dsh-github-copilot 0.4.0-alpha.80 → 0.4.0-alpha.82

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 CHANGED
@@ -240,6 +240,20 @@ Never say GPT-6/search works merely because settings, typecheck or a package imp
240
240
 
241
241
  ## Changing capabilities
242
242
 
243
+ Completed-turn Usage diagnostics (#259) are read-only plugin conversation data and an additive assistant-actions explanation. Official rc.2 hides its total if any attempt lacks complete accounting; the plugin's finish-only pre-dispatch pressure block can trigger this rule even after recovery. Never fabricate zero usage or replace it with successful-step partial totals. Identify local interception only from its exact recorded finish-only diagnostic, keep other missing samples and paged evidence uncertain, and suppress the explanation whenever native `tokenUsage` exists. Preserve native accounting, history, pressure and recovery; see `docs/copilot-usage.md`.
244
+
245
+ Auto task routing (#258) uses current `model_picker_category` facts, never model
246
+ names, effort or context capacity as a quality category. Hard eligibility/input
247
+ fit precede task/preference policy; unknown category and uncertain task evidence
248
+ remain explicit. Keep suitable previous-model continuity and equal-weight
249
+ tie-breaking within the first available category. Capture actual reasons once
250
+ per turn without durable decision events or changing Usage. The optional
251
+ `autoSemanticAssessment` experiment is off by default: one bounded concrete
252
+ native-adapter request, no tools/Auto recursion/new retry, explicit failures and
253
+ omitted-context conservatism. Extra supplier charges are separate from Chat
254
+ Usage. Require labeled-corpus calibration before enabling by default; synthetic
255
+ tests do not prove GitHub-private routing equivalence. See `docs/auto-task-routing.md`.
256
+
243
257
  The turn selection Host endpoint must use `TurnSelectionController`, extending
244
258
  public `TypertRemoteService` with `@Remote get`. A plain `ctx.provide` object
245
259
  cannot supply the native gateway binding or source method discovery. Test both
package/CHANGELOG.md CHANGED
@@ -1,5 +1,18 @@
1
1
  # Changelog
2
2
 
3
+ ## 0.4.0-alpha.82 (prepared)
4
+
5
+ - Explain missing native Turn Usage on completed Copilot replies through a plugin-owned read-only conversation projection and a compact, keyboard-accessible footer disclosure (#259).
6
+ - Distinguish the exact recorded local pre-dispatch input-budget block, other settlements without usage, and incomplete history. A recovered turn can contain reported successful steps without a provable whole-turn total.
7
+ - Preserve native Usage when complete; do not fabricate zero usage, aggregate partial totals, change history or replace compaction/retry ownership. Record the observed plugin/Core integration gap and its limits.
8
+
9
+ ## 0.4.0-alpha.81 (prepared)
10
+
11
+ - Route Auto through current supplier Powerful/Versatile/Lightweight categories and contextual task evidence instead of advertised-capacity bands (#258). Preserve hard input/image/exclusion/account admission and native no-fit recovery.
12
+ - Use Lightweight for isolated greetings even with Intelligence, retain capable targets for short hard work, prefer suitable previous-model continuity, and distribute only equal-category candidates.
13
+ - Explain the actual captured turn choice, category fallback and continuity/tie-break in the footer; old records explicitly lack detailed reasons rather than displaying generic reconstructed text.
14
+ - Add an off-by-default bounded semantic-assessment experiment through the existing native adapter. Strict output, omitted-context conservatism, cancellation and diagnostic fallback remain; extra calls add latency and supplier charges outside native Chat Usage. Synthetic tests do not establish calibrated accuracy or GitHub-private parity.
15
+
3
16
  ## 0.4.0-alpha.80 (prepared)
4
17
 
5
18
  - Observe real HTTP/SSE byte progress on managed requests, retaining the original five-minute byte-idle bound while allowing at most ten minutes without a native assistant chunk (#253).
package/README.md CHANGED
@@ -20,6 +20,7 @@ Core still owns image conversion and history. See [image compatibility limits](.
20
20
  - **Search routing:** choose the primary and fallback search providers under **Plugins → dsh-github-copilot → Details**. The page intentionally exposes only those user-facing choices; protocol probes, timeouts, allowlists, and transport switches are not presented as routine UI. It also keeps the rc.2 lifecycle warning visible before disable/remove/upgrade actions. If the updated routed service is not active yet, the card reports **Restart required** instead of presenting a broken form.
21
21
  - **Usage:** eligible Copilot sessions get an optional **Credits** control beside the composer, with provider-reported billing-cycle data and explicit unavailable/stale states. The quota request alone combines Node default and OS trust roots without disabling certificate verification or relying on Desktop launcher environment inheritance.
22
22
  - **Context evidence:** managed requests suppress terminal all-zero usage on failure or cancellation. A separate composer disclosure can identify historical failed-zero readings and show the last applicable input sample, including cache. This is not current occupancy, does not replace the native `0%` meter, and does not rewrite history. See [context evidence boundaries](./docs/copilot-usage.md#historical-context-evidence).
23
+ - **Missing Turn Usage:** a completed Copilot reply without a native total can show **Turn Usage unavailable**. Its read-only explanation distinguishes recorded local pre-dispatch budget blocks, other missing samples and incomplete history; it never invents zero usage or a partial turn total. [Observation and boundaries](./docs/copilot-usage.md#missing-completed-turn-usage).
23
24
  - **Compatibility:** the plugin applies Copilot-only tool-schema, reasoning-presentation, request-budget, and account-model compatibility through published DSH APIs. It does not patch Core or maintain a second model catalog.
24
25
 
25
26
  ![Current Web search provider routing on the plugin details page](./docs/images/copilot-search-routing.png)
@@ -108,7 +109,7 @@ The Host reads GitHub's internal quota endpoint using the existing canonical OAu
108
109
 
109
110
  ## Install and sign in
110
111
 
111
- The commands below target the package version `0.4.0-alpha.80`. Versioned URLs describe the intended release artifacts, not proof that publication or local activation has completed; use them only once that Release and its checksums are available. Install into the profile you use (replace `web` when targeting another profile):
112
+ The commands below target the package version `0.4.0-alpha.82`. Versioned URLs describe the intended release artifacts, not proof that publication or local activation has completed; use them only once that Release and its checksums are available. Install into the profile you use (replace `web` when targeting another profile):
112
113
 
113
114
  Before installing/updating, unpack the **checksum-verified** archive into a temporary directory and run its read-only composition preflight (replace all paths with absolute paths for the intended profile):
114
115
 
@@ -121,10 +122,10 @@ Include any launcher patch files with repeated `--patch /absolute/file` argument
121
122
  For a standalone named profile only, the supported CLI command is:
122
123
 
123
124
  ```sh
124
- dsh plugin --profile web add https://github.com/cloga/dsh-github-copilot/releases/download/v0.4.0-alpha.80/dsh-github-copilot-0.4.0-alpha.80.tgz
125
+ dsh plugin --profile web add https://github.com/cloga/dsh-github-copilot/releases/download/v0.4.0-alpha.82/dsh-github-copilot-0.4.0-alpha.82.tgz
125
126
  ```
126
127
 
127
- For Desktop profiles, use Desktop's native package manager with `dsh-github-copilot@0.4.0-alpha.80` after the same preflight and only when registry access is permitted. Desktop reserves its `desktop` profile from CLI management. If registry access is blocked or unavailable, stop; the documented [controlled offline CLI procedure](./docs/npm-distribution.md#controlled-offline-cli-maintenance-for-standalone-profiles) is not a Desktop workaround.
128
+ For Desktop profiles, use Desktop's native package manager with `dsh-github-copilot@0.4.0-alpha.82` after the same preflight and only when registry access is permitted. Desktop reserves its `desktop` profile from CLI management. If registry access is blocked or unavailable, stop; the documented [controlled offline CLI procedure](./docs/npm-distribution.md#controlled-offline-cli-maintenance-for-standalone-profiles) is not a Desktop workaround.
128
129
 
129
130
  Then open the Models UI listed above, find **GitHub Copilot**, select **Sign in**, and complete the GitHub device-code flow. Plugin installation changes the selected profile; activation follows that profile's normal reload/restart policy.
130
131
 
@@ -142,10 +143,14 @@ If preferences cannot be read, available account rows remain visible but read-on
142
143
 
143
144
  **One setting for subagents:** in the plugin configuration page, turn on **Subagent models → Follow parent model** and Save. It applies to existing and new supported Copilot subagents and Team mates, without Session IDs or per-child setup. A fixed parent model takes effect on each child's next turn; Auto inherits the same preference but chooses using each child's own context. Running turns stay unchanged and child-owned manual selections still win. Default Off (`github-copilot.followParentModel: false`). Enabling replaces creation-time model choices; it never clears explicit child selection history. Legacy `parentModelFollow` bindings remain active independently and are not migrated. See [requirements, mockup and boundaries](./docs/parent-model-follow.md). Native Team creation labels are not rewritten.
144
145
 
145
- The managed GitHub Copilot group includes **Auto · Balance**, **Auto · Efficiency**, and **Auto · Intelligence** as flat virtual picker rows. They share one account-verified eligible pool after applying exact model exclusions. Before preference, routing evaluates input headroom against each candidate's hard input budget, filtering out candidates that cannot accommodate current messages. Soft preferences distribute across upper/lower/center candidate bands using deterministic in-memory seeds to avoid monopolizing a single top candidate across Sessions. If no candidate fits, Auto selects the largest input capacity model and triggers compaction pressure on compressible history; if exclusions leave no eligible model, Auto fails closed. Auto resolves once per Core turn to a real model; every tool step, retry and automatic compaction recovery in that turn stays on that model, while a later Chat or Goal turn may choose again. Image turns require a non-excluded account-verified image-capable candidate. New turns retain Core's actual model/usage details without adding unsafe plugin events. Search keeps its independent provider-owned model policy. Today, native subagents and Team teammates initialize from the parent's latest recorded request route (or creation options before its first request); this is not a live link, and a route snapshot does not prove Auto intent. The requested per-child-turn inheritance behavior and Core/plugin boundary are documented in the [policy](./docs/automatic-model-routing.md#requested-parent-to-child-selection-inheritance). The nested management disclosure is plugin-owned; Core's picker remains flat.
146
+ The managed GitHub Copilot group includes **Auto · Balance**, **Auto · Efficiency**, and **Auto · Intelligence** as flat virtual picker rows. They share one account-verified eligible pool after applying exact model exclusions and input-fit filtering. Routing now uses the account's **Powerful / Versatile / Lightweight** classification and conservative contextual task assessment, not reasoning effort or context capacity as a quality ranking. Even Intelligence uses Lightweight for an isolated greeting; short difficult reasoning stays Powerful. Unknown tasks prefer Versatile with Balance/Efficiency and Powerful with Intelligence. Suitable previous-model continuity wins before stable equal-weight category tie-breaking. Missing categories and fallback remain explicit. If no candidate fits, Auto preserves largest-input-budget native recovery; if none is eligible, it fails closed. One decision stays fixed across the turn's tool steps, retries and compaction. Images retain their independent verified admission. Search, actual model/Usage and histories remain native-owned. See the [task policy and limits](./docs/auto-task-routing.md) and [parent-inheritance boundary](./docs/automatic-model-routing.md#requested-parent-to-child-selection-inheritance). Core's picker remains flat.
147
+
148
+ The optional `github-copilot.autoSemanticAssessment: true` experiment recognizes contextual tasks with one bounded auxiliary model call; it is **off by default** until calibrated. It adds latency and supplier charges not included in native chat Usage. Failure or omitted assessment context never means the task is simple; it retains conservative local evidence. This is a local routing approximation, not GitHub's private Auto service.
146
149
 
147
150
  Completed Copilot replies show **Auto (preference)** and an expandable recorded reason after native Usage/time, or **Manual** only with captured explicit fixed-selection evidence. A successful read with no retained evidence shows **Selection unknown**, never a guess from today's picker. Failed reads instead show **Selection unavailable → Retry**; retry only reads the same turn's record, not another model request. Explicit Session/turn arguments remain stable inside a session-bound Chat, avoiding the former native gateway argument-count failure on fresh master/lead replies. There is no repeated model name or separate Model details button: actual model display stays in native Usage, whose missing-route behavior is unchanged. New selection records are bounded and Host-local; restart, disposal or eviction loses them. Compatible historical Auto records remain readable. The public assistant-actions slot supports internal wrapping only, not changing the native fixed-height row.
148
151
 
152
+ The information button explains this turn's captured task/category choice, fallback and continuity/tie-breaking. Supporting assessment facts sit in a disclosure; older records without details explicitly say that the reason was not retained.
153
+
149
154
  The selection footer mounts independently of optional conversation projections. Native completed-turn evidence lets retained Auto/Manual records remain visible if those projections are unavailable; incomplete attribution stays explicit in the information disclosure. This does not restore lost selection records or prove that an installed update is loaded in Desktop.
150
155
 
151
156
  Alpha.79 also fixes the Host half of selection reads: the retained decision store is now exposed through the public Remote service binding and method marker, not a plain provided object rejected by the native gateway. Actual Client-to-Host roundtrips cover both source discovery and strict descriptors. This complements alpha.76's Client argument fix; it does not recover lost records or change Auto model selection.
@@ -179,7 +184,7 @@ Agents should treat the browser authorization as a human handoff, not as a token
179
184
  5. Confirm **Signed in** and inspect the automatic discovery result before asking the user to choose a model. Already-signed-in Models opening ensures missing/stale metadata automatically; fresh ready cache makes no request. Use visible **Retry** for errors or **Manage → Refresh models** for an intentional forced update, not routine setup. Status alone does not discover, and login, metadata and successful model calls remain separate evidence.
180
185
  6. Use **Sign out** only when the user explicitly asks to disconnect the account. It deletes the Copilot credential record but preserves route settings.
181
186
 
182
- GitHub Releases and npm are the default distribution channels for each new version, using the same verified tarball. Pin the version and verify the evidence for the channel used: Release `SHA256SUMS`, and npm `dist.integrity` when installing from npm. Install into Desktop profiles only through Desktop's native package manager; the reserved `desktop` profile is not managed by the DSH CLI. After npm publication is verified, the native manager accepts `dsh-github-copilot@0.4.0-alpha.80`, not a URL or file. Controlled offline CLI maintenance is documented for standalone named profiles only, never Desktop-managed profiles. Follow the mandatory search-composition preflight, backup, single-writer and post-install checks; do not bypass a corporate registry ban, disable TLS verification or restart without separate approval. See [distribution and publication requirements](./docs/npm-distribution.md).
187
+ GitHub Releases and npm are the default distribution channels for each new version, using the same verified tarball. Pin the version and verify the evidence for the channel used: Release `SHA256SUMS`, and npm `dist.integrity` when installing from npm. Install into Desktop profiles only through Desktop's native package manager; the reserved `desktop` profile is not managed by the DSH CLI. After npm publication is verified, the native manager accepts `dsh-github-copilot@0.4.0-alpha.82`, not a URL or file. Controlled offline CLI maintenance is documented for standalone named profiles only, never Desktop-managed profiles. Follow the mandatory search-composition preflight, backup, single-writer and post-install checks; do not bypass a corporate registry ban, disable TLS verification or restart without separate approval. See [distribution and publication requirements](./docs/npm-distribution.md).
183
188
 
184
189
  **Desktop package-update and Web-lifecycle boundaries:** the native manager's `restart-required` result for an in-place installed-package upgrade is authoritative and distinct from plugin-bundle removal/re-add behavior. Any disable, removal, or upgrade involving service recomposition (`cordis.patch.yml`, such as the isolated routed web search layer) may leave dependent entries pending until Desktop is fully exited and cold-restarted. If the native manager reports only entries such as `pending (waiting for service: web)` while this plugin already shows **Off**, the requested disabled state has been persisted; cold-restart Desktop instead of repeatedly toggling or reinstalling the plugin. Other errors are not covered by this guidance and must still be investigated.
185
190
 
@@ -461,8 +466,8 @@ Report the published Release URL, version, tag/commit and verified asset SHA-256
461
466
  `package.json` declares public npm distribution. A release tag must equal `v${package.json.version}`. Versions use standard SemVer prerelease labels (`alpha`, `beta`, or `rc`), each with its matching npm dist-tag; only stable versions use `latest`. The Release workflow performs the frozen install and complete verification gate, packs once (or recovers the original archive on retry), verifies `SHA256SUMS`, publishes the immutable GitHub Release and then publishes those same bytes to npm through OIDC. Either channel failing means delivery is incomplete. First package creation needs an authorized maintainer; staging requires an existing package and is not a first-package bootstrap. Historical releases are not republished.
462
467
 
463
468
  ```sh
464
- curl -LO https://github.com/cloga/dsh-github-copilot/releases/download/v0.4.0-alpha.80/dsh-github-copilot-0.4.0-alpha.80.tgz
465
- curl -LO https://github.com/cloga/dsh-github-copilot/releases/download/v0.4.0-alpha.80/SHA256SUMS
469
+ curl -LO https://github.com/cloga/dsh-github-copilot/releases/download/v0.4.0-alpha.82/dsh-github-copilot-0.4.0-alpha.82.tgz
470
+ curl -LO https://github.com/cloga/dsh-github-copilot/releases/download/v0.4.0-alpha.82/SHA256SUMS
466
471
  sha256sum --check SHA256SUMS
467
472
  ```
468
473
 
@@ -470,7 +475,7 @@ PowerShell can verify the same two downloaded files with:
470
475
 
471
476
  ```powershell
472
477
  $expected = (Get-Content .\SHA256SUMS).Split()[0]
473
- $actual = (Get-FileHash .\dsh-github-copilot-0.4.0-alpha.80.tgz -Algorithm SHA256).Hash.ToLowerInvariant()
478
+ $actual = (Get-FileHash .\dsh-github-copilot-0.4.0-alpha.82.tgz -Algorithm SHA256).Hash.ToLowerInvariant()
474
479
  if ($actual -cne $expected) { throw 'Release checksum mismatch' }
475
480
  ```
476
481
 
package/README.zh.md CHANGED
@@ -20,13 +20,14 @@ Core 管理。详见[图片兼容性与限制](./docs/image-input-compatibility.
20
20
  - **搜索路由:**在 **插件 → dsh-github-copilot → 详情**选择主搜索 Provider 与兜底 Provider。页面只保留用户真正需要的两项选择,不再把协议探测、超时、白名单或传输开关作为日常配置暴露;同时会在停用/移除/升级前常驻显示 rc.2 生命周期提示。更新后的路由服务尚未激活时,卡片会明确显示 **需要重启生效(Restart required)**,而不是继续展示无法使用的表单。
21
21
  - **用量:**符合条件的 Copilot 会话会在输入框旁显示可选的 **Credits / 额度** 控件,展示供应方明确返回的计费周期数据,并如实区分不可用与历史快照。仅额度请求组合 Node 默认证书与系统信任根,无需依赖 Desktop launcher 的环境变量继承,且保持 TLS 验签。
22
22
  - **上下文证据:**托管请求失败或取消时,不再传递终止前的全零用量。独立的输入框提示可识别历史失败零值,并显示仍适用的最后输入采样(含缓存);它不是当前占用率,不替换原生 `0%` 表盘,也不改写历史。详见[上下文证据边界](./docs/copilot-usage.md#historical-context-evidence)。
23
+ - **整轮 Usage 缺失:**已完成的 Copilot 回复没有原生总量时,可显示**本轮 Usage 不可用**。只读说明区分已记录的插件发出前预算拦截、其他缺失采样和不完整历史;不会补造零用量或把部分步骤合计当成整轮总量。详见[观察与边界](./docs/copilot-usage.md#missing-completed-turn-usage)。
23
24
  - **兼容处理:**插件通过 DSH 已发布公开接口提供 Copilot 专用 Tool Schema、Thinking 展示、请求预算及账号模型兼容,不修改 Core,也不维护第二套模型目录。
24
25
 
25
26
  ![当前插件详情页的 Web 搜索 Provider 路由](./docs/images/copilot-search-routing.png)
26
27
 
27
28
  截图来自当前构建 Client 的隔离浏览器 fixture,使用合成 Provider。它只证明已发布配置界面的形态,不证明真实凭据、账号权限、搜索成功或当前 Desktop 已加载。
28
29
 
29
- > 当前文档与安装命令对应版本 `0.4.0-alpha.80`。源码、发布制品、已安装版本和实际加载运行时需分别确认;本地升级和中断会话的重启仍需用户批准。
30
+ > 当前文档与安装命令对应版本 `0.4.0-alpha.82`。源码、发布制品、已安装版本和实际加载运行时需分别确认;本地升级和中断会话的重启仍需用户批准。
30
31
 
31
32
  Web search 路由现在优先显示在插件详情页;宿主不支持详情页插槽时才回退到 **设置 → 模型**或独立 Web search section。搜索子 Fiber 仍独立等待 Remote,不影响账号控件和既有搜索安全检查。
32
33
 
@@ -123,10 +124,10 @@ node package/scripts/check-search-composition.mjs --profile-dir /absolute/profil
123
124
  仅对独立的具名 profile,获准且网络可用时,可通过受支持的 CLI 命令安装:
124
125
 
125
126
  ```sh
126
- dsh plugin --profile web add https://github.com/cloga/dsh-github-copilot/releases/download/v0.4.0-alpha.80/dsh-github-copilot-0.4.0-alpha.80.tgz
127
+ dsh plugin --profile web add https://github.com/cloga/dsh-github-copilot/releases/download/v0.4.0-alpha.82/dsh-github-copilot-0.4.0-alpha.82.tgz
127
128
  ```
128
129
 
129
- Desktop profile 请在完成同样的预检后,使用 Desktop 原生包管理器安装 `dsh-github-copilot@0.4.0-alpha.80`。Desktop 保留 `desktop` profile,CLI 不负责管理。若 registry 被公司封禁或不可用,不要更换网络绕行或使用离线 CLI;请停止。[受控离线 CLI 流程](./docs/npm-distribution.md#controlled-offline-cli-maintenance-for-standalone-profiles)仅适用于独立具名 profile,不适用于 Desktop。
130
+ Desktop profile 请在完成同样的预检后,使用 Desktop 原生包管理器安装 `dsh-github-copilot@0.4.0-alpha.82`。Desktop 保留 `desktop` profile,CLI 不负责管理。若 registry 被公司封禁或不可用,不要更换网络绕行或使用离线 CLI;请停止。[受控离线 CLI 流程](./docs/npm-distribution.md#controlled-offline-cli-maintenance-for-standalone-profiles)仅适用于独立具名 profile,不适用于 Desktop。
130
131
 
131
132
  随后打开上表对应的 Models UI,找到 **GitHub Copilot**,点击 **Sign in** 并完成 GitHub device-code 流程。安装会修改指定 profile;是否立即激活取决于该 profile 的常规 reload/restart 策略。
132
133
 
@@ -144,10 +145,14 @@ Desktop profile 请在完成同样的预检后,使用 Desktop 原生包管理
144
145
 
145
146
  **子代理只需设置一次:** 在插件配置页打开 **Subagent models → Follow parent model** 并保存,即应用于已有和新创建的受支持 Copilot 子代理及 Team mate,无需填写 Session ID 或逐个登记。父会话为固定模型时,子会话从下一轮跟随;父会话为 Auto 时,继承相同偏好并按各自上下文独立选模。正在执行的轮次不变,子会话自己的手动选择仍优先。默认关闭(`github-copilot.followParentModel: false`)。开启意味着接管创建时保存的模型选择,但不清除子会话的显式选择历史。旧 `parentModelFollow` 绑定独立保留,不自动迁移。参见[需求、mockup 与边界](./docs/parent-model-follow.md)。原生 Team 的创建时模型标签不会被改写。
146
147
 
147
- 托管 GitHub Copilot 分组提供三个平铺的虚拟选择:**Auto · Balance**(均衡)、**Auto · Efficiency**(效率)和 **Auto · Intelligence**(智能)。三档先应用精确模型排除,再共用当前账号验证过的合格模型池。在应用偏好前,路由会根据候选模型的硬输入预算计算输入余量,剔除无法容纳当前消息的候选。软偏好通过确定性内存种子在候选区间(上段/下段/中段)内分布,避免单一顶配模型跨 Session 被垄断。若所有候选均超限,Auto 会选择输入容量最大的模型并对可压缩历史触发 compaction pressure;若排除后没有任何合格模型,则 Auto 关闭式失败。Auto 在每个 Core turn 只解析一次;同轮工具步骤、重试和自动 compact 恢复使用同一真实模型,新轮次可重新选择。图片轮次必须有未被排除且已验证的图片候选。新轮次保留 Core 的真实模型/用量信息,不追加不兼容的插件事件。搜索仍由独立策略管理。目前原生子代理与 Team 队友会以 master 最近记录的请求 route 初始化(master 尚未发出请求时回退到创建配置);这不是随 master 后续选择自动变化的动态关联,route 快照也不能证明 Auto 意图已继承。每轮跟随 master 固定模型、或让子代理按自身上下文独立执行 Auto 的目标行为及 Core/插件边界,见[设计记录](./docs/automatic-model-routing.md#requested-parent-to-child-selection-inheritance)。嵌套管理折叠区由插件提供;Core picker 仍保持平铺。
148
+ 托管 GitHub Copilot 分组提供三个平铺的虚拟选择:**Auto · Balance**(均衡)、**Auto · Efficiency**(效率)和 **Auto · Intelligence**(智能)。三档共用精确排除与输入适配后的账号合格模型池。现在依据账号返回的 **Powerful / Versatile / Lightweight** 分类和保守的上下文任务判断选择,不再把推理档位或上下文容量当作品质排名。即使 Intelligence,独立问候也用 Lightweight;简短但困难的推理仍用 Powerful。无法判断的任务,Balance/Efficiency 优先 Versatile,Intelligence 优先 Powerful。上一轮模型仍符合目标分类与输入要求时优先保持它,否则仅在同类候选间稳定等权选择。分类缺失、分类回退均明确说明。没有可容纳候选时保留最大输入预算的原生恢复;没有合格候选则明确失败。同轮工具步骤、重试和自动压缩保持同一决策。图片验证独立保留;搜索、真实模型/Usage 和历史仍由原有服务所有。参见[任务策略与边界](./docs/auto-task-routing.md)及[父子继承边界](./docs/automatic-model-routing.md#requested-parent-to-child-selection-inheritance)。Core picker 仍保持平铺。
149
+
150
+ 可选实验 `github-copilot.autoSemanticAssessment: true` 通过一次有界的辅助模型请求判断上下文任务;校准前**默认关闭**。它会增加延迟和供应方费用,不计入原生 Chat Usage。失败或判断上下文不完整不会被当作简单任务,而是保留保守本地依据。这是插件本地近似策略,不是 GitHub 的私有 Auto 服务。
148
151
 
149
152
  已完成的 Copilot 回复在原生 Usage/时间后显示 **Auto(偏好)**,点信息按钮查看本轮记录的选择原因;只有捕获到明确固定选择才显示**手动**。成功读取但没有保留记录时显示**选择方式未知**,不借当前 picker 推断历史。读取失败则显示**选择记录读取失败 → 重试**;重试只重新读取同一轮记录,不重新调用模型。显式会话 ID/轮次参数不再被 Chat 的自动会话绑定缩减,修复新 master/lead 回复遭遇原生网关参数数量错误的问题。不重复模型名、不另设 Model details 按钮:实际模型仍由原生 Usage 显示,其缺失归属行为未改变。新选择记录仅限当前 Host 内存且有容量上限,重启、释放 Agent 或淘汰后会丢失;兼容历史 Auto 记录仍可读取。公开 assistant-actions slot 只允许插件内部换行,不能改变原生固定高度尾栏。
150
153
 
154
+ 信息按钮展示本轮实际捕获的任务/分类依据、回退和保持原模型/同类分配原因;辅助判断信息收纳在折叠区。旧记录没有详细依据时明确说明未保留原因,不用今天的设置反推。
155
+
151
156
  选择标记独立于可选会话投影挂载。投影不可用时,原生已完成回合证据仍允许显示保留的 Auto/手动记录;归属证据不完整会在信息弹层中明确说明。这不会恢复丢失的选择记录,也不证明已安装的更新已在 Desktop 加载。
152
157
 
153
158
  Alpha.79 进一步修复选模读取的 Host 端:通过官方 Remote 服务绑定与方法标记暴露现有决策存储,不再使用会被原生网关拒绝的普通对象。实际 Client→Host 连通回归覆盖源码发现和严格描述符两种方式。这补齐 alpha.76 的 Client 参数修复,不恢复已丢失记录,也不改变 Auto 的选模逻辑。
@@ -181,7 +186,7 @@ Agent 应把浏览器授权视为需要用户完成的 handoff,而不是自行
181
186
  5. 确认 **Signed in** 并检查自动发现结果,再请用户选择模型。已登录时打开 Models 会自动确保缺失/过期元数据,新鲜 ready 缓存不发请求。错误可使用 **Retry**,有意强制更新时使用 **Manage → Refresh models**,不作为常规设置步骤。状态读取本身不发现;登录、元数据与真实调用成功是独立证据。
182
187
  6. 只有用户明确要求断开账号时才使用 **Sign out**;它会删除 Copilot credential record,但保留 route settings。
183
188
 
184
- 每个新版本默认同时分发到 GitHub Releases 和 npm,两个渠道使用同一份已验证 tarball。应固定版本并核对所用渠道的证据:Release 的 `SHA256SUMS`;从 npm 安装时另核对 `dist.integrity`。Desktop profile 只能通过原生 Desktop 包管理器安装;获准且 registry 可用时输入 `dsh-github-copilot@0.4.0-alpha.80`,不是 URL 或本地文件。Desktop 保留 `desktop` profile,不能通过 DSH CLI 管理。受控离线 CLI 维护仅适用于独立具名 profile,不得作为 Desktop 安装绕行方案。[双渠道发布与 OIDC 要求](./docs/npm-distribution.md)保持不变。
189
+ 每个新版本默认同时分发到 GitHub Releases 和 npm,两个渠道使用同一份已验证 tarball。应固定版本并核对所用渠道的证据:Release 的 `SHA256SUMS`;从 npm 安装时另核对 `dist.integrity`。Desktop profile 只能通过原生 Desktop 包管理器安装;获准且 registry 可用时输入 `dsh-github-copilot@0.4.0-alpha.82`,不是 URL 或本地文件。Desktop 保留 `desktop` profile,不能通过 DSH CLI 管理。受控离线 CLI 维护仅适用于独立具名 profile,不得作为 Desktop 安装绕行方案。[双渠道发布与 OIDC 要求](./docs/npm-distribution.md)保持不变。
185
190
 
186
191
  **Desktop package 更新与 Web 生命周期边界:**原生 manager 对已安装 package 原位升级返回的 `restart-required` 是权威结果,与移除/重加 plugin bundle 的行为是两个不同问题。涉及服务重编排补丁(`cordis.patch.yml`,如本插件所采用的隔离路由 Web 搜索层)的停用、移除或升级,可能让依赖条目保持 pending,直到完全退出并冷重启 Desktop。若原生 manager 只报告类似 `pending (waiting for service: web)` 的条目,同时本插件已经显示 **Off**,说明请求的停用状态已经落盘;请冷重启 Desktop,不要反复切换或重装插件。其它错误不属于此提示范围,仍需单独排查。
187
192
 
@@ -472,8 +477,8 @@ node scripts/agent.mjs attribution "DeepSeek Harness (DSH)"
472
477
  `package.json` 声明公开 npm 分发。Release tag 必须严格等于 `v${package.json.version}`。预发布使用 `alpha`、`beta` 或 `rc` 及对应 npm dist-tag,只有稳定版使用 `latest`。Release workflow 执行 frozen install 和完整门禁,只打包一次(重试恢复原始归档),验证 `SHA256SUMS`,发布不可变 GitHub Release,再通过 OIDC 将同一份字节发布到 npm。任一渠道失败都表示交付未完成。首次建包须由获准环境中的维护者完成;staging 要求包已存在,不能代替首次建包。不会批量补发历史版本。
473
478
 
474
479
  ```sh
475
- curl -LO https://github.com/cloga/dsh-github-copilot/releases/download/v0.4.0-alpha.80/dsh-github-copilot-0.4.0-alpha.80.tgz
476
- curl -LO https://github.com/cloga/dsh-github-copilot/releases/download/v0.4.0-alpha.80/SHA256SUMS
480
+ curl -LO https://github.com/cloga/dsh-github-copilot/releases/download/v0.4.0-alpha.82/dsh-github-copilot-0.4.0-alpha.82.tgz
481
+ curl -LO https://github.com/cloga/dsh-github-copilot/releases/download/v0.4.0-alpha.82/SHA256SUMS
477
482
  sha256sum --check SHA256SUMS
478
483
  ```
479
484
 
@@ -481,7 +486,7 @@ PowerShell 可以对已下载的同一组文件执行:
481
486
 
482
487
  ```powershell
483
488
  $expected = (Get-Content .\SHA256SUMS).Split()[0]
484
- $actual = (Get-FileHash .\dsh-github-copilot-0.4.0-alpha.80.tgz -Algorithm SHA256).Hash.ToLowerInvariant()
489
+ $actual = (Get-FileHash .\dsh-github-copilot-0.4.0-alpha.82.tgz -Algorithm SHA256).Hash.ToLowerInvariant()
485
490
  if ($actual -cne $expected) { throw 'Release checksum mismatch' }
486
491
  ```
487
492
 
@@ -8,7 +8,7 @@
8
8
  },
9
9
  "package": {
10
10
  "name": "dsh-github-copilot",
11
- "version": "0.4.0-alpha.80"
11
+ "version": "0.4.0-alpha.82"
12
12
  },
13
13
  "evidence": {
14
14
  "kind": "source-and-synthetic-test-inventory",
@@ -288,6 +288,25 @@
288
288
  ]
289
289
  },
290
290
  "capabilities": [
291
+ {
292
+ "id": "missing-turn-usage-explanation",
293
+ "required": true,
294
+ "sourceMarkers": [
295
+ {"file": "src/turn-usage-evidence.ts", "marker": "github-copilot-turn-usage-evidence"},
296
+ {"file": "src/turn-usage-evidence.ts", "marker": "tail.tokenUsage !== undefined"},
297
+ {"file": "src/turn-usage-notice.ts", "marker": "Turn Usage unavailable"},
298
+ {"file": "src/auto-model-presentation.ts", "marker": "turnUsageEvidenceDefinition"},
299
+ {"file": "scripts/verify-tagged-core.mjs", "marker": "tests/fixtures/turn-usage-core.fixture.ts"}
300
+ ],
301
+ "tests": [
302
+ {"file": "tests/turn-usage-evidence.spec.ts", "name": "explains a recovered local block alongside reported successful steps without summing tokens"},
303
+ {"file": "tests/turn-usage-evidence.spec.ts", "name": "does not label generic overflow, provider errors or ambiguous streams as pre-dispatch"},
304
+ {"file": "tests/turn-usage-notice.spec.ts", "name": "keeps the fixed-height footer compact and opens a bounded nonmodal explanation"},
305
+ {"file": "tests/turn-usage-integration.spec.ts", "name": "explains native missing usage even when the independent selection Remote fails"},
306
+ {"file": "tests/fixtures/turn-usage-core.fixture.ts", "name": "reproduces the finish-only local block and explains it without changing native accounting"},
307
+ {"file": "tests/fixtures/reasoning-presentation-core.fixture.ts", "name": "explains missing turn Usage through native assembly and scoped actions (%s)"}
308
+ ]
309
+ },
291
310
  {
292
311
  "id": "historical-context-evidence",
293
312
  "required": true,
@@ -1061,6 +1080,12 @@
1061
1080
  "id": "managed-copilot-auto-model-routing",
1062
1081
  "required": true,
1063
1082
  "sourceMarkers": [
1083
+ {"file": "src/account-model-catalog.ts", "marker": "model_picker_category"},
1084
+ {"file": "src/auto-task-assessment.ts", "marker": "COPILOT_AUTO_ASSESSMENT_CONTEXT_OMITTED"},
1085
+ {"file": "src/auto-task-classifier.ts", "marker": "classifyTaskWithAdapter"},
1086
+ {"file": "src/auto-model-routing.ts", "marker": "targetCategory"},
1087
+ {"file": "src/config.ts", "marker": "autoSemanticAssessment"},
1088
+ {"file": "src/auto-selection-explanation.ts", "marker": "No candidate fits"},
1064
1089
  {"file": "src/preview-route.ts", "marker": "GITHUB_COPILOT_AUTO_MODEL_ID"},
1065
1090
  {"file": "src/auto-model-host.ts", "marker": "keep attribution out of the log"},
1066
1091
  {"file": "src/parent-model-follow.ts", "marker": "resolveParentModel"},
@@ -1081,7 +1106,14 @@
1081
1106
  "tests": [
1082
1107
  {"file": "tests/auto-model-host.spec.ts", "name": "follows fixed and Auto parents using %s without writing child selections"},
1083
1108
  {"file": "tests/scripts/parent-model-follow.test.mjs", "name": "explicit same-model and different-model child selections both win"},
1084
- {"file": "tests/auto-model-routing.spec.ts", "name": "classifies only the latest user turn and selects a deterministic capacity tier"},
1109
+ {"file": "tests/auto-model-routing.spec.ts", "name": "retains legacy length evidence but does not use it as task or supplier category evidence"},
1110
+ {"file": "tests/auto-model-routing.spec.ts", "name": "uses the task/category policy with one hard-eligible pool for all preferences"},
1111
+ {"file": "tests/auto-model-routing.spec.ts", "name": "preserves suitable continuity but never lets it override task category or input fit"},
1112
+ {"file": "tests/auto-model-routing.spec.ts", "name": "discloses category fallbacks and never guesses missing category from ID or capacity"},
1113
+ {"file": "tests/auto-task-assessment.spec.ts", "name": "bounds context, never reads attachment/reasoning bodies, and does not downshift omitted context"},
1114
+ {"file": "tests/auto-task-assessment.spec.ts", "name": "preserves caller cancellation instead of using a success-shaped fallback"},
1115
+ {"file": "tests/preview-route.spec.ts", "name": "assesses Auto through one concrete public native adapter call and freezes the captured reason across retries"},
1116
+ {"file": "tests/turn-selection-card.spec.ts", "name": "explains a captured task/category decision with progressive details (%s)"},
1085
1117
  {"file": "tests/auto-model-routing.spec.ts", "name": "does not discard a complete Core assistant message when it contains reasoning"},
1086
1118
  {"file": "tests/auto-model-routing.spec.ts", "name": "retains the native current-surface and tool-envelope measurement as an input floor"},
1087
1119
  {"file": "tests/fixtures/compaction-pressure-core.fixture.ts", "name": "prices Core assistant text and reasoning with the unchanged native token meter"},
@@ -0,0 +1,114 @@
1
+ # Contextual Auto routing and captured selection reasons
2
+
3
+ Tracking: [#258](https://github.com/cloga/dsh-github-copilot/issues/258).
4
+
5
+ ## Requirements
6
+
7
+ Auto must explain why the initiating turn chose its actual route, not display a
8
+ generic description of prompt length. Intelligence prioritizes response quality
9
+ but can use a lightweight model for demonstrably simple work. Efficiency favors
10
+ economical capability without knowingly underpowering difficult work. Balance
11
+ trades those priorities. All preferences share the same account model pool.
12
+
13
+ The authenticated account `/models` response supplies `model_picker_category`.
14
+ The October 3, 2026 bounded live probe observed `powerful`, `versatile` and
15
+ `lightweight`; some records lacked the field. This is supplier classification,
16
+ not measured latency, quality, price or health. Official documentation's
17
+ `category` is a different source and must not override live account facts.
18
+ Do not derive category from IDs, names, reasoning effort or context capacity.
19
+
20
+ Hard constraints remain first: entitlement, exclusions, current proof, text/image
21
+ capability and actual image format, estimated input headroom and final native
22
+ admission. No preference overrides these. No new Core owner, adapter, credential
23
+ record, general transport, global model mutation or durable decision event.
24
+
25
+ ## Task assessment
26
+
27
+ Assessment concerns the task in context, not the last sentence's length.
28
+ An isolated greeting may be low demand. A short request to prove a theorem or
29
+ continue an unresolved investigation is not evidence of low demand. Long input
30
+ can be mechanical. Prior assistant/tool work prevents a greeting/continuation
31
+ shortcut from claiming an isolated simple task.
32
+
33
+ Use a conservative local assessment as the default. Unknown is an explicit
34
+ outcome, never silently translated into simple. Optional semantic assessment is
35
+ an experiment until evaluated against annotated tasks; it must not claim
36
+ equivalence to GitHub's private router or calibrated confidence.
37
+
38
+ The semantic experiment uses one fixed concrete account candidate through the
39
+ existing managed native adapter, never an Auto ID and never `ctx.llm` routing
40
+ recursion. It has no tools, bounded text-only assessment context, bounded JSON
41
+ output, one deadline, cancellation and no new retry loop. Text from messages is
42
+ untrusted data, not router instructions. Files/images/reasoning/replay bodies
43
+ are not read; omitted context makes classification uncertain and cannot justify
44
+ downshifting. Assessment input is a separate projection: the actual chat
45
+ messages, system prompt, tools, attachments and history stay untouched.
46
+
47
+ The strict result contains demand (`simple`, `routine`, `complex`, `unknown`)
48
+ and finite evidence signals, not a model ID, raw reasoning or an uncalibrated
49
+ confidence percentage. Invalid/truncated/failed results retain the local
50
+ assessment with an explicit diagnostic. Caller cancellation aborts the turn
51
+ instead of falling back. Credential, metadata and account revocation remain
52
+ fatal through existing lease/admission guards. Extra calls incur supplier
53
+ charges and latency, separate from native chat Usage and without inferred cost.
54
+
55
+ ## Category policy and continuity
56
+
57
+ | Task demand | Efficiency | Balance | Intelligence |
58
+ | --- | --- | --- | --- |
59
+ | Proven simple | Lightweight | Lightweight | Lightweight |
60
+ | Routine | Lightweight | Versatile | Powerful |
61
+ | Complex | Powerful | Powerful | Powerful |
62
+ | Unknown | Versatile | Versatile | Powerful |
63
+
64
+ These are declared routing policies, not benchmark rankings. Efficiency's
65
+ routine lightweight selection is an economical approximation, not proof of
66
+ equal output quality. Intelligence never treats short length alone as simple.
67
+
68
+ When the target has no fitting candidates, use the disclosed category fallback:
69
+ powerful -> versatile -> lightweight; versatile -> powerful -> lightweight;
70
+ lightweight -> versatile -> powerful. Unclassified candidates are a final
71
+ explicit fallback only, not fabricated categories. Unknown/new category values
72
+ remain unclassified with a bounded catalog diagnostic.
73
+
74
+ Within the first available category, retain the previous managed model only
75
+ when it remains eligible and suitable. Otherwise use stable equal-weight
76
+ Session/turn selection with deterministic ID ordering as a tie breaker. Do not
77
+ infer quality from that ordering or assign fabricated weights. Once admitted,
78
+ all steps/retries share the turn decision. Continuity reduces unnecessary
79
+ switches but is not a claim of supplier cache savings.
80
+
81
+ If no model fits, preserve the existing largest-admissible-input recovery
82
+ branch and its native compaction diagnostics. Explain that no candidate fits;
83
+ never label this as a successful preference match or silently trim history.
84
+
85
+ ## Explanation contract
86
+
87
+ Capture task source/demand/signals, assessment diagnostic, target/selected
88
+ category, fitting category count, category fallback and continuity/tie-break or
89
+ no-fit selection at decision time. Client reads the explicit viewed Session and
90
+ turn through the existing strict Remote. It does not rerun classification or
91
+ consult today's picker. Show a decisive first sentence, bounded supporting
92
+ facts and optional policy detail. Do not repeat the model name or change Usage.
93
+
94
+ Legacy records without detailed evidence show that the reason was not retained.
95
+ Host-lifetime evidence may be lost on restart; do not invent durable history or
96
+ promise cold-history reconstruction.
97
+
98
+ ## Acceptance and rollout
99
+
100
+ Regress category normalization (missing/new/invalid values), short difficult
101
+ tasks, isolated greetings, contextual continuation, long mechanical input,
102
+ category fallback, previous-model eligibility, deterministic tie breaking,
103
+ no-fit recovery, one-call semantic success/invalid/failure/timeout/cancellation,
104
+ omitted context, strict Remote and footer English/Chinese explanations.
105
+
106
+ Retain native gateway, parent following, manual selection, exclusion admission,
107
+ account proof, image admission and compaction fixture coverage. Full protected
108
+ Windows/Linux CI remains mandatory before merge/release. No live inference
109
+ calls are needed for synthetic acceptance.
110
+
111
+ Before making semantic assessment the default, evaluate a reviewed labeled
112
+ corpus for complex-to-simple errors, extra latency, token use and stability
113
+ across languages and prompt injection. Synthetic passing tests are not this
114
+ evaluation. Publication does not authorize profile installation or restart.
@@ -1,6 +1,6 @@
1
1
  # GitHub Copilot Auto model routing MVP
2
2
 
3
- **Tracking:** [#174](https://github.com/cloga/dsh-github-copilot/issues/174), [#192](https://github.com/cloga/dsh-github-copilot/issues/192), [#199](https://github.com/cloga/dsh-github-copilot/issues/199)
3
+ **Tracking:** [#174](https://github.com/cloga/dsh-github-copilot/issues/174), [#192](https://github.com/cloga/dsh-github-copilot/issues/192), [#199](https://github.com/cloga/dsh-github-copilot/issues/199), [#258](https://github.com/cloga/dsh-github-copilot/issues/258)
4
4
 
5
5
  **Target:** official DSH and Windows Desktop `0.2.0-rc.2`
6
6
 
@@ -31,7 +31,7 @@ MIME are not wire-format proof. See [image compatibility and limits](./image-inp
31
31
 
32
32
  The account `/models` catalog is authoritative for entitlement and hard capabilities: protocol, context and output limits, image input, tool support, reasoning efforts, and server policy. Unknown or incomplete capability metadata fails closed.
33
33
 
34
- The catalog does not provide trustworthy quality, latency, price, or global-health rankings. The routing policy therefore does **not** infer model quality from a marketing name, model-ID prefix, provider order, or token price. It does not claim to reproduce GitHub's private Auto algorithm.
34
+ The catalog supplies `model_picker_category` (`powerful`, `versatile`, `lightweight`), but not trustworthy measured quality, latency, price, or global-health rankings. Missing, invalid and new category values remain unclassified with diagnostics. The routing policy does **not** infer category from a marketing name, model-ID prefix, reasoning effort, context capacity or token price. It does not claim to reproduce GitHub's private Auto algorithm.
35
35
 
36
36
  ### Hard input fit check before preference
37
37
 
@@ -49,28 +49,35 @@ Before applying soft preferences, routing checks whether each candidate can acco
49
49
  - `compaction-eligible`: prior history is compressible and downstream dispatch will trigger compaction pressure.
50
50
  5. **Final framing limitation:** The fresh entered-message estimate and current native envelope floor are not the final candidate-specific serialized request. Later framing, changed tools and native attachment projection can still exceed a candidate's budget. The native provider guard (`inspectRequest`) remains authoritative and truthfully rejects such a request with `COPILOT_CONTEXT_BUDGET_EXCEEDED` without bypassing safety boundaries.
51
51
 
52
- ### Soft capacity preferences across candidate bands
52
+ ### Task-aware supplier category preferences
53
53
 
54
- All three virtual preferences use the *same* eligible account models. Account entitlement, verified input capability (including image input), and input headroom are hard filters; preference changes where the routing policy lands in the deterministic, advertised-capacity order across candidate bands:
54
+ All three preferences use the *same* hard-eligible account pool. The default local
55
+ assessment recognizes isolated greetings and explicit difficult reasoning or
56
+ investigation; other tasks remain uncertain rather than guessing from length.
57
+ An optional `github-copilot.autoSemanticAssessment: true` experiment makes one
58
+ bounded auxiliary native-adapter call to a concrete, non-excluded supplier
59
+ Lightweight model. It adds latency and supplier charges not included in native
60
+ chat Usage. It is off by default pending labeled-corpus calibration.
55
61
 
56
- - **Intelligence:** Biased toward higher advertised capacity across the upper candidate band.
57
- - **Efficiency:** Biased toward lower advertised capacity across the lower candidate band, though demanding tasks allow higher capacity.
58
- - **Balance:** Central candidate band.
59
-
60
- To avoid monopolizing a single top candidate (such as Grok 4.7) across all turns and Sessions, band selection uses an in-memory deterministic seed derived from `${sessionId}:${turn}` (or turn content when Session context is absent). This guarantees:
61
-
62
- - Selection is **100% frozen** across all steps and retries within the same turn.
63
- - Unrelated Sessions or turns sample across the candidate band rather than concentrating on one model.
64
- - No sensitive prompt text or seed hashes are persisted or sent over the network.
65
- - When only a single candidate is eligible or fitting, it is returned unchanged.
66
-
67
- | Latest-turn class | Efficiency | Balance (`auto`) | Intelligence |
62
+ | Task demand | Efficiency | Balance (`auto`) | Intelligence |
68
63
  | --- | --- | --- | --- |
69
- | Fast | low band (index 0) | low band (index 0) | middle band |
70
- | Balanced | lower band | center band | upper band |
71
- | Strong | center band | upper band | upper band |
72
-
73
- These are **capacity preferences**, not measured quality, speed, price, or inference-cost preferences. They cannot promise that Intelligence is smarter or Efficiency faster or cheaper. The actual request-budget and compaction checks still apply after selection; no cost or speed metadata is fabricated.
64
+ | Proven simple | Lightweight | Lightweight | Lightweight |
65
+ | Routine | Lightweight | Versatile | Powerful |
66
+ | Complex | Powerful | Powerful | Powerful |
67
+ | Unknown | Versatile | Versatile | Powerful |
68
+
69
+ If the target category has no fitting model, follow the documented category
70
+ fallback order, using unclassified models only last. Keep the previous managed
71
+ model if it remains in the first available category and fits this turn.
72
+ Otherwise use stable equal-weight Session/turn allocation within that category;
73
+ ID ordering breaks ties, not quality rankings. Capacity and reasoning-effort
74
+ bands no longer determine task quality. All steps/retries retain the decision.
75
+
76
+ These are declared policies, not measured quality, speed or cost guarantees.
77
+ Captured footer reasons show task evidence, selected/target category, fallback,
78
+ continuity or equal tie-breaking, and no-fit recovery. Legacy records without
79
+ details say the reasons were not retained; they never reconstruct history from
80
+ today's settings. See [requirements, assessment limits and rollout](./auto-task-routing.md).
74
81
 
75
82
  The exact virtual preference stays in the Session's pending selection across turns, while each real request header and transport record the chosen account model. Switching to a real model ends automatic routing for subsequent turns. New turns no longer write a separate decision event; compatible historical events without a preference field remain Balance for display.
76
83
 
@@ -30,6 +30,24 @@ An additive composer disclosure identifies missing or unreliable context evidenc
30
30
 
31
31
  Local synthetic tests establish guarded-stream, historical-fold and presentation behavior only. Exact published-adapter and public-registry acceptance, complete CI, publication and loaded Desktop state must be verified separately. No live provider call or Session restart is needed for these tests.
32
32
 
33
+ ## Missing completed-turn Usage
34
+
35
+ ### Observed plugin/Core integration gap (#259)
36
+
37
+ One observed completed turn contained 57 successful assistant messages with reported usage and one failed `assistant/attempt` without a usage sample. That attempt recorded the plugin's exact local estimated-input-budget diagnostic: 796299 estimated tokens exceeded a 781113-token input budget, before provider dispatch. Native compaction recovered and the subsequent steps completed. No conversation content, credentials or opaque replay is needed to explain this observation; the exact loaded plugin bytes were not independently attested.
38
+
39
+ Official rc.2 `deriveTurnTokenUsage` requires complete lifecycle and exact usage for every attempt. One attempt without a sample makes its whole-turn result unavailable, so `TurnTailNodeView` omits native Usage. This is Core's existing fail-closed accounting rule triggered by our preventive pressure path, not evidence that all requests had no usage and not a regression in single-model exclusion saves.
40
+
41
+ ### Plugin-local explanation
42
+
43
+ A plugin-owned public conversation definition retains only bounded counts of settlements without reported samples and finish-only attempts with the exact recorded local diagnostic. On a completed Copilot reply whose native tail has no `tokenUsage`, the additive assistant-actions control opens **Turn Usage unavailable**. It uses the incumbent nonmodal popover pattern rather than expanding the native fixed-height footer.
44
+
45
+ The explanation distinguishes recorded local budget blocks, other settlements without samples, and missing turn-start/projection evidence. Generic provider overflow is never labeled pre-dispatch. Invalid reported samples and missing lifecycle boundaries can also make native accounting unavailable; if no cause is established, the explanation remains unknown rather than reimplementing the native validator. Projection and selection Remote failures remain independent, with named diagnostics; Retry only rereads selection evidence, never inference or accounting.
46
+
47
+ The control is historical evidence, not current context occupancy, account credits, dispatch attestation or a supplier receipt. It adds no token sum or fabricated zero usage, changes no native projection key, reads no opaque replay, and rewrites no history. Native complete Usage suppresses the explanation. Ordinary pressure protection, compaction transactions, errors, cancellation and retry remain unchanged. Removing the plugin removes the explanation without a data migration.
48
+
49
+ Synthetic unit/UI tests and exact unchanged Core accounting/registry fixtures prove these boundaries, not live provider receipts or loaded Desktop behavior. A native representation for unbilled pre-dispatch attempts may eventually remove the gap, but would require a separately requested upstream scope; it is not a prerequisite for this plugin-local explanation. Retire the companion diagnostic when native public presentation communicates the same uncertainty and the supported-version acceptance is verified.
50
+
33
51
  ## Account data boundary
34
52
 
35
53
  The Host uses the existing `llm-pi-ai/github-copilot` OAuth grant through public credential APIs. GitHub's `copilot_internal/user` endpoint uses the GitHub grant, not a second sign-in or a newly pasted token. Access-token renewal remains owned by pi-ai; quota inspection must not introduce a refresh implementation or credential persistence.