dsh-github-copilot 0.1.0-alpha.1.dsh.0.2.1.alpha.5
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/AGENTS.md +439 -0
- package/CHANGELOG.md +875 -0
- package/CONTRIBUTING.md +55 -0
- package/LICENSE +21 -0
- package/README.md +366 -0
- package/README.zh.md +341 -0
- package/SECURITY.md +29 -0
- package/cordis.patch.yml +16 -0
- package/deployment-baseline.json +1580 -0
- package/docs/account-management-experience.md +143 -0
- package/docs/agent-readiness.md +110 -0
- package/docs/auto-high-cost.md +255 -0
- package/docs/auto-iteration-review.json +22 -0
- package/docs/auto-review-sampling.md +60 -0
- package/docs/auto-task-routing.md +341 -0
- package/docs/automatic-model-routing.md +236 -0
- package/docs/copilot-accounts.md +424 -0
- package/docs/copilot-compaction.md +148 -0
- package/docs/copilot-usage.md +133 -0
- package/docs/dual-model.md +66 -0
- package/docs/evidence-driven-iteration.md +86 -0
- package/docs/image-input-compatibility.md +64 -0
- package/docs/images/copilot-accounts-credits.png +0 -0
- package/docs/images/copilot-accounts.png +0 -0
- package/docs/images/copilot-current-provenance.json +39 -0
- package/docs/images/copilot-model-preferences.png +0 -0
- package/docs/images/copilot-search-routing.png +0 -0
- package/docs/images/copilot-session-continuation.png +0 -0
- package/docs/images/dual-model-desktop.png +0 -0
- package/docs/images/dual-model-mobile.png +0 -0
- package/docs/images/dual-model-provenance.json +35 -0
- package/docs/images/parent-model-follow.png +0 -0
- package/docs/manual-compaction-recovery.md +109 -0
- package/docs/mockups/copilot-account-dropdown.html +71 -0
- package/docs/mockups/copilot-account-management.html +393 -0
- package/docs/model-compatibility-acceptance.md +182 -0
- package/docs/npm-distribution.md +202 -0
- package/docs/official-first-016-alpha2.md +66 -0
- package/docs/official-first-020-rc1.md +32 -0
- package/docs/official-first-020-rc2.md +17 -0
- package/docs/official-first-021-alpha4.md +157 -0
- package/docs/official-first-021-alpha5.md +83 -0
- package/docs/plugin-diagnostics.md +605 -0
- package/docs/session-continuation.md +141 -0
- package/docs/session-search-routing.md +133 -0
- package/docs/single-route-migration.md +131 -0
- package/docs/web-lifecycle-rc2.md +55 -0
- package/lib/client.js +13236 -0
- package/lib/client.js.map +1 -0
- package/lib/copilot-identity-DRZY_d2X.js +15 -0
- package/lib/diagnostics-analysis-io.js +78 -0
- package/lib/diagnostics-analysis.js +427 -0
- package/lib/diagnostics-types-CF-qg8Do.js +708 -0
- package/lib/index.js +15661 -0
- package/lib/manual-compaction-recovery.js +485 -0
- package/lib/remote.js +1507 -0
- package/lib/request-budget-DLY_aRq0.js +87 -0
- package/lib/routed-web.js +120 -0
- package/lib/search-routing-CV2hs8as.js +792 -0
- package/lib/types/account-dropdown.d.ts +22 -0
- package/lib/types/account-model-auth.d.ts +15 -0
- package/lib/types/account-model-catalog.d.ts +70 -0
- package/lib/types/account-model-source.d.ts +85 -0
- package/lib/types/agent-compaction.d.ts +4 -0
- package/lib/types/authorization-code-clipboard.d.ts +4 -0
- package/lib/types/authorization-controller.d.ts +121 -0
- package/lib/types/auto-allocation-card.d.ts +9 -0
- package/lib/types/auto-allocation-diagnostics.d.ts +126 -0
- package/lib/types/auto-allocation-evidence.d.ts +44 -0
- package/lib/types/auto-allocation.d.ts +47 -0
- package/lib/types/auto-assessment-evidence.d.ts +111 -0
- package/lib/types/auto-model-host.d.ts +30 -0
- package/lib/types/auto-model-intent.d.ts +36 -0
- package/lib/types/auto-model-presentation.d.ts +57 -0
- package/lib/types/auto-model-routing.d.ts +64 -0
- package/lib/types/auto-review-sampling.d.ts +6 -0
- package/lib/types/auto-selection-explanation.d.ts +9 -0
- package/lib/types/auto-task-assessment.d.ts +49 -0
- package/lib/types/auto-task-classifier.d.ts +10 -0
- package/lib/types/background-compaction.d.ts +30 -0
- package/lib/types/client.d.ts +129 -0
- package/lib/types/compact-account.d.ts +42 -0
- package/lib/types/compaction-continuation-ui.d.ts +29 -0
- package/lib/types/compaction-continuation.d.ts +48 -0
- package/lib/types/compaction-input-estimate.d.ts +28 -0
- package/lib/types/compaction-pressure.d.ts +23 -0
- package/lib/types/compaction-replay.d.ts +18 -0
- package/lib/types/compatibility.d.ts +17 -0
- package/lib/types/composer-notice-style.d.ts +7 -0
- package/lib/types/config.d.ts +141 -0
- package/lib/types/content-file.d.ts +10 -0
- package/lib/types/context-evidence-ui.d.ts +28 -0
- package/lib/types/context-evidence.d.ts +256 -0
- package/lib/types/context-usage.d.ts +4 -0
- package/lib/types/copilot-account-presentation.d.ts +7 -0
- package/lib/types/copilot-accounts-card.d.ts +43 -0
- package/lib/types/copilot-accounts-host.d.ts +93 -0
- package/lib/types/copilot-accounts-remote.d.ts +85 -0
- package/lib/types/copilot-accounts-types.d.ts +43 -0
- package/lib/types/copilot-auth.d.ts +31 -0
- package/lib/types/copilot-grant.d.ts +11 -0
- package/lib/types/copilot-identity.d.ts +11 -0
- package/lib/types/copilot-request.d.ts +11 -0
- package/lib/types/copilot-stream-liveness.d.ts +23 -0
- package/lib/types/copilot-usage-card.d.ts +51 -0
- package/lib/types/copilot-usage-host.d.ts +61 -0
- package/lib/types/copilot-usage-normalize.d.ts +9 -0
- package/lib/types/copilot-usage-remote.d.ts +63 -0
- package/lib/types/copilot-usage-types.d.ts +20 -0
- package/lib/types/copilot-usage-ui.d.ts +4 -0
- package/lib/types/current-provider.d.ts +59 -0
- package/lib/types/deepseek-search-fallback.d.ts +11 -0
- package/lib/types/diagnostics-analysis-io.d.ts +23 -0
- package/lib/types/diagnostics-analysis.d.ts +298 -0
- package/lib/types/diagnostics-card.d.ts +12 -0
- package/lib/types/diagnostics-client.d.ts +13 -0
- package/lib/types/diagnostics-collector.d.ts +50 -0
- package/lib/types/diagnostics-host.d.ts +183 -0
- package/lib/types/diagnostics-remote.d.ts +113 -0
- package/lib/types/diagnostics-types.d.ts +893 -0
- package/lib/types/dual-model-host.d.ts +119 -0
- package/lib/types/dual-model-remote.d.ts +63 -0
- package/lib/types/dual-model-types.d.ts +34 -0
- package/lib/types/external-link.d.ts +3 -0
- package/lib/types/failure.d.ts +38 -0
- package/lib/types/http.d.ts +48 -0
- package/lib/types/image-input-admission.d.ts +5 -0
- package/lib/types/index.d.ts +39 -0
- package/lib/types/manual-compaction-recovery.d.ts +67 -0
- package/lib/types/migration-status.d.ts +50 -0
- package/lib/types/model-exclusions.d.ts +15 -0
- package/lib/types/model-preferences-card.d.ts +9 -0
- package/lib/types/model-protocol.d.ts +31 -0
- package/lib/types/native-select-style.d.ts +4 -0
- package/lib/types/native-settings-style.d.ts +10 -0
- package/lib/types/parent-model-follow-card.d.ts +32 -0
- package/lib/types/parent-model-follow.d.ts +44 -0
- package/lib/types/pi-provider-bridge.d.ts +22 -0
- package/lib/types/plan.d.ts +148 -0
- package/lib/types/pre-step-pressure.d.ts +8 -0
- package/lib/types/preview-provider.d.ts +77 -0
- package/lib/types/preview-route.d.ts +76 -0
- package/lib/types/probe.d.ts +39 -0
- package/lib/types/remote-codec.d.ts +10 -0
- package/lib/types/remote.d.ts +202 -0
- package/lib/types/replay-recovery-host.d.ts +28 -0
- package/lib/types/replay-recovery-remote.d.ts +14 -0
- package/lib/types/replay-recovery-types.d.ts +24 -0
- package/lib/types/replay-recovery-ui.d.ts +28 -0
- package/lib/types/replay-recovery.d.ts +17 -0
- package/lib/types/request-body-composition.d.ts +17 -0
- package/lib/types/request-body-evidence.d.ts +37 -0
- package/lib/types/request-body-timeout-marker.d.ts +2 -0
- package/lib/types/request-body-timeout.d.ts +4 -0
- package/lib/types/request-budget.d.ts +58 -0
- package/lib/types/request-diagnostics.d.ts +278 -0
- package/lib/types/request-upload-evidence.d.ts +23 -0
- package/lib/types/response-error-body.d.ts +3 -0
- package/lib/types/responses-reasoning-text.d.ts +24 -0
- package/lib/types/responses-reasoning.d.ts +7 -0
- package/lib/types/responses-replay-compat.d.ts +31 -0
- package/lib/types/responses-request-compression.d.ts +33 -0
- package/lib/types/route-ownership.d.ts +70 -0
- package/lib/types/routed-web.d.ts +50 -0
- package/lib/types/search-backend.d.ts +16 -0
- package/lib/types/search-routing-host.d.ts +19 -0
- package/lib/types/search-routing-policy.d.ts +18 -0
- package/lib/types/search-routing-remote.d.ts +20 -0
- package/lib/types/search-routing.d.ts +43 -0
- package/lib/types/search-tool-routing.d.ts +22 -0
- package/lib/types/serialize.d.ts +106 -0
- package/lib/types/session-accounts-controller.d.ts +20 -0
- package/lib/types/session-accounts-host.d.ts +33 -0
- package/lib/types/session-accounts-remote.d.ts +114 -0
- package/lib/types/session-accounts.d.ts +38 -0
- package/lib/types/session-continuation-host.d.ts +45 -0
- package/lib/types/session-continuation-remote.d.ts +22 -0
- package/lib/types/session-continuation-types.d.ts +62 -0
- package/lib/types/session-continuation-ui.d.ts +49 -0
- package/lib/types/session-continuation.d.ts +14 -0
- package/lib/types/settings-reader.d.ts +16 -0
- package/lib/types/sse.d.ts +28 -0
- package/lib/types/temporary-models.d.ts +25 -0
- package/lib/types/tool-schema-compat.d.ts +17 -0
- package/lib/types/traditional-search.d.ts +17 -0
- package/lib/types/turn-model-provenance.d.ts +45 -0
- package/lib/types/turn-request-models.d.ts +55 -0
- package/lib/types/turn-selection-card.d.ts +24 -0
- package/lib/types/turn-selection-host.d.ts +35 -0
- package/lib/types/turn-selection-remote.d.ts +17 -0
- package/lib/types/turn-selection.d.ts +186 -0
- package/lib/types/turn-usage-evidence.d.ts +66 -0
- package/lib/types/turn-usage-notice.d.ts +7 -0
- package/lib/types/types.d.ts +69 -0
- package/lib/types/usage.d.ts +26 -0
- package/lib/types/watchdog.d.ts +22 -0
- package/lib/types/web-delegate.d.ts +12 -0
- package/lib/types/web-search-routing-card.d.ts +16 -0
- package/lib/types/web-search-routing-config.d.ts +12 -0
- package/lib/types/wire-anthropic.d.ts +37 -0
- package/lib/types/wire.d.ts +56 -0
- package/lib/web-delegate.js +9 -0
- package/output/auto-model-mockup/index.html +207 -0
- package/package.json +261 -0
- package/scripts/analyze-diagnostics.mjs +91 -0
- package/scripts/check-search-composition.mjs +191 -0
- package/scripts/iteration-review.mjs +53 -0
- package/scripts/release-policy.mjs +353 -0
- package/scripts/select-core-release.mjs +11 -0
|
@@ -0,0 +1,341 @@
|
|
|
1
|
+
# Contextual Auto routing and captured selection reasons
|
|
2
|
+
|
|
3
|
+
Tracking: [#258](https://github.com/cloga/dsh-github-copilot/issues/258) and
|
|
4
|
+
[evidence refinement #262](https://github.com/cloga/dsh-github-copilot/issues/262).
|
|
5
|
+
|
|
6
|
+
**Current contract:** source candidate `0.4.2-alpha.10`, targeting official DSH/Windows
|
|
7
|
+
Desktop `0.2.0-rc.2`. Semantic assessment is enabled by default with bounded
|
|
8
|
+
auxiliary timing/output evidence. Neither publication nor
|
|
9
|
+
these explanations prove the version loaded in a particular Host.
|
|
10
|
+
|
|
11
|
+
## Using Auto
|
|
12
|
+
|
|
13
|
+
Choose **Auto · Balance**, **Auto · Efficiency**, or **Auto · Intelligence**
|
|
14
|
+
under the managed **GitHub Copilot** provider in DSH's ordinary picker. This is
|
|
15
|
+
the plugin's DSH routing policy, not Copilot App's Auto or GitHub's private
|
|
16
|
+
router. Selecting a real model exits Auto. Every step/retry in an admitted
|
|
17
|
+
turn keeps its chosen model; the next turn can choose differently.
|
|
18
|
+
|
|
19
|
+
Semantic assessment is already enabled for locally unknown tasks. To opt out,
|
|
20
|
+
set `autoSemanticAssessment: false` in the plugin's `github-copilot` settings
|
|
21
|
+
namespace; it is not an extra switch on the Web search card. This changes
|
|
22
|
+
subsequent assessments, not an already-admitted turn. Local known-task rules
|
|
23
|
+
still work when auxiliary inference is disabled.
|
|
24
|
+
|
|
25
|
+
The native plugin configuration exposes these profile-wide settings under
|
|
26
|
+
`github-copilot` (not the Web search routing card):
|
|
27
|
+
|
|
28
|
+
| Setting | Default | Meaning |
|
|
29
|
+
|---|---|---|
|
|
30
|
+
| `autoSemanticAssessmentTimeoutMs` | `30000` | Integer total preparation/response budget from 1000 to 120000 ms. `8000` restores the earlier budget; zero/unlimited is not supported. |
|
|
31
|
+
| `autoSemanticAssessmentModel` | `''` | Empty uses metadata-driven automatic classifier selection. A nonempty value is an exact current-account model ID, not a display name or virtual Auto choice. |
|
|
32
|
+
|
|
33
|
+
For example, keep automatic selection while allowing the new default budget:
|
|
34
|
+
|
|
35
|
+
```yaml
|
|
36
|
+
github-copilot:
|
|
37
|
+
autoSemanticAssessment: true
|
|
38
|
+
autoSemanticAssessmentTimeoutMs: 30000
|
|
39
|
+
autoSemanticAssessmentModel: ''
|
|
40
|
+
```
|
|
41
|
+
|
|
42
|
+
These fields require the new plugin release; adding them to an older version
|
|
43
|
+
does not implement the behavior. Changes apply to subsequent assessments,
|
|
44
|
+
not an admitted turn's steps/retries. The 30-second default is a user-approved
|
|
45
|
+
trial, not calibrated latency or proof that past unknown results were timeouts.
|
|
46
|
+
Larger budgets can increase time before the answer begins and auxiliary charges.
|
|
47
|
+
Native Chat/search deadlines and user cancellation are independent.
|
|
48
|
+
|
|
49
|
+
An explicit classifier may use any account-advertised category, but still must
|
|
50
|
+
be an entitled, unexcluded concrete model with text capability, validated protocol
|
|
51
|
+
and the existing input/output headroom. A missing, excluded or unsuitable choice
|
|
52
|
+
produces `COPILOT_AUTO_CLASSIFIER_CONFIGURED_MODEL_UNAVAILABLE`, then the existing
|
|
53
|
+
unknown-demand/unavailable diagnostic; no silent alternate classifier is called.
|
|
54
|
+
Answer selection remains a separate Auto decision. The classifier result does
|
|
55
|
+
not choose its own answer model.
|
|
56
|
+
|
|
57
|
+
Open the reply's selection explanation for the captured decision. Evidence is
|
|
58
|
+
Host-lifetime and bounded; after restart or eviction it can be unavailable.
|
|
59
|
+
Do not infer an old decision from the current picker.
|
|
60
|
+
|
|
61
|
+
| Reading | Meaning |
|
|
62
|
+
|---|---|
|
|
63
|
+
| `uncertain` / unknown demand | The available evidence does not establish task demand; it is not a failed answer or a claim that the conversation is empty. |
|
|
64
|
+
| `insufficient-evidence` | A local rule or classifier could not justify a demand category. Longer context alone does not guarantee classification. |
|
|
65
|
+
| Semantic timeout | No validated result within the captured configured end-to-end budget (30 seconds by default). Demand remains unknown; Efficiency/Balance target Versatile, Intelligence targets Powerful, subject to eligibility and category fallback. |
|
|
66
|
+
| Missing auxiliary milestone | That milestone was not observed before assessment settled, not proof the supplier never sent data. |
|
|
67
|
+
| Selection unknown/unavailable | Captured selection evidence is absent or could not be read, distinct from unknown task demand. |
|
|
68
|
+
|
|
69
|
+
There is no assessment retry or automatic deadline extension. Caller cancellation
|
|
70
|
+
aborts the turn rather than invoking fallback. Auxiliary calls can add latency
|
|
71
|
+
and supplier charges outside native answer Usage.
|
|
72
|
+
|
|
73
|
+
Classifier capability checks accept only explicit account-advertised disable
|
|
74
|
+
labels: `off`, protocol-native `none` for Responses/Completions, or `disabled`
|
|
75
|
+
for Anthropic. No aliases are applied to ordinary Chat model maps. Missing or
|
|
76
|
+
empty effort metadata does not prove absence of reasoning.
|
|
77
|
+
|
|
78
|
+
Strict request-local `semantic.candidateScan` evidence records at most 512
|
|
79
|
+
candidates, eligible counts and first-rejection counts for exclusions, explicit
|
|
80
|
+
ID mismatch, category, text input, unproven disable capability and budget.
|
|
81
|
+
Counts partition the examined pool, not all possible rejection reasons per model.
|
|
82
|
+
No candidate IDs, prompts or raw metadata are retained in these counts.
|
|
83
|
+
`preparationFailure: native-off-unavailable` separates native off preparation
|
|
84
|
+
from no-fit selection. No-fit Host warnings use fixed codes only. These fields
|
|
85
|
+
are ephemeral selection evidence, not a new persisted diagnostic population;
|
|
86
|
+
absent legacy evidence cannot be reconstructed.
|
|
87
|
+
|
|
88
|
+
## Requirements
|
|
89
|
+
|
|
90
|
+
Auto must explain why the initiating turn chose its actual route, not display a
|
|
91
|
+
generic description of prompt length. Intelligence prioritizes response quality
|
|
92
|
+
but can use a lightweight model for demonstrably simple work. Efficiency favors
|
|
93
|
+
economical capability without knowingly underpowering difficult work. Balance
|
|
94
|
+
trades those priorities. All preferences share the same account model pool.
|
|
95
|
+
|
|
96
|
+
The authenticated account `/models` response supplies `model_picker_category`.
|
|
97
|
+
The October 3, 2026 bounded live probe observed `powerful`, `versatile` and
|
|
98
|
+
`lightweight`; some records lacked the field. This is supplier classification,
|
|
99
|
+
not measured latency, quality, price or health. Official documentation's
|
|
100
|
+
`category` is a different source and must not override live account facts.
|
|
101
|
+
Do not derive category from IDs, names, reasoning effort or context capacity.
|
|
102
|
+
|
|
103
|
+
Hard constraints remain first: entitlement, exclusions, current proof, text/image
|
|
104
|
+
capability and actual image format, estimated input headroom and final native
|
|
105
|
+
admission. No preference overrides these. No new Core owner, adapter, credential
|
|
106
|
+
record, general transport, global model mutation or durable decision event.
|
|
107
|
+
|
|
108
|
+
## Task assessment
|
|
109
|
+
|
|
110
|
+
Assessment concerns the task in context, not the last sentence's length.
|
|
111
|
+
An isolated greeting may be low demand. A short request to prove a theorem or
|
|
112
|
+
continue an unresolved investigation is not evidence of low demand. Long input
|
|
113
|
+
can be mechanical. Prior assistant/tool work prevents a greeting/continuation
|
|
114
|
+
shortcut from claiming an isolated simple task.
|
|
115
|
+
|
|
116
|
+
Use conservative local assessment first. Unknown is an explicit outcome, never
|
|
117
|
+
silently translated into simple. Semantic assessment is enabled by default for
|
|
118
|
+
locally unknown demand at the user's explicit rollout request (#267), not
|
|
119
|
+
because labeled-corpus calibration has completed. Set
|
|
120
|
+
`github-copilot.autoSemanticAssessment: false` to disable auxiliary inference;
|
|
121
|
+
an existing explicit false remains respected. It must not claim equivalence to
|
|
122
|
+
GitHub's private router or calibrated confidence.
|
|
123
|
+
|
|
124
|
+
Local rules recognize an isolated, explicit JSON-to-CSV/CSV-to-JSON conversion
|
|
125
|
+
or line sort only when the entire request is at most 1600 characters, contains
|
|
126
|
+
one fenced payload and no other work or non-text content. This establishes
|
|
127
|
+
routine, not simple. An exact acknowledgement/continue request can inherit
|
|
128
|
+
explicit difficult-task evidence from the nearest non-continuation user request
|
|
129
|
+
among the preceding twelve user requests. A new task or an expired anchor stops
|
|
130
|
+
inheritance; simple/routine demand is never inherited.
|
|
131
|
+
|
|
132
|
+
`uncertain` remains a valid task result. `insufficient-evidence` means the local
|
|
133
|
+
rules cannot establish demand, not that the conversation has no context.
|
|
134
|
+
Attachments, unfamiliar instructions and genuinely ambiguous continuations can
|
|
135
|
+
still remain unknown. These refinements do not change category preference,
|
|
136
|
+
previous-model eligibility or semantic safeguards.
|
|
137
|
+
|
|
138
|
+
The semantic assessment uses one fixed concrete account candidate through the
|
|
139
|
+
existing managed native adapter, never an Auto ID and never `ctx.llm` routing
|
|
140
|
+
recursion. It has no tools, bounded text-only assessment context, bounded JSON
|
|
141
|
+
output, one deadline, cancellation and no new retry loop. Text from messages is
|
|
142
|
+
untrusted data, not router instructions. Files/images/reasoning/replay bodies
|
|
143
|
+
are not read; limited task text cannot justify downshifting. Assessment input is a separate projection: the actual chat
|
|
144
|
+
messages, system prompt, tools, attachments and history stay untouched.
|
|
145
|
+
|
|
146
|
+
Only locally unknown tasks invoke the enabled assessment. Context packing keeps
|
|
147
|
+
the current and nearest preceding user requests before recent output. An explicit
|
|
148
|
+
continuation also reserves the nearest substantive user anchor among the previous
|
|
149
|
+
twelve user messages and the intervening user chain, then restores chronological
|
|
150
|
+
order. No expired anchor, task summary or semantic task boundary is invented.
|
|
151
|
+
At most twelve complete JSON rows, 1600 text
|
|
152
|
+
characters per row and 8000 serialized characters are retained; JSON escape
|
|
153
|
+
expansion is included in this bound. Any truncation, skipped row or non-text
|
|
154
|
+
content still marks projection omission. A missing current request/continuation
|
|
155
|
+
anchor, a missing reserved user row, truncated reserved task text, or nontext
|
|
156
|
+
content in the reserved user chain separately marks `taskTextLimited` and prevents
|
|
157
|
+
a semantic simple/routine downshift. Incidental assistant/tool/history omissions
|
|
158
|
+
alone no longer force that downgrade. The classifier must still judge whether
|
|
159
|
+
the retained evidence establishes demand; retained text is not semantic
|
|
160
|
+
completeness and missing required details must remain unknown.
|
|
161
|
+
|
|
162
|
+
The instruction requests a schema without a pre-filled unknown answer and
|
|
163
|
+
distinguishes projected data omissions from missing relevant task evidence.
|
|
164
|
+
Unknown is permitted for genuinely indeterminate demand, not required merely
|
|
165
|
+
because a long conversation exceeds the projection bounds. No reasoning/tool
|
|
166
|
+
enablement, second inference, larger context/output budget or full-history
|
|
167
|
+
serialization is introduced.
|
|
168
|
+
|
|
169
|
+
The strict result contains demand (`simple`, `routine`, `complex`, `unknown`)
|
|
170
|
+
and finite evidence signals, not a model ID, raw reasoning or an uncalibrated
|
|
171
|
+
confidence percentage. Invalid/truncated/failed results retain the local
|
|
172
|
+
assessment with an explicit diagnostic. Caller cancellation aborts the turn
|
|
173
|
+
instead of falling back. Credential, metadata and account revocation remain
|
|
174
|
+
fatal through existing lease/admission guards. Extra calls incur supplier
|
|
175
|
+
charges and latency, separate from native chat Usage and without inferred cost.
|
|
176
|
+
|
|
177
|
+
Strict ephemeral explanation evidence optionally records retained row count
|
|
178
|
+
(0–12), the projection omission flag, task-text limitation, fixed continuation
|
|
179
|
+
anchor status and the classifier's decoded demand. No text, attachment values
|
|
180
|
+
or new persistent population is recorded. Legacy absent fields stay absent.
|
|
181
|
+
These fields distinguish a valid classifier `unknown` from plugin-side
|
|
182
|
+
simple/routine blocking; they do not establish accuracy or a live unknown rate.
|
|
183
|
+
|
|
184
|
+
## Auxiliary latency evidence and bounded output (#272)
|
|
185
|
+
|
|
186
|
+
The assessment has one **configured end-to-end deadline, 30 seconds by default**, including account
|
|
187
|
+
discovery/preparation and the native call. It does not retry on timeout or try a
|
|
188
|
+
second classifier. The response budget is now 128 output tokens, with compact
|
|
189
|
+
demand JSON and at most three fixed signals requested. The strict decoder
|
|
190
|
+
retains its seven-signal compatibility bound; raw explanation, confidence
|
|
191
|
+
percentages and reasoning are not requested. Truncation remains failure, never
|
|
192
|
+
accepted partial JSON.
|
|
193
|
+
|
|
194
|
+
One monotonic deadline is checked after context packing, discovery and native
|
|
195
|
+
preparation, before native request admission, on delivered chunks and before
|
|
196
|
+
accepting a decoded result. A delayed event-loop timer cannot authorize a
|
|
197
|
+
late auxiliary request or result. It can still delay observed settlement:
|
|
198
|
+
this is not a hard real-time guarantee or evidence of a supplier latency cause.
|
|
199
|
+
The owned timer and abort listener are disposed when assessment settles.
|
|
200
|
+
|
|
201
|
+
Among eligible supplier Lightweight text candidates with the existing input
|
|
202
|
+
headroom and explicit reasoning-disable proof, automatic selection prefers
|
|
203
|
+
unmarked cost, then deterministic ID ordering. An explicit exact-ID setting selects only that
|
|
204
|
+
eligible model, without category or cost re-ranking.
|
|
205
|
+
This does not infer speed, health or quality from effort or model names, and
|
|
206
|
+
does not launch a second candidate if native support is absent.
|
|
207
|
+
|
|
208
|
+
Automatic selection is metadata-driven, not an empirical fastest/best model
|
|
209
|
+
search. No additional benchmark requests, adaptive deadlines, per-model health
|
|
210
|
+
ranking, cooldown or cross-turn cache are introduced. Those require separately
|
|
211
|
+
reviewed representative latency/accuracy/cost observations and consent, not a
|
|
212
|
+
guess from model names or a timeout suspicion. Review outcomes under #334 before
|
|
213
|
+
claiming the increased default improves live classification.
|
|
214
|
+
|
|
215
|
+
The classifier is eligible only when current account metadata advertises
|
|
216
|
+
an explicit protocol-compatible disable mode and the isolated, public prepared
|
|
217
|
+
model exposes `off`. It then
|
|
218
|
+
always requests `off`; otherwise assessment fails closed without dispatching
|
|
219
|
+
the classifier. The classifier-only model mapping serializes the native
|
|
220
|
+
disable form supported by the published adapter: Responses
|
|
221
|
+
`reasoning.effort: "none"`, OpenAI-compatible `reasoning_effort: "none"`, or
|
|
222
|
+
Anthropic `thinking.type: "disabled"`. This mapping is not added to shared
|
|
223
|
+
account models, so ordinary Chat behavior remains unchanged.
|
|
224
|
+
|
|
225
|
+
The auxiliary adapter also does not inherit configured Chat reasoning effort:
|
|
226
|
+
its isolated profile omits that default while ordinary Chat retains it. Tools
|
|
227
|
+
and an execution loop are absent. These checks establish the requested wire
|
|
228
|
+
payload in synthetic tests, not that every supplier honors the control
|
|
229
|
+
internally. Models lacking either advertised or native support are not used;
|
|
230
|
+
do not substitute `low` or `minimal`, which can enable reasoning.
|
|
231
|
+
|
|
232
|
+
Tools and an execution loop are absent. Only locally unknown tasks use one
|
|
233
|
+
classifier call per admitted turn; subsequent steps/retries reuse its result.
|
|
234
|
+
Text projection stays bounded to twelve rows, 1600 characters per row and
|
|
235
|
+
8000 serialized characters; files, images and replay bodies are not included.
|
|
236
|
+
The requested output budget remains 128 tokens. Neither those bounds nor the
|
|
237
|
+
waiting deadline establishes a hard supplier reasoning-token, credit or dollar
|
|
238
|
+
cap. Measure accuracy, elapsed time and auxiliary usage separately before
|
|
239
|
+
reducing context/output or claiming a cheapest/fastest classifier.
|
|
240
|
+
|
|
241
|
+
Captured Host-lifetime selection evidence now includes a bounded public
|
|
242
|
+
classifier ID, configured waiting budget, total monotonic assessment elapsed
|
|
243
|
+
time, adapter-invocation start, first nonempty text time, native stop/non-stop
|
|
244
|
+
finish observation and strict result-validation status. Preparation is included
|
|
245
|
+
in total time; adapter start is **not HTTP dispatch**, first text is **not first
|
|
246
|
+
network byte**, and these milestones cannot distinguish connection, supplier
|
|
247
|
+
queueing or computation. Absence says no milestone was observed before
|
|
248
|
+
settlement, not that the provider never emitted data. Counts are characters,
|
|
249
|
+
not tokens/credits; no prompt, text, replay, credential, error body or hash is
|
|
250
|
+
captured. Late callbacks cannot rewrite a settled decision.
|
|
251
|
+
|
|
252
|
+
The existing strict explicit-Session/turn Remote carries this optional evidence;
|
|
253
|
+
old explanations remain decodable without it. Numeric/model evidence is copied
|
|
254
|
+
and frozen in the bounded store and routing decision, not durably appended or
|
|
255
|
+
reconstructed after restart. The footer says assessment timed out, demand
|
|
256
|
+
remained unknown, and preference policy selected the category as a fallback.
|
|
257
|
+
Progressive details show milestones separately from native answer Usage.
|
|
258
|
+
Continuity copy states the fitting category-pool count and explicitly disclaims
|
|
259
|
+
semantic ranking; all-input-fit counts are not the final category pool.
|
|
260
|
+
|
|
261
|
+
This is instrumentation and bounded output optimization, not a measured latency
|
|
262
|
+
improvement or classifier calibration. No latency-ranked classifier pool,
|
|
263
|
+
cross-turn assessment cache, cooldown, adaptive deadline or extra probe is
|
|
264
|
+
introduced without reviewing real observations. Existing local evidence,
|
|
265
|
+
omitted-context downshift protection, cancellation, account/proof revocation,
|
|
266
|
+
same-turn single-flight, eligibility and category policy remain unchanged.
|
|
267
|
+
|
|
268
|
+
## Category policy and continuity
|
|
269
|
+
|
|
270
|
+
| Task demand | Efficiency | Balance | Intelligence |
|
|
271
|
+
| --- | --- | --- | --- |
|
|
272
|
+
| Proven simple | Lightweight | Lightweight | Lightweight |
|
|
273
|
+
| Routine | Lightweight | Versatile | Powerful |
|
|
274
|
+
| Complex | Powerful | Powerful | Powerful |
|
|
275
|
+
| Unknown | Versatile | Versatile | Powerful |
|
|
276
|
+
|
|
277
|
+
These are declared routing policies, not benchmark rankings. Efficiency's
|
|
278
|
+
routine lightweight target is an economical approximation, not proof of
|
|
279
|
+
equal output quality. Intelligence never treats short length alone as simple.
|
|
280
|
+
|
|
281
|
+
The target and remaining categories have this declared weight order:
|
|
282
|
+
powerful -> versatile -> lightweight; versatile -> powerful -> lightweight;
|
|
283
|
+
lightweight -> versatile -> powerful. Unclassified candidates retain a small
|
|
284
|
+
explicit opportunity, not fabricated categories. Unknown/new category values
|
|
285
|
+
remain unclassified with a bounded catalog diagnostic.
|
|
286
|
+
|
|
287
|
+
Across every eligible fitting category, use stable positive weighted Session/turn
|
|
288
|
+
allocation. The task/preference matrix is a preferred target, not exclusion.
|
|
289
|
+
Multiply per-model cost/recency weights by category factors 1 / 0.25 / 0.05
|
|
290
|
+
in the existing target/second/third order; unknown category uses 0.02 explicitly.
|
|
291
|
+
These factors are an uncalibrated trial. Ordinary weight 1 has finite
|
|
292
|
+
previous-model multiplier 1.5; high-cost base weight 0.2 decays tenfold to 0.02
|
|
293
|
+
only for the previous model. Every candidate keeps positive opportunity; a lone
|
|
294
|
+
candidate still has share 1. A different model's turn resets the penalty, without
|
|
295
|
+
extra streak state. Cost is independent of task difficulty and supplier
|
|
296
|
+
category; input fit does not prove task success. Non-target selection is
|
|
297
|
+
intentional weighted opportunity, not absence of a target candidate or proof
|
|
298
|
+
of task suitability. Never automatically escalate to establish whether
|
|
299
|
+
another model can solve the task.
|
|
300
|
+
These are reviewed policy parameters, not measured quality or prices. Once
|
|
301
|
+
admitted, all steps/retries share the decision. No supplier cache savings are
|
|
302
|
+
claimed. See [high-cost requirements](./auto-high-cost.md) and
|
|
303
|
+
[evidence-driven iteration](./evidence-driven-iteration.md).
|
|
304
|
+
|
|
305
|
+
If no model fits, preserve the existing largest-admissible-input recovery
|
|
306
|
+
branch and its native compaction diagnostics. Explain that no candidate fits;
|
|
307
|
+
never label this as a successful preference match or silently trim history.
|
|
308
|
+
|
|
309
|
+
## Explanation contract
|
|
310
|
+
|
|
311
|
+
Capture task source/demand/signals, assessment diagnostic, target/selected
|
|
312
|
+
category, fitting cross-category count, target absence and continuity/tie-break or
|
|
313
|
+
no-fit selection at decision time. Client reads the explicit viewed Session and
|
|
314
|
+
turn through the existing strict Remote. It does not rerun classification or
|
|
315
|
+
consult today's picker. Show a decisive first sentence, bounded supporting
|
|
316
|
+
facts and optional policy detail. Do not repeat the model name or change Usage.
|
|
317
|
+
|
|
318
|
+
Legacy records without detailed evidence show that the reason was not retained.
|
|
319
|
+
Host-lifetime evidence may be lost on restart; do not invent durable history or
|
|
320
|
+
promise cold-history reconstruction.
|
|
321
|
+
|
|
322
|
+
## Acceptance and rollout
|
|
323
|
+
|
|
324
|
+
Regress category normalization (missing/new/invalid values), short difficult
|
|
325
|
+
tasks, isolated greetings, contextual continuation, long mechanical input,
|
|
326
|
+
category fallback, previous-model eligibility, deterministic tie breaking,
|
|
327
|
+
no-fit recovery, one-call semantic success/invalid/failure/timeout/cancellation,
|
|
328
|
+
omitted context, strict Remote and footer English/Chinese explanations.
|
|
329
|
+
|
|
330
|
+
Retain native gateway, parent following, manual selection, exclusion admission,
|
|
331
|
+
account proof, image admission and compaction fixture coverage. Full protected
|
|
332
|
+
Windows/Linux CI remains mandatory before merge/release. No live inference
|
|
333
|
+
calls are needed for synthetic acceptance.
|
|
334
|
+
|
|
335
|
+
Default enablement is a user-authorized rollout, not an accuracy qualification.
|
|
336
|
+
Evaluate a reviewed labeled corpus for complex-to-simple errors, extra latency,
|
|
337
|
+
token use and stability across languages and prompt injection before claiming
|
|
338
|
+
calibrated performance. Synthetic passing tests are not this evaluation.
|
|
339
|
+
Missing settings adopt the enabled default after upgrade; explicit false is
|
|
340
|
+
not overwritten and no profile migration is performed. Publication does not
|
|
341
|
+
authorize profile installation or restart.
|
|
@@ -0,0 +1,236 @@
|
|
|
1
|
+
# GitHub Copilot Auto routing contracts
|
|
2
|
+
|
|
3
|
+
Target: official DSH / Windows Desktop `0.2.0-rc.2`. Implementation is plugin-only.
|
|
4
|
+
For ordinary use, start with [task assessment and explanations](./auto-task-routing.md)
|
|
5
|
+
and [parent following](./parent-model-follow.md).
|
|
6
|
+
|
|
7
|
+
## User contract
|
|
8
|
+
|
|
9
|
+
The managed `github-copilot-preview` group provides **Auto · Balance** (`auto`),
|
|
10
|
+
**Auto · Efficiency** and **Auto · Intelligence** in Core's flat picker.
|
|
11
|
+
Only a real model selection exits Auto. Native selection also saves the future
|
|
12
|
+
global default; the plugin does not change that Core behavior.
|
|
13
|
+
|
|
14
|
+
Each admitted Core turn resolves to one real account model. Tool continuations,
|
|
15
|
+
native request retries and compaction retries retain it; a new user or Goal
|
|
16
|
+
turn can choose differently. Transport, replay, native Usage and billing retain
|
|
17
|
+
the actual model identity. Auto never crosses providers/accounts or changes an
|
|
18
|
+
explicit real-model selection.
|
|
19
|
+
|
|
20
|
+
The virtual preference normally remains the Session's pending selection while actual
|
|
21
|
+
request headers record real models. A strict plugin-owned public Session projection
|
|
22
|
+
also retains the latest owner-explicit native `model/selection`: if a virtual
|
|
23
|
+
request header consumes native pending, later concrete headers cannot erase that
|
|
24
|
+
Auto intent. Cold refolding restores intent, not old decisions or reasons.
|
|
25
|
+
Copied fork/child prefixes alone do not establish an owner-explicit choice;
|
|
26
|
+
native pending and the existing parent-following rules remain independent.
|
|
27
|
+
Missing or malformed continuity evidence produces a named diagnostic instead of
|
|
28
|
+
guessing Auto from a real model. A new explicit fixed or other-provider selection
|
|
29
|
+
replaces intent for future turns, even after its native pending is consumed.
|
|
30
|
+
Already admitted Auto turns retain their real route through late picker changes.
|
|
31
|
+
This projection neither writes history/settings nor changes native projection ownership.
|
|
32
|
+
For an otherwise unselected Session that
|
|
33
|
+
inherits default Auto, the first request persists the exact virtual
|
|
34
|
+
`model/selection` intent. A later explicit selection wins. Existing fixed
|
|
35
|
+
Sessions are not converted because the global default changes.
|
|
36
|
+
|
|
37
|
+
## Eligibility and input fit
|
|
38
|
+
|
|
39
|
+
Current authenticated account metadata establishes entitlement, endpoint,
|
|
40
|
+
context/input/output limits, tool/image support and reasoning capabilities.
|
|
41
|
+
Missing/invalid metadata produces diagnostics, not guessed capabilities.
|
|
42
|
+
Exact exclusions and independent account/proof/cancellation guards apply first.
|
|
43
|
+
|
|
44
|
+
The public native TokenMeter estimates every entered message, including
|
|
45
|
+
reasoning and tool history. Its current-surface measurement is an envelope
|
|
46
|
+
floor, not a substitute for candidate-specific final serialization.
|
|
47
|
+
Missing/invalid measurement fails with
|
|
48
|
+
`COPILOT_AUTO_TOKEN_METER_UNAVAILABLE` /
|
|
49
|
+
`COPILOT_AUTO_TOKEN_ESTIMATE_INVALID`.
|
|
50
|
+
|
|
51
|
+
Candidates fit when estimated input does not exceed the hard budget after
|
|
52
|
+
output reservation and safety allowance. If none fits, choose the largest
|
|
53
|
+
input budget for native recovery and explain whether content is fixed,
|
|
54
|
+
compaction unavailable, already attempted, or eligible. This is not permission
|
|
55
|
+
to dispatch an oversized request. The native managed `inspectRequest` guard
|
|
56
|
+
remains authoritative after final framing/attachment projection.
|
|
57
|
+
|
|
58
|
+
Task preference uses supplier `model_picker_category`, not capacity, model
|
|
59
|
+
names or reasoning effort as quality rankings. [Task policy](./auto-task-routing.md)
|
|
60
|
+
owns the demand matrix, default semantic assessment, category fallback,
|
|
61
|
+
high-cost one-turn recency decay, ordinary finite continuity and stable
|
|
62
|
+
positive-weight cross-category allocation. Every eligible fitting model has
|
|
63
|
+
positive opportunity; the task target biases per-model weights instead of
|
|
64
|
+
hard-filtering other categories. Unknown category stays explicit.
|
|
65
|
+
User-marked high cost is orthogonal to demand and category; see
|
|
66
|
+
[high-cost requirements](./auto-high-cost.md). Future policy changes must read
|
|
67
|
+
the [latest review](./auto-iteration-review.json) before implementation.
|
|
68
|
+
|
|
69
|
+
## Images, compaction and search
|
|
70
|
+
|
|
71
|
+
Image eligibility includes historical user/tool image blocks in all entered
|
|
72
|
+
messages, not only current user text. Text-only candidates are removed; Auto
|
|
73
|
+
advertises image input only with an eligible image model. Actual fixed/Auto
|
|
74
|
+
dispatch independently checks explicit account MIME restrictions against
|
|
75
|
+
Core's projected image. Missing MIME lists stay unverified.
|
|
76
|
+
[Image admission limits](./image-input-compatibility.md).
|
|
77
|
+
|
|
78
|
+
Auto does not trim history or own compaction transactions. Estimated pressure
|
|
79
|
+
uses the initiating Agent's public bound recovery service and preserves the
|
|
80
|
+
same real model. If the summary input already exceeds every eligible budget,
|
|
81
|
+
stock one-shot compaction cannot rescue it. [Budget admission](./copilot-compaction.md)
|
|
82
|
+
and [manual recovery](./manual-compaction-recovery.md) describe separate paths.
|
|
83
|
+
|
|
84
|
+
Chat Auto does not own independent hosted search. A nonempty `searchModel`
|
|
85
|
+
override remains authoritative; otherwise that provider uses bounded
|
|
86
|
+
account-owned Responses selection and capability proof. Auto-following search
|
|
87
|
+
sees the initiating real request route. No virtual model reaches transport,
|
|
88
|
+
and the final query is not replayed across candidates.
|
|
89
|
+
[Search routing](./session-search-routing.md).
|
|
90
|
+
|
|
91
|
+
## Attribution and explanation
|
|
92
|
+
|
|
93
|
+
Completed replies show **Auto (preference)** with captured reasons, **Manual**
|
|
94
|
+
only with explicit fixed-selection evidence, or **Selection unknown** when a
|
|
95
|
+
successful read has no record. Failed reads show **Selection unavailable → Retry**.
|
|
96
|
+
Retry rereads the same Session/turn; it never runs inference.
|
|
97
|
+
|
|
98
|
+
The public assistant-actions slot adds the plugin item after native Usage/time.
|
|
99
|
+
It cannot change the native fixed-height parent row and adds no Model details
|
|
100
|
+
or footer button. The existing info dialog is **Turn model and selection evidence**.
|
|
101
|
+
When native Usage lacks complete model routes, model evidence comes first and the
|
|
102
|
+
original selection explanation remains in a collapsed disclosure below:
|
|
103
|
+
|
|
104
|
+
- **Recorded models** lists independent successful same-turn `assistant/message`
|
|
105
|
+
sources, preserving multiple models/providers. Failed attempts, retries and
|
|
106
|
+
paged/missing history retain an explicit incomplete-attribution warning; known
|
|
107
|
+
sources do not prove which models every failed attempt used.
|
|
108
|
+
- **Requested model** is the fallback when no successful source exists. An
|
|
109
|
+
independent public Session projection folds native `turn/start`, `step/start`,
|
|
110
|
+
`request/header`, `step/end` and `turn/end`; only headers inside a recorded open
|
|
111
|
+
step bind to that turn. Native headers have no turn field and are not emitted
|
|
112
|
+
for every unchanged request. Therefore absent headers stay unknown, never
|
|
113
|
+
inherited from an earlier turn or today's `Session.requestHeader()`. Configuration
|
|
114
|
+
proves neither dispatch nor execution, usage or billing.
|
|
115
|
+
- Missing/invalid evidence stays **Unknown**. Requested-model read failures are
|
|
116
|
+
distinct from missing evidence and have same-Session/turn read-only Retry.
|
|
117
|
+
Selection mode and model evidence remain independent.
|
|
118
|
+
|
|
119
|
+
Complete native Usage routes avoid redundant model presentation. A turn without
|
|
120
|
+
a closing message has no assistant-actions anchor; pure failed turns without one
|
|
121
|
+
are not promised an injected entry. Optional conversation projections do not
|
|
122
|
+
control selection mounting. The request projection has its own `stateVersion: 1`,
|
|
123
|
+
retains at most 128 turns and 32 distinct routes per turn (overflow invalidates
|
|
124
|
+
that turn), and stores no message/tool/replay/credential content. Cold folding
|
|
125
|
+
uses existing recorded events only, not custom events or history conversion.
|
|
126
|
+
|
|
127
|
+
The bounded Host store retains 64 Agents and 128 turns per Agent. Disposal,
|
|
128
|
+
eviction and restart lose new records. Compatible historical Auto events remain
|
|
129
|
+
readable; missing historical preference/reasons are not reconstructed.
|
|
130
|
+
New optional `github-copilot/auto-model-decision` events are **not** emitted:
|
|
131
|
+
official rc.2 `Session.append()` has no public ignorable-envelope option.
|
|
132
|
+
|
|
133
|
+
`TurnSelectionController` extends public `TypertRemoteService` with `@Remote get`.
|
|
134
|
+
Client calls carry explicit viewed Session ID and turn, with no automatic
|
|
135
|
+
Client scope projection that would remove the ID. Native Host Agent lookup,
|
|
136
|
+
access checks and strict codecs remain. Additive `requestedModels` preserves the
|
|
137
|
+
existing `get` descriptor and strict selection codec unchanged. Actual Client/Host gateway regressions
|
|
138
|
+
cover reachability, bound contexts and missing/denied identities; mock RPC
|
|
139
|
+
success is not sufficient evidence.
|
|
140
|
+
|
|
141
|
+
Captured evidence includes task/category, eligible/fitting counts,
|
|
142
|
+
fallback/continuity and optional auxiliary milestones. It proves a decision,
|
|
143
|
+
not execution, current occupancy, usage or supplier billing. Native accounting
|
|
144
|
+
remains independent. [Usage limits](./copilot-usage.md).
|
|
145
|
+
|
|
146
|
+
### Recovering affected histories
|
|
147
|
+
|
|
148
|
+
Only histories containing the old incompatible decision event need this
|
|
149
|
+
compatibility utility. Upgrading prevents new writes but does not repair old logs.
|
|
150
|
+
From a source checkout, use Node 24 LTS and the pinned frozen dependencies:
|
|
151
|
+
|
|
152
|
+
```powershell
|
|
153
|
+
# Read-only validation; no conversation content is printed.
|
|
154
|
+
node scripts\repair-auto-model-history.mjs 'C:\absolute\path\session.v4.jsonl.zstd'
|
|
155
|
+
|
|
156
|
+
# Optional detached copy in a NEW private directory, never the live source.
|
|
157
|
+
node scripts\repair-auto-model-history.mjs 'C:\absolute\path\session.v4.jsonl.zstd' --write-copy 'C:\private\new-recovery-directory'
|
|
158
|
+
```
|
|
159
|
+
|
|
160
|
+
The utility validates the entire v4 log with the official rc.2 catalog, marks
|
|
161
|
+
only recognized decision envelopes `ignorable: true`, and preserves all other
|
|
162
|
+
logical fields. Unknown events, malformed relationships, torn tails and inputs
|
|
163
|
+
over 128 MiB fail closed. Copy mode refuses existing destinations and creates a
|
|
164
|
+
byte-exact original, validated copy and SHA-256 receipt; keep both logs private.
|
|
165
|
+
|
|
166
|
+
The corrected writer preserves a separate first zstd header frame for Desktop
|
|
167
|
+
discovery. The original utility combined header/events and could make a copy
|
|
168
|
+
disappear despite logical validation. `reframedHeader: true` reports physical
|
|
169
|
+
repair; zero changed event sequences alone is not readability proof.
|
|
170
|
+
|
|
171
|
+
**Applying a copy requires separate approval and stopped writers.** Recheck the
|
|
172
|
+
live source against `originalSha256`; regenerate if changed. Replace only that
|
|
173
|
+
log, compare `repairedSha256` and reopen with the official reader. Do not replay
|
|
174
|
+
business requests as a test. The utility never installs, restarts, writes the
|
|
175
|
+
source, clears credentials or performs automatic migration.
|
|
176
|
+
|
|
177
|
+
## Requested parent-to-child selection inheritance
|
|
178
|
+
|
|
179
|
+
The implemented opt-in [profile-wide switch](./parent-model-follow.md) follows
|
|
180
|
+
supported children on each new turn. Fixed parent selection is passed directly;
|
|
181
|
+
Auto passes the exact preference and evaluates the child's own context.
|
|
182
|
+
Child-owned explicit selections and admitted-turn freezing remain authoritative.
|
|
183
|
+
Legacy `parentModelFollow` bindings stay independent.
|
|
184
|
+
|
|
185
|
+
With both policies off, ordinary native children retain Core's creation-time
|
|
186
|
+
route snapshot. Native resolved options/descriptor v3 do not retain whether an
|
|
187
|
+
equal creation route was an explicit override. Enabling the plugin policy
|
|
188
|
+
authorizes replacing that snapshot; it does not reconstruct original intent,
|
|
189
|
+
fabricate descriptor provenance or rewrite Team labels.
|
|
190
|
+
|
|
191
|
+
The original [investigation #229](https://github.com/cloga/dsh-github-copilot/issues/229)
|
|
192
|
+
and [one-switch follow-up #234](https://github.com/cloga/dsh-github-copilot/issues/234)
|
|
193
|
+
record the design history. Upstream native rules/settings UI is separate work,
|
|
194
|
+
not a dependency or claimed shipped feature of this plugin.
|
|
195
|
+
|
|
196
|
+
## Model exclusions and richer picker boundary
|
|
197
|
+
|
|
198
|
+
**Settings → Models → GitHub Copilot → Manage → Model preferences** offers
|
|
199
|
+
search, All/Enabled/Excluded filters and immediate Exclude/Restore actions.
|
|
200
|
+
Temporarily absent excluded IDs remain restorable. Selected models can also be
|
|
201
|
+
excluded; legacy `lockedModelIds` remains empty.
|
|
202
|
+
|
|
203
|
+
`setModelExcluded(modelId, excluded)` returns a narrow strict preferences result
|
|
204
|
+
after path-level native CAS persistence. It reads no credentials, discovers no
|
|
205
|
+
models and enumerates no Sessions. Unknown settings make rows read-only with
|
|
206
|
+
named diagnostics; status-only Retry cannot imply an empty exclusion set.
|
|
207
|
+
The hidden volatile Config leaf persists through native projection/restart.
|
|
208
|
+
|
|
209
|
+
Directory, Auto pools/modalities, search facts and new-turn admission remove
|
|
210
|
+
exact excluded IDs. An admitted Session/turn/model with the exact native request
|
|
211
|
+
signal may continue steps/retries; `turn/end` clears admission. New turns,
|
|
212
|
+
unbound calls and stale picker rows fail `COPILOT_PREVIEW_MODEL_EXCLUDED`.
|
|
213
|
+
Account/proof/cancellation guards remain revocable. A fixed selection is never
|
|
214
|
+
changed automatically; restore it or select another model before the next turn.
|
|
215
|
+
Zero eligible candidates fails `COPILOT_AUTO_NO_ELIGIBLE_MODEL`.
|
|
216
|
+
|
|
217
|
+
Core owns flat grouping and picker search. The old
|
|
218
|
+
[nested-picker prototype](../output/auto-model-mockup/tiers.html) is historical
|
|
219
|
+
design exploration: its capacity quality scores, nested menu and simulated
|
|
220
|
+
responses are not implementation or model evaluations.
|
|
221
|
+
|
|
222
|
+
## Public seams and evidence limits
|
|
223
|
+
|
|
224
|
+
`PreviewAdapter.listModels/resolveModel`, `agent/pre-step`, public model-selection
|
|
225
|
+
middleware and `agent/request` compose the real route before request-header
|
|
226
|
+
persistence. Unresolved virtual Auto fails in adapter preparation/streaming.
|
|
227
|
+
Per-Agent/turn caching, native adapter leases, OAuth, replay and cancellation
|
|
228
|
+
keep their existing owners.
|
|
229
|
+
|
|
230
|
+
Required unchanged rc.2 source/published-artifact fixtures on Windows/Linux
|
|
231
|
+
cover eligibility, real headers, durable intent, multi-step/retry freezing,
|
|
232
|
+
images, native compaction, exclusions and gateway access. They do not establish
|
|
233
|
+
live model availability, quality, latency, price, account access or Desktop
|
|
234
|
+
activation. No Core/dependency/private-registry/shared-catalog/history patch is
|
|
235
|
+
permitted. Measured ranking, cross-provider Auto, failure-driven model switching,
|
|
236
|
+
side-effect replay and native picker redesign remain out of scope.
|