feat(web): support reasoning effort selection
This commit is contained in:
Yichen Jiang
@@ -2,5 +2,5 @@
|
||||
# side as of the last confirmed-consistent state. Both languages carry equal authority;
|
||||
# after editing either side, bring the other along and re-record with:
|
||||
# pnpm run verify-translation-pairing --write .agents/notes/implemented/feature/2026-07-24-web-session-model-selector.md
|
||||
2026-07-24-web-session-model-selector.md: b776a2e5ddf7339170ac64fb8daa3ec314dc45c2
|
||||
2026-07-24-web-session-model-selector.zh.md: 1068ad50ad34e52cc4ffbd52c2287e7c5018588e
|
||||
2026-07-24-web-session-model-selector.md: 4bd52bba8d3ba16c01fc59d2b561516f6fd0bb87
|
||||
2026-07-24-web-session-model-selector.zh.md: 35f3525b80e7d27fb573e854aefc317ed354df53
|
||||
|
||||
@@ -10,15 +10,15 @@ The Web conversation displayed and sent through the Host's fixed provider/model
|
||||
|
||||
## Decision
|
||||
|
||||
The Web Host reuses `installAgentLlmTarget` for every created or resumed agent. The target starts from the latest `request/header` when the session has used a model, otherwise from the Host default. `session.selectModel` changes the session-local mutable target, and prompt assembly captures it with request routing; a switch during a running step therefore applies to the next assembled step. The next consumed route persists through the existing full `request/header` snapshot, while a choice that has not reached a request remains process-local.
|
||||
The Web Host reuses `installAgentLlmTarget` for every created or resumed agent. The provider/model/reasoning target starts from the latest `request/header` when the session has used a model, otherwise from the Host default route. `session.selectModel` changes the session-local mutable target, and prompt assembly captures it with request routing; a switch during a running step therefore applies to the next assembled step. The next consumed target persists through the existing full `request/header` snapshot, while a choice that has not reached a request remains process-local.
|
||||
|
||||
The session RPC domain exposes `session.history`'s current `modelTarget`, a `session.models` directory, and `session.selectModel`. The directory is built dynamically from the LLM registry and grouped by provider. Provider catalogs load concurrently and fail independently, so successful groups remain usable alongside retryable failure records. Catalog membership stays advisory: the current model is inserted as an unlisted row when its registered provider omits it, and selecting an unlisted model under a registered provider remains valid.
|
||||
The session RPC domain exposes a `session.models` directory and `session.selectModel`. The directory is built dynamically from the LLM registry and grouped by provider; each listed model's exact metadata adds adapter-owned reasoning effort ids, names, descriptions, and optional default. Provider catalogs and exact metadata load concurrently by provider and fail independently, so successful groups remain usable alongside retryable failure records. Catalog membership stays advisory: the current model is inserted as an unlisted row when its registered provider omits it, while exact resolution decides whether a route and explicit effort are available. Selection uses `resolveCallConfig` to reject unsupported effort ids and materialize an adapter-configured default before updating the target.
|
||||
|
||||
The browser `Session` object owns the current target, grouped catalog, provider failures, operation error, and `idle`/`loading`/`ready`/`selecting`/`error` state. A Host session primes the directory when its selector mounts so the compact trigger can resolve a catalog name, and each menu open refreshes it. The resident shell has no session model route before Workspace selection connects or reuses a Host session, so its disabled no-session input dispatches no selector. Directory and selection calls share an operation generation so older responses cannot replace a newer result; a separate target-change generation lets concurrent history restore the logged model across a mount-time directory refresh without allowing old history to overwrite a user selection. Failures retain the previous current target and usable groups.
|
||||
The browser `ModelService` owns one `ModelDirectory` per live session. Its snapshot contains the current complete target, grouped catalog, provider failures, operation error, and `idle`/`loading`/`ready`/`selecting`/`error` state. Mounting primes the trigger label and each menu open refreshes the directory. Directory and selection calls share an operation generation so older responses cannot replace a newer result; connection reset discards the process-local projection before restoring the Host target. Failures retain the previous current target and usable groups.
|
||||
|
||||
`@deepseek-ai/dsh-client-ui-conversation` declares the session-scoped single slot `conversation.input.model` as a child of its composer-bar entry. InputBar renders the seat in its trailing controls immediately before the pending indicator and primary button; the seat receives the bar's `locked` owner prop and the session standard kit. `@deepseek-ai/dsh-client-ui-model-selector` occupies that dedicated seat, including for a Host-owned blank-session hero. Its compact trigger and radio rows display the catalog name, falling back to the model id for an unlisted current target, while the upward menu displays provider headings once with keyboard navigation, dismissal, retry states, and current selection marking.
|
||||
`@deepseek-ai/dsh-client-ui-conversation` declares the session-scoped single slot `conversation.input.model` as a child of its composer-bar entry. InputBar renders the seat in its trailing controls immediately before the pending indicator and primary button; the seat receives the bar's `locked` owner prop and session scope. `@deepseek-ai/dsh-client-ui-model` occupies that seat and also contributes `/model` over the same directory. Its compact trigger displays the catalog model name and effective reasoning label, falling back to ids when metadata is absent. The upward menu first offers Model and, when the current exact model supports it, Effort; Model drills into provider groups, while Effort drills into the adapter-ordered levels. The provider-default row appears only when the adapter does not configure a model default.
|
||||
|
||||
The production browser roster is the flat config tree in `apps/cli/cordis.yml`; the selector is one `dshClient` row rather than a package hardcoded in Web boot code. Its package manifest still declares the graph edge on `ui-conversation`, while cordis service availability governs activation.
|
||||
The production browser roster is the flat config tree in `apps/cli/cordis.yml`; the model feature is one `dshClient` row rather than a package hardcoded in Web boot code. Its package manifest orders it after the runtime and command feature, while Cordis service injection waits for the conversation slot before registering the composer occupant.
|
||||
|
||||
## Alternatives considered
|
||||
|
||||
@@ -26,6 +26,8 @@ The production browser roster is the flat config tree in `apps/cli/cordis.yml`;
|
||||
|
||||
**Hardcode the current DeepSeek catalog in the Web client.** This would drift from registered adapters and exclude deployment-owned providers. The LLM registry remains the source of provider and model metadata, including partial lookup failures.
|
||||
|
||||
**Keep `High`/`Max` as client-local UI state.** Static DeepSeek labels cannot represent `off`, pi-ai provider vocabularies, adapter defaults, validation, resume, or the next provider request. Exact-model metadata owns the selectable vocabulary, and the session target owns the selected id.
|
||||
|
||||
**Make the selection a global default.** A global mutation would unexpectedly redirect other open conversations. The target belongs to one live session, while Host configuration remains the default for sessions without a logged request.
|
||||
|
||||
**Reject changes while an agent is running.** The shared atomic target already separates the assembled step from the next selection. Keeping the selector available lets the user prepare the following step without altering the in-flight request.
|
||||
@@ -34,8 +36,8 @@ The production browser roster is the flat config tree in `apps/cli/cordis.yml`;
|
||||
|
||||
## Consequences
|
||||
|
||||
Any Host-backed Web conversation, including a blank session, can switch among dynamically discovered provider groups without displaying duplicated `provider/model` labels, and the current used route survives resume and reconnect. Catalog names remain presentation-only; selection and persistence continue to use provider/model ids. A provider catalog outage degrades only that group. Selection changes can reduce provider-side cache reuse when the route changes, but the selector adds no prompt content and does not disturb the in-flight step. The resident shell uses the Host default and exposes no selector only while it has no current session.
|
||||
Any Host-backed Web conversation, including a blank session, can switch among dynamically discovered provider groups and adapter-owned reasoning levels without displaying duplicated `provider/model` labels. The current consumed target survives resume and reconnect; catalog names remain presentation-only, while selection and persistence use provider/model/effort ids. A provider catalog or exact-metadata outage degrades only that group. Route changes can reduce provider-side cache reuse, but the selector adds no prompt content and does not disturb the in-flight step. A model without reasoning metadata has no Effort row.
|
||||
|
||||
## Testing
|
||||
|
||||
Host tests pin grouped discovery, duplicate-catalog isolation, partial provider failure, logged restoration, unlisted current targets, unavailable-provider rejection, and next-assembly switching. Client tests pin state transitions, failure preservation, transport errors, stale-response fencing, mount/open overlap, history restoration, and snapshot reference stability. UI tests pin the dedicated model-seat lifecycle and lock propagation, catalog-name labels with id fallback, provider grouping, radio semantics, retry/error states, successful and failed selection, outside dismissal, and Arrow/Home/End/Escape navigation. The keyless built-app fixture loads the selector through the production-shaped boot graph, selects OpenAI's GPT-5, sends a turn, and verifies that the next generated response reports the selected route.
|
||||
Host tests pin grouped discovery, catalog and exact-metadata failure isolation, logged effort restoration, unlisted current targets, unsupported effort rejection, default materialization, and next-assembly switching. Client tests pin the shared directory, reconnect restoration, and complete-target submission. Component tests pin dynamic effort labels, descriptions, provider-default exposure, and effort submission. The keyless built-app fixture loads the production model plugin, selects OpenAI's GPT-5 and its Max effort, sends a turn, and verifies that the next generated response reports both ids.
|
||||
|
||||
@@ -10,15 +10,15 @@ Web 对话原本通过 Host 固定的提供方与模型路由显示并发送消
|
||||
|
||||
## 决策
|
||||
|
||||
Web Host 为每个新建或恢复的 agent（智能体）复用 `installAgentLlmTarget`。如果会话已经使用过模型，目标从最新的 `request/header` 开始；否则采用 Host 默认值。`session.selectModel` 会更改会话级可变目标，提示词组装则将该目标与请求路由一并捕获，因此运行中步骤发生的切换会应用于下一个组装步骤。下一条实际采用的路由通过现有的完整 `request/header` 快照持久化；尚未进入请求的选择则仅保存在当前进程中。
|
||||
Web Host 为每个新建或恢复的 agent（智能体）复用 `installAgentLlmTarget`。如果会话已经使用过模型，提供方／模型／推理（reasoning）目标从最新的 `request/header` 开始；否则采用 Host 默认路由。`session.selectModel` 会更改会话级可变目标，提示词组装则将该目标与请求路由一并捕获，因此运行中步骤发生的切换会应用于下一个组装步骤。下一条实际采用的目标通过现有的完整 `request/header` 快照持久化；尚未进入请求的选择则仅保存在当前进程中。
|
||||
|
||||
会话 RPC 领域公开 `session.history` 的当前 `modelTarget`、`session.models` 模型目录与 `session.selectModel`。该目录从 LLM（大语言模型）注册表动态构建，并按提供方分组。各提供方目录会并发加载，且彼此独立失败，因此成功加载的分组仍可与可重试的失败记录一同使用。模型是否位于目录仅供参考：如果当前模型的已注册提供方没有列出该模型，系统会将其作为未列出行插入；在已注册提供方下选择未列出的模型仍然有效。
|
||||
会话 RPC 领域公开 `session.models` 模型目录与 `session.selectModel`。该目录从 LLM（大语言模型）注册表动态构建，并按提供方分组；每个已列出模型的精确元数据还会加入由适配器持有的推理强度 ID、名称、说明和可选默认值。各提供方的目录与精确元数据会按提供方并发加载，且彼此独立失败，因此成功加载的分组仍可与可重试的失败记录一同使用。模型是否位于目录仅供参考：如果当前模型的已注册提供方没有列出该模型，系统会将其作为未列出行插入；精确解析则决定路由与显式推理强度是否可用。选择操作通过 `resolveCallConfig` 拒绝不支持的推理强度 ID，并在更新目标前具体化适配器配置的默认值。
|
||||
|
||||
浏览器中的 `Session` 对象持有当前目标、分组目录、提供方失败记录、操作错误，以及 `idle`、`loading`、`ready`、`selecting`、`error` 状态。Host 会话的选择器挂载时会预加载目录，使紧凑型触发器能够解析目录名称；此后每次打开菜单都会刷新目录。常驻壳在 Workspace 选择连接或复用 Host 会话之前没有会话模型路由，因此其禁用的无会话输入栏不会分发选择器。目录与选择调用共用操作代次，防止较早响应覆盖较新结果；另设目标变更代次，使历史恢复即使与挂载时的目录刷新并发，也能还原日志记录的模型，同时防止旧历史覆盖用户选择。失败时保留先前的当前目标和可用分组。
|
||||
浏览器中的 `ModelService` 为每个实时会话持有一个 `ModelDirectory`。其快照包含当前完整目标、分组目录、提供方失败记录、操作错误，以及 `idle`、`loading`、`ready`、`selecting`、`error` 状态。挂载时会预先填充触发器标签，此后每次打开菜单都会刷新目录。目录与选择调用共用操作代次，防止较早响应覆盖较新结果；连接重置会先丢弃当前进程中的投影，再恢复 Host 目标。失败时保留先前的当前目标和可用分组。
|
||||
|
||||
`@deepseek-ai/dsh-client-ui-conversation` 将会话作用域的单实例 slot `conversation.input.model` 声明为其输入栏 entry 的子 slot。InputBar 在尾部控件区将该 seat 渲染于 pending 指示器与主按钮之前；该 seat 接收输入栏的 `locked` owner prop 与会话标准工具包。`@deepseek-ai/dsh-client-ui-model-selector` 占用该专用 seat，Host 拥有的空白会话 hero 也包括在内。其紧凑型触发器和单选菜单项显示目录名称；当前目标未列出时则回退到模型 ID。向上展开的菜单只显示一次提供方标题，同时提供键盘导航、关闭操作、重试状态和当前选择标记。
|
||||
`@deepseek-ai/dsh-client-ui-conversation` 将会话作用域的单实例 slot `conversation.input.model` 声明为其输入栏 entry 的子 slot。InputBar 在尾部控件区将该 seat 渲染于 pending 指示器与主按钮之前；该 seat 接收输入栏的 `locked` owner prop 与会话作用域。`@deepseek-ai/dsh-client-ui-model` 占用该 seat，并在同一目录上提供 `/model`。其紧凑型触发器显示目录中的模型名称与生效的推理强度标签；元数据缺失时则回退到相应 ID。向上展开的菜单首先提供 Model，并在当前精确模型支持时提供 Effort；Model 可深入提供方分组，Effort 可深入适配器排序的级别。仅当适配器没有配置模型默认值时，才显示提供方默认值行。
|
||||
|
||||
生产环境的浏览器名册是 `apps/cli/cordis.yml` 中的平铺 config tree；选择器对应其中一行 `dshClient` 配置项，而不是 Web boot 代码中硬编码的包。其包 manifest（元数据清单）仍声明对 `ui-conversation` 的图依赖边，激活则由 Cordis 服务可用性驱动。
|
||||
生产环境的浏览器名册是 `apps/cli/cordis.yml` 中的平铺 config tree；模型功能对应其中一行 `dshClient` 配置项，而不是 Web boot 代码中硬编码的包。其包 manifest（元数据清单）将加载顺序置于运行时与命令功能之后；Cordis 服务注入则等待 conversation slot 可用，再注册 composer 占用方。
|
||||
|
||||
## 考虑过的替代方案
|
||||
|
||||
@@ -26,6 +26,8 @@ Web Host 为每个新建或恢复的 agent（智能体）复用 `installAgentLlm
|
||||
|
||||
**在 Web 客户端中硬编码当前 DeepSeek 目录。** 该目录会与已注册适配器发生偏离，也会排除部署自有的提供方。LLM 注册表继续作为提供方与模型元数据的真源，也负责呈现部分查询失败。
|
||||
|
||||
**将 `High`／`Max` 保留为客户端本地 UI 状态。** 静态 DeepSeek 标签无法覆盖 `off`、pi-ai 的提供方词汇、适配器默认值与校验，也不能参与恢复或下一次提供方请求。精确模型元数据拥有可选词汇，会话目标则拥有已选择的 ID。
|
||||
|
||||
**将选择设为全局默认值。** 全局变更会意外改道其他已打开的对话。目标仅属于一个实时会话；对于没有已记录请求的会话，Host 配置仍是默认值。
|
||||
|
||||
**agent 运行期间拒绝更改。** 共享原子目标已经将当前组装步骤与下一次选择分离。保持选择器可用，可以让用户为下一个步骤预先选择模型，而不会改变正在执行的请求。
|
||||
@@ -34,8 +36,8 @@ Web Host 为每个新建或恢复的 agent（智能体）复用 `installAgentLlm
|
||||
|
||||
## 影响
|
||||
|
||||
任何由 Host 支撑的 Web 对话（包括空白会话）都可以在动态发现的提供方分组之间切换，而无需显示重复的 `provider/model` 标签；当前实际使用的路由会在恢复和重连后保留。目录名称仅用于呈现；选择和持久化仍然使用提供方/模型 ID。某个提供方的目录不可用时，只有相应分组会降级。路由变更可能降低提供方侧的缓存复用率，但选择器不会添加任何提示词内容，也不会干扰正在执行的步骤。常驻壳仅在没有当前会话时使用 Host 默认值且不暴露选择器。
|
||||
任何由 Host 支撑的 Web 对话（包括空白会话）都可以在动态发现的提供方分组和适配器持有的推理级别之间切换，而无需显示重复的 `provider/model` 标签。当前实际采用的目标会在恢复和重连后保留；目录名称仅用于呈现，而选择和持久化使用提供方／模型／推理强度 ID。某个提供方的目录或精确元数据不可用时，只有相应分组会降级。路由变更可能降低提供方侧的缓存复用率，但选择器不会添加任何提示词内容，也不会干扰正在执行的步骤。没有推理元数据的模型不显示 Effort 行。
|
||||
|
||||
## 测试
|
||||
|
||||
Host 测试固定分组发现、重复目录项隔离、部分提供方失败、已记录目标恢复、当前未列出目标、不可用提供方拒绝，以及切换仅影响下一次组装。客户端测试固定状态转换、失败时保留原状态、传输错误、过时响应栅栏、挂载与打开重叠、从历史记录恢复，以及快照引用稳定性。UI 测试固定专用模型 seat 的生命周期与锁定状态传播、显示目录名称并回退到 ID、提供方分组、单选语义、重试与错误状态、选择成功与失败、点击外部关闭，以及 Arrow/Home/End/Escape 键盘导航。无密钥 built-app fixture（测试前置数据）通过与生产环境同形的 boot 图加载选择器，选择 OpenAI 的 GPT-5，发起一个轮次，并验证下一条生成的响应会报告所选路由。
|
||||
Host 测试固定分组发现、目录与精确元数据失败隔离、已记录推理强度恢复、当前未列出目标、不支持的推理强度拒绝、默认值具体化，以及切换仅影响下一次组装。客户端测试固定共享目录、重连恢复与完整目标提交。组件测试固定动态推理强度标签、说明、提供方默认值展示与推理强度提交。无密钥 built-app fixture（测试前置数据）加载生产模型插件，选择 OpenAI 的 GPT-5 及其 Max 推理强度，发起一个轮次，并验证下一条生成的响应会报告两个 ID。
|
||||
|
||||
Reference in New Issue
Block a user