+
{
|
|
936
|
+
"file": "tests/resident-memory-progress.test.ts",
|
|
937
|
+
"name": "external effect progress at the real runner stuck detector > retains genuine stalled malformed detection"
|
|
938
|
+
},
|
|
939
|
+
{
|
|
940
|
+
"file": "tests/resident-memory-progress.test.ts",
|
|
941
|
+
"name": "external effect progress at the real runner stuck detector > keeps stale-result detection even when recent provider effects are fresh"
|
|
942
|
+
},
|
|
943
|
+
{
|
|
944
|
+
"file": "tests/run-observability.test.ts",
|
|
945
|
+
"name": "run observability > expires or suppresses non-project failures before trajectory feedback"
|
|
946
|
+
},
|
|
947
|
+
{
|
|
948
|
+
"file": "tests/run-observability.test.ts",
|
|
949
|
+
"name": "run observability > retires an expired failure obligation durably without claiming verification"
|
|
950
|
+
},
|
|
951
|
+
{
|
|
952
|
+
"file": "tests/run-observability.test.ts",
|
|
953
|
+
"name": "run observability > records a bounded runtime receipt when dynamic guidance is sanitized"
|
|
954
|
+
},
|
|
955
|
+
{
|
|
956
|
+
"file": "tests/runner-verification-bundles.test.ts",
|
|
957
|
+
"name": "ordered verifier bundles through the runner evidence boundary > durably records rejected task_complete candidate/review bundle without delivery (review=false)"
|
|
958
|
+
},
|
|
959
|
+
{
|
|
960
|
+
"file": "tests/runner-verification-bundles.test.ts",
|
|
961
|
+
"name": "ordered verifier bundles through the runner evidence boundary > durably records rejected task_complete candidate/review bundle without delivery (review=true)"
|
|
962
|
+
},
|
|
963
|
+
{
|
|
964
|
+
"file": "tests/small-model-capability.test.ts",
|
|
965
|
+
"name": "WO-INF-03: Goal Re-injection After Compaction > failed approaches are included in goal re-injection"
|
|
966
|
+
},
|
|
967
|
+
{
|
|
968
|
+
"file": "tests/small-model-capability.test.ts",
|
|
969
|
+
"name": "WO-RLM-03: Placeholder Function Scaffolding > large tier defers non-core tool schemas behind tool_search"
|
|
970
|
+
},
|
|
971
|
+
{
|
|
972
|
+
"file": "tests/structured-tool-authority.test.ts",
|
|
973
|
+
"name": "structured tool authority > text protocol retains allowed argument contracts and filters blocked tools, retry=true"
|
|
974
|
+
},
|
|
975
|
+
{
|
|
976
|
+
"file": "tests/structured-tool-authority.test.ts",
|
|
977
|
+
"name": "structured tool authority > tools-unsupported retry uses ordinary admission, standalone=false"
|
|
978
|
+
},
|
|
979
|
+
{
|
|
980
|
+
"file": "tests/structured-tool-authority.test.ts",
|
|
981
|
+
"name": "structured tool authority > tools-unsupported retry uses ordinary admission, standalone=true"
|
|
982
|
+
},
|
|
983
|
+
{
|
|
984
|
+
"file": "tests/telegram-visible-response-authority.test.ts",
|
|
985
|
+
"name": "Telegram visible-response completion authority > rejects summary-only task_complete after pre-transport Telegram text"
|
|
986
|
+
},
|
|
987
|
+
{
|
|
988
|
+
"file": "tests/telegram-visible-response-authority.test.ts",
|
|
989
|
+
"name": "Telegram visible-response completion authority > accepts a typed Telegram delivery receipt as already-visible transport evidence"
|
|
990
|
+
},
|
|
991
|
+
{
|
|
992
|
+
"file": "tests/tool-search-preparatory-progress.test.ts",
|
|
993
|
+
"name": "host-authored tool discovery preparatory receipts > does not grant progress for cached renewals, aliases, contextual synonyms, or a changed query for the same activation"
|
|
994
|
+
},
|
|
995
|
+
{
|
|
996
|
+
"file": "tests/tool-search-preparatory-progress.test.ts",
|
|
997
|
+
"name": "host-authored tool discovery preparatory receipts > excludes already inline tools while annotating new registered names in a mixed selection"
|
|
998
|
+
},
|
|
999
|
+
{
|
|
1000
|
+
"file": "tests/tool-search-preparatory-progress.test.ts",
|
|
1001
|
+
"name": "host-authored tool discovery preparatory receipts > annotates a subset's first activation and leaves its cached renewal without preparatory credit"
|
|
1002
|
+
}
|
|
1003
|
+
]
|
|
1004
|
+
},
|
|
1005
|
+
"limits": [
|
|
1006
|
+
"Optional same-slot auxiliary calls defer. Required task review/extraction explicitly hands off the primary cache and records a restart. Independent clients can still evict a shared broker lane.",
|
|
1007
|
+
"The shell probe uses a rendering-pressure fixture; it does not claim to fill the real model window with the synthetic occupied-input packet."
|
|
1008
|
+
],
|
|
1009
|
+
"reproduce": [
|
|
1010
|
+
"pnpm -r build",
|
|
1011
|
+
"pnpm --filter @omnius/orchestrator exec vitest run tests/multi-section-edit-runner-regression.test.ts tests/file-working-set-regression.test.ts tests/task-frame-admission-retention.test.ts tests/append-only-outbound.test.ts tests/auxiliary-output-budget.test.ts",
|
|
1012
|
+
"After fresh GPU hardware preflight: node scripts/live-multi-section-edit-harness.mjs --run --endpoint http://127.0.0.1:11434 --model robit/qwen3.8-27b-obliterated-e03:27b --gpu GPU-3538758b-6ba7-fa1b-a80c-e315a606c3f0 --output /tmp/omnius-edit-live.json",
|
|
1013
|
+
"node scripts/live-shell-preview-harness.mjs --run --edit-report /tmp/omnius-edit-live.json --output /tmp/omnius-shell-live.json"
|
|
1014
|
+
],
|
|
1015
|
+
"sourceSha256": {
|
|
1016
|
+
"packages/orchestrator/src/agenticRunner.ts": "bbc364325b2570d5433da683c36d853ae27affa3623027864b15e93110ff583d",
|
|
1017
|
+
"packages/orchestrator/src/canonical-context-projection.ts": "a036bb21c22ff6d4c785aad0f6cdf96658b74be13fd3fea8bf35a5ebc326528c",
|
|
1018
|
+
"packages/orchestrator/src/append-only-outbound.ts": "110e1347d3ddbeb99af117d0c70373d28d93607557fb832c1bccb85106e84df4",
|
|
1019
|
+
"packages/orchestrator/src/evidenceLedger.ts": "94f1847018acfe639affcc84a72334ec16e03a1e22007f934d03c53140ea8f49",
|
|
1020
|
+
"packages/orchestrator/src/contextWindowDump.ts": "255f44e11c9e4a057c66f2acfc7957f42ac2e316f6d81a3eaba9c71327599ec4",
|
|
1021
|
+
"packages/execution/src/output-preview.ts": "a9df2ea196da12ee1d30e5b791d3ab6a4041cdf30ad1b56623c8d5f427171f3b",
|
|
1022
|
+
"packages/execution/src/tools/log-packet.ts": "ac556993730704a3252a5566c456673b6501249f2dec0dec822ff3fdd9d1924b"
|
|
1023
|
+
}
|
|
1024
|
+
}
|