@zq-silk/yui 2.7.1 → 2.8.1
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/cli/commandCatalog.js +8 -7
- package/dist/cli/updateCommand.js +3 -0
- package/dist/cli/updatePorts.js +20 -0
- package/dist/commands/configCommands.js +1 -0
- package/dist/commands/globalRoleCommands.js +42 -35
- package/dist/commands/inboxCommands.js +60 -40
- package/dist/commands/taskCommands.js +7 -3
- package/dist/commands/taskRoleRuntimeStatus.js +2 -0
- package/dist/context/runContextPack.js +48 -3
- package/dist/context/taskContext.js +36 -12
- package/dist/controlPlaneCli.js +64 -17
- package/dist/controller/controller.js +5 -0
- package/dist/controller/fileSchedulerStoreAdapter.js +9 -4
- package/dist/controller/globalInputDelivery.js +17 -5
- package/dist/controller/globalRuntimeAttention.js +135 -5
- package/dist/controller/runtime.js +4 -0
- package/dist/controller/structuredProviderObservation.js +51 -15
- package/dist/controller/taskHostTerminalRecovery.js +33 -0
- package/dist/executor/fileRoleLaunchPlanner.js +4 -9
- package/dist/message/inboxReceipt.js +30 -0
- package/dist/message/taskInboxOwnership.js +25 -3
- package/dist/models/modelManagement.js +469 -0
- package/dist/models/modelManagementConfig.js +46 -0
- package/dist/models/modelManagementFiles.js +90 -0
- package/dist/models/modelManagementHttp.js +90 -0
- package/dist/models/modelManagementTypes.js +1 -0
- package/dist/output/boundedRead.js +43 -14
- package/dist/runtime/agentEndpointIdentity.js +8 -4
- package/dist/runtime/agentEndpointOwnership.js +5 -6
- package/dist/runtime/agentHost.js +62 -10
- package/dist/runtime/agentHostImplementation.js +18 -0
- package/dist/runtime/managedCaller.js +25 -1
- package/dist/runtime/managedIdentity.js +1 -1
- package/dist/runtime/providerRetry.js +19 -1
- package/dist/runtime/sessionCaller.js +53 -0
- package/dist/storage/contextRecords.js +28 -0
- package/dist/storage/maintenanceCaller.js +69 -0
- package/dist/storage/sqliteSchema.js +12 -2
- package/dist/storage/sqliteStore.js +9 -5
- package/dist/storage/storageVersions.js +1 -1
- package/dist/storage/taskStore.js +4 -0
- package/dist/task/completionReadiness.js +3 -3
- package/dist/web/assets/assetManifest.js +28 -2
- package/dist/web/assets/client/app/manage.js +6 -0
- package/dist/web/assets/client/app/models.js +20 -0
- package/dist/web/assets/client/app/settings.js +41 -9
- package/dist/web/assets/client/i18n/en.js +180 -2
- package/dist/web/assets/client/i18n/zh-CN.js +178 -2
- package/dist/web/assets/client/index.js +8 -0
- package/dist/web/assets/client/layout/manageShell.js +94 -0
- package/dist/web/assets/client/lib/api.js +13 -1
- package/dist/web/assets/client/lib/i18n.js +5 -3
- package/dist/web/assets/client/views/dock/conversation.js +231 -80
- package/dist/web/assets/client/views/models.js +777 -0
- package/dist/web/assets/client/views/skills.js +22 -15
- package/dist/web/assets/shared/icons.js +2 -0
- package/dist/web/assets/shared/manageNav.js +13 -0
- package/dist/web/assets/shell/manage.js +54 -0
- package/dist/web/assets/shell/models.js +7 -0
- package/dist/web/assets/shell/settings.js +1 -35
- package/dist/web/assets/shell/settingsContent.js +30 -0
- package/dist/web/assets/shell/sidebar.js +1 -0
- package/dist/web/assets/styles/index.js +6 -0
- package/dist/web/assets/styles/views/conversation.js +133 -0
- package/dist/web/assets/styles/views/manage.js +71 -0
- package/dist/web/assets/styles/views/models.js +129 -0
- package/dist/web/assets/styles/views/settings.js +28 -14
- package/dist/web/assets/styles/views/sidebar.js +1 -0
- package/dist/web/webServer.js +59 -9
- package/docs/cli-information-contract.md +42 -11
- package/docs/cli-information-contract.zh-CN.md +28 -7
- package/docs/managed-turn-and-session-runtime.md +47 -1
- package/docs/managed-turn-and-session-runtime.zh-CN.md +32 -0
- package/docs/release-workflow.md +80 -7
- package/docs/release-workflow.zh-CN.md +55 -4
- package/docs/storage-baseline.md +6 -1
- package/docs/storage-baseline.zh-CN.md +4 -1
- package/docs/testing/review-evaluation.md +148 -0
- package/docs/testing/review-evaluation.zh-CN.md +117 -0
- package/docs/testing/verification-levels.md +6 -2
- package/docs/testing/verification-levels.zh-CN.md +3 -0
- package/package.json +1 -1
- package/skills/yui-leader/VERSION +1 -1
- package/skills/yui-leader/references/execution.md +29 -0
- package/skills/yui-leader/references/replicated-execution.md +7 -0
- package/skills/yui-operator/SKILL.md +5 -1
- package/skills/yui-operator/VERSION +1 -1
- package/skills/yui-reviewer/SKILL.md +35 -0
- package/skills/yui-reviewer/VERSION +1 -1
- package/skills/yui-runtime/SKILL.md +53 -7
- package/skills/yui-runtime/VERSION +1 -1
- package/skills/yui-runtime/references/recovery.md +24 -5
|
@@ -90,6 +90,35 @@ request a follow-up Round yourself.
|
|
|
90
90
|
Use Project Skills, Policy and Knowledge for project-specific checks. The
|
|
91
91
|
Task Contract and frozen scope determine acceptance and required evidence.
|
|
92
92
|
|
|
93
|
+
## Derive scenarios, then inspect the complete candidate
|
|
94
|
+
|
|
95
|
+
Read the frozen Task/WorkItem requirements, non-goals and applicable decisions
|
|
96
|
+
before using the author's explanation. The frozen Brief is a summary and material
|
|
97
|
+
index, not the contract body. Follow its relevant same-Task commit-pinned artifact
|
|
98
|
+
references with `task artifact read <task> <path> <commit>`; expand Context pointers
|
|
99
|
+
and their content pages as required. A path in prose grants no additional access.
|
|
100
|
+
Report missing or conflicting evidence instead of replacing it with a newer head.
|
|
101
|
+
|
|
102
|
+
Derive observable expectations from requirements and independent contracts, not
|
|
103
|
+
from the current implementation or its mocks. Trace the complete authorized
|
|
104
|
+
change and necessary callers, consumers, state and side effects. Select
|
|
105
|
+
complementary perspectives according to risk: user paths, producer/consumer
|
|
106
|
+
contracts, state/data invariants, failure and partial recovery, authorization,
|
|
107
|
+
and whether verification tests the actual contract. Multiple perspectives do
|
|
108
|
+
not require multiple Agents or every perspective on every task.
|
|
109
|
+
|
|
110
|
+
Turn applicable perspectives into concrete scenarios, expected behavior and
|
|
111
|
+
evidence. Challenge important assumptions with the smallest reachable
|
|
112
|
+
counterexample; check normal controls and non-goals too. A shared assumption
|
|
113
|
+
between a mock and implementation is not independent confirmation. Reuse valid
|
|
114
|
+
checks; do not run external-resource validation without its exact authorization.
|
|
115
|
+
Stop neither at the changed lines nor after finding a few defects.
|
|
116
|
+
|
|
117
|
+
For an ordinary full re-review, inspect the current complete frozen candidate
|
|
118
|
+
and requirements, using the repair delta, prior finding dispositions and still
|
|
119
|
+
applicable evidence as guidance, not a scope ceiling. An explicitly assigned
|
|
120
|
+
Delta Recheck retains its narrower equivalence contract above.
|
|
121
|
+
|
|
93
122
|
Review design complexity against the current Task Contract and reachable
|
|
94
123
|
operating paths. Report abstractions, state, indirection, fallback, or module
|
|
95
124
|
boundaries whose complete lifecycle cost exceeds their demonstrated value.
|
|
@@ -117,6 +146,12 @@ frozen Review or its original report.
|
|
|
117
146
|
|
|
118
147
|
A helpful default result layout is Conclusion, Material findings, Verification,
|
|
119
148
|
Uncertainty, and Recommended next action. Markdown or JSON are both acceptable.
|
|
149
|
+
Include a short account of the perspectives/scenarios actually inspected and
|
|
150
|
+
material unexamined risks, not a copied checklist. Each confirmed finding should
|
|
151
|
+
identify its trigger, violated expectation, impact and supporting evidence;
|
|
152
|
+
merge one root cause without losing its affected entry points. Distinguish
|
|
153
|
+
uncertain reachability or missing verification from a confirmed defect. There
|
|
154
|
+
is no finding quota, mandatory perspective count or claim of exhaustive coverage.
|
|
120
155
|
Yui preserves the original text and does not parse a verdict or reject it for
|
|
121
156
|
missing headings, field names, or formatting. Follow `yui-runtime`'s optional
|
|
122
157
|
`Yui-Summary:` first-line convention for attributed notification text; retain
|
|
@@ -1 +1 @@
|
|
|
1
|
-
|
|
1
|
+
3
|
|
@@ -79,6 +79,15 @@ installation with the same protocol and storage versions. If the entry or Home
|
|
|
79
79
|
is wrong or unavailable, report that fact and use the explicitly authorized
|
|
80
80
|
entry/environment; do not guess another installation or clear Session identity.
|
|
81
81
|
|
|
82
|
+
Desktop commands may carry only `CODEX_THREAD_ID`. The CLI resolves that native
|
|
83
|
+
identity in the selected Home's current Session registry before routing Task or
|
|
84
|
+
Global operations. Environment and Manifest values are location hints, not
|
|
85
|
+
continuing grants. Conflicting, unregistered, replaced or revoked identities
|
|
86
|
+
do not become public-user callers; a plain user terminal has no managed/native
|
|
87
|
+
identity. `inbox list` reports the resolved current identity. Do not add or clear
|
|
88
|
+
environment fields to repair a mismatch. Without a Manifest environment pointer,
|
|
89
|
+
`session skills` returns current Skill paths and the Session's captured profile.
|
|
90
|
+
|
|
82
91
|
For every explicitly dispatched managed Task AgentRun:
|
|
83
92
|
|
|
84
93
|
1. Read the exact AgentRun identity from the newest Bootstrap Envelope.
|
|
@@ -93,8 +102,15 @@ For every explicitly dispatched managed Task AgentRun:
|
|
|
93
102
|
and Adapter match the Envelope and Session Manifest. Stop and report a
|
|
94
103
|
context-load failure if the pack is missing, stale, unauthorized, malformed,
|
|
95
104
|
or mismatched. Never request an inline/full-prompt fallback.
|
|
96
|
-
4. Use the
|
|
97
|
-
the
|
|
105
|
+
4. Use the entry's `pointers` (including each pointer's summary) first. They are
|
|
106
|
+
a bounded window into the frozen readable-ref inventory. `pointerPage`
|
|
107
|
+
reports total/complete/nextCursor; discover further relevant refs with
|
|
108
|
+
`task run context <task>/<run> --view refs [--cursor <cursor>]`.
|
|
109
|
+
`deltaRefs` samples changed identities, with `deltaRefsTotal` reporting coverage.
|
|
110
|
+
`--view full` explicitly reads the complete pack when needed; do not load or
|
|
111
|
+
traverse every historical reference merely to make an entry window complete.
|
|
112
|
+
If `input.directiveOmitted` is true, read the complete assigned input with
|
|
113
|
+
`--view input` before acting; its entry summary is not the instruction body.
|
|
98
114
|
Expand only an authorized ref when
|
|
99
115
|
its full value is needed, selecting it by the pointer's exact `store` and
|
|
100
116
|
`refId`:
|
|
@@ -173,13 +189,20 @@ TUI, with the configured Agent, permissions and workspace unchanged. A live
|
|
|
173
189
|
unmanaged Session is not silently replaced or adopted; use an explicit Session
|
|
174
190
|
lifecycle action before enabling controlled delivery.
|
|
175
191
|
|
|
176
|
-
Context reads never consume queue entries. Global Context returns bounded
|
|
177
|
-
`pending`
|
|
192
|
+
Context reads never consume queue entries. Global Context returns a bounded
|
|
193
|
+
`pending` discovery page, not Message bodies, Profile bodies, recent history or
|
|
194
|
+
accumulated retry details. The Manifest retains the exact Role profile.
|
|
195
|
+
Read a referenced
|
|
178
196
|
original with `role message show <role> <message-id> --json`. Follow a pending
|
|
179
197
|
page's `nextCursor` using `role message list <role> --pending --cursor <cursor>`
|
|
180
198
|
when it is incomplete; accepted delivery does not erase the Message from the
|
|
181
199
|
separate all-message list. Native/transport acceptance is not implementation, and
|
|
182
200
|
`interrupt-requested` is not a stopped Turn or stopped background resources.
|
|
201
|
+
`role message list <role> --diagnostics` projects repeated classified failures
|
|
202
|
+
from original evidence with first/latest occurrence, count and recovery evidence.
|
|
203
|
+
`--diagnostic <key>` pages the original affected observations. Counts cover
|
|
204
|
+
distinct observations, not repeated polling of one identity. A recovery notice
|
|
205
|
+
is neither an input acknowledgement nor proof that prior effects were undone.
|
|
183
206
|
Saving a Task Message or Brief is not notification delivery to Operator. A
|
|
184
207
|
Task Role uses the existing Task InputRequest path for a genuinely missing
|
|
185
208
|
user decision, not global Role input controls or fabricated progress questions.
|
|
@@ -238,15 +261,32 @@ Retirement preserves the original claim and records the management decision;
|
|
|
238
261
|
it neither fabricates an acknowledgement nor proves native execution stopped.
|
|
239
262
|
An unfinished claim fences its Assignment, not unrelated work of the same Role.
|
|
240
263
|
|
|
241
|
-
In-flight, unknown,
|
|
264
|
+
In-flight, unknown, not-yet-delivered explicit interrupt-then and
|
|
242
265
|
unsettled steer inputs retain their existing owner/receipt and report a bounded
|
|
243
266
|
reason instead of being taken over. Use existing exact-receipt recovery when
|
|
244
267
|
needed; do not change request IDs, clear identity, or copy input to bypass it.
|
|
245
268
|
A steer with no control receipt was not submitted; it may be explicitly handled
|
|
246
269
|
and acknowledged in its authorized inbox, without converting it into a queue.
|
|
247
270
|
An accepted notification still follows its normal handling path, not another
|
|
248
|
-
|
|
249
|
-
|
|
271
|
+
native delivery. Accepted Global inputs, delivered Leader windows and current
|
|
272
|
+
Assignment continuations remain in the inbox until explicitly handled. Claim the
|
|
273
|
+
received entry, inspect its originals and existing effects, then ack with a
|
|
274
|
+
concrete disposition (including result or effect receipt references when useful).
|
|
275
|
+
This claim never sends the input again. A current rightful successor may settle
|
|
276
|
+
an accepted Global input or Leader window while retaining the original delivery
|
|
277
|
+
target; it must inspect prior effects first. Pending/unknown acceptance stays
|
|
278
|
+
blocked and cannot use claim as a replay path.
|
|
279
|
+
|
|
280
|
+
An ack belongs to the exact original and revision: fixed windows pin
|
|
281
|
+
`message.updated` event identities, including same-timestamp edits. A later
|
|
282
|
+
revision still needs its own handling. Before Task completion, settle handled
|
|
283
|
+
user-input windows; native acceptance or a terminal does not settle them.
|
|
284
|
+
Original Messages, wake windows and handling receipts remain readable after
|
|
285
|
+
acknowledgement. Lost context, Session continuation and installation upgrades
|
|
286
|
+
do not reset receipts. If ack fails after an external effect, read back the
|
|
287
|
+
original claim and the external receipt before retrying the same ack; do not
|
|
288
|
+
repeat the effect. This is durable handling evidence, not arbitrary external
|
|
289
|
+
exactly-once execution.
|
|
250
290
|
|
|
251
291
|
## Read bounded information completely when required
|
|
252
292
|
|
|
@@ -259,6 +299,12 @@ family, then `task context inspect <task> --store <store> --ref <refId>
|
|
|
259
299
|
--digest <digest>` to read its exact current value. Common Task lists use the
|
|
260
300
|
same `items`, `total`, `complete`, and `nextCursor` contract. Do not treat the
|
|
261
301
|
first page or a summary as complete requirements.
|
|
302
|
+
The 32 KiB compact-JSON read budget includes nested data, cursors and structure;
|
|
303
|
+
summaries are at most 512 UTF-8 bytes. Current Task Message discovery excludes
|
|
304
|
+
explicit acknowledgements and retirements, not uncertain delivery. Use
|
|
305
|
+
`--view current` for its continuation and the default all view for history.
|
|
306
|
+
Skill checks remain metadata-only; unusually large checks use the same explicit
|
|
307
|
+
document pagination. A small default window is not a request to fetch all history.
|
|
262
308
|
|
|
263
309
|
Long detail reads return `contentPage` instead of the ordinary value (under
|
|
264
310
|
`data.context` for Run context/expand, `data.contextDelta` for Run Context delta). Repeat the same read with
|
|
@@ -1 +1 @@
|
|
|
1
|
-
|
|
1
|
+
6
|
|
@@ -143,11 +143,30 @@ If exact native stop or inspection fails, read the persisted diagnostic and
|
|
|
143
143
|
resolve that resource boundary; do not kill arbitrary processes, clear unknown
|
|
144
144
|
execution records, or modify managed refs, tmux Sessions or state files.
|
|
145
145
|
|
|
146
|
-
|
|
147
|
-
|
|
148
|
-
|
|
149
|
-
|
|
150
|
-
|
|
146
|
+
Compatible upgrades preserve logical Sessions. Build generations and historical
|
|
147
|
+
Skill packages are provenance, not continuation gates. Load current stable
|
|
148
|
+
Skills and current authority; never change a user's model, account, native
|
|
149
|
+
permissions or workspace to make an upgrade work.
|
|
150
|
+
|
|
151
|
+
Within Host control/event v1, status advertises `controlCapabilities` and the
|
|
152
|
+
loaded `hostImplementation`, independently of `endpointImplementation`. Missing
|
|
153
|
+
capabilities or `eventDelivery` are unreported, not healthy. A Controller must
|
|
154
|
+
not repeatedly send an unsupported `reconcile-turn`; the existing raw Inbox
|
|
155
|
+
still permits exact event consumption without executing the input again.
|
|
156
|
+
Current Hosts retain unpersisted original facts during a filesystem failure;
|
|
157
|
+
keep that process alive until persistence recovers. Those volatile bytes cannot
|
|
158
|
+
be recovered after process loss merely by restarting the Controller.
|
|
159
|
+
|
|
160
|
+
Program/storage upgrade success does not mean Host replacement succeeded.
|
|
161
|
+
Compatible live Hosts may continue, including busy ones. A new Host may resume
|
|
162
|
+
the same native Session only through the existing authorized lifecycle after
|
|
163
|
+
exact execution, background resources and writer release are proven. An idle
|
|
164
|
+
status, a stopped proxy or a missing control field is insufficient. Unknown
|
|
165
|
+
effects preserve the original Session, input and resources; do not force-kill,
|
|
166
|
+
replay, hot-replace loaded modules or silently create a new Session. Existing
|
|
167
|
+
explicit stop/replacement authority is separate from an ordinary upgrade.
|
|
168
|
+
Unsupported protocols require an explicit bounded recovery plan, not arbitrary
|
|
169
|
+
historical Host adoption. Storage migration remains independently fenced.
|
|
151
170
|
|
|
152
171
|
After repeated failure of the same bounded recovery, report the observed
|
|
153
172
|
cause, impact and smallest remaining options. Do not broaden cleanup or add
|