Unified model input controls
统一模型输入控件
Separate DeepSeek and pi-ai row renderers let labels, icons, and spacing diverge. An automatic column count moves image input between the capacity row and a second row with small width changes. The image-support selector also cannot express image-only input, although both adapters accept it.
Source compatibility: Chinese title uses a compatible non-canonical header
English
Problem
Separate DeepSeek and pi-ai row renderers let labels, icons, and spacing diverge. An automatic column count moves image input between the capacity row and a second row with small width changes. The image-support selector also cannot express image-only input, although both adapters accept it.
Decision
Both catalogs render a shared model row. The two capacity fields occupy two columns, and Input types occupies the full next row. Text and Image use the controlled native Checkbox in ui-primitives, with caller-owned localized labels and native keyboard behavior. At least one type remains selected; the last checked type is disabled until another is selected.
An undeclared or empty input field displays the installed model input types, then the route default, then Text, without materializing an override on open. Pi-ai catalog discovery carries optional inputModalities through the LLM service and Remote response; adopted candidates copy it to input. Known provider editors read the installed catalog without endpoint I/O, so existing capacity-only rows also display inherited vision support. Catalog reads for a previous provider are discarded, and input controls wait for the current read to settle. The first checkbox edit writes the exact nonempty selection to DeepSeek's inputModalities or pi-ai's input. Existing explicit selections, including image-only input, remain visible. Unrelated metadata survives edits; disabling DeepSeek images removes its image request limits because the adapter rejects those limits without image input. Saving uses the existing catalog-array mutation and adapter validation.
Alternatives considered
Independent model-row renderers. Adapter-specific defaults and model discovery remain with their editors, but duplicate presentation creates avoidable visual drift. Shared rows keep those behaviors independent of field layout.
Keep the image-support selector. Settings-file-only capability editing leaves custom vision models effectively text-only. A selector with an explicit Default option preserves visible pi-ai catalog inheritance; a two-state image switch would misrepresent inherited vision support as disabled. The checkbox interface exposes both input types directly, resolving inherited capabilities before enabling edits. Merely opening a row still preserves inheritance; restoring it after a checkbox edit requires removing the input field in settings or resetting the catalog override.
Allow no checked types. DeepSeek rejects an empty list and pi-ai treats it as inheritance. Keeping a nonempty selection prevents the same gesture from meaning different things across adapters.
Keep image limits when disabling DeepSeek images. This leaves a configuration that the adapter refuses to save. Clearing the image-specific limits makes the selected text-only state valid while preserving unrelated model fields.
Consequences
Users can edit text-only, image-only, and combined declarations through one layout. Configuration declares upstream capabilities; it does not add image processing to a text-only model. Removing a pi-ai input declaration can change effective capabilities as catalog or provider defaults change. DeepSeek image limits need reconfiguration after images are disabled. Provider routing and catalog recovery retain their existing owners. Component checks cover defaults, exact selections, metadata preservation, and disabled controls; browser scenarios cover saved adapter capabilities, reopened selections, the separate input-type row, and installed vision metadata across discovery, adoption, and reopening.
中文
问题
DeepSeek 和 pi-ai 分别渲染模型行，导致标签、图标和间距不一致。自动列数会在宽度稍有变化时，将图片输入移到容量字段同一行或下一行。图片支持选择器也无法表达仅图片输入，而两个适配器都接受这种声明。
决策
两个模型目录渲染同一个模型行组件。两个容量字段占据两列，输入类型独占下一整行。文本和图片使用 ui-primitives 的受控原生 Checkbox（复选框），由调用方提供本地化标签，并保留原生键盘行为。至少保留一种输入类型；最后一个勾选项保持禁用，直到选中另一项。
未声明或为空的输入字段优先显示已安装模型的输入类型，其次是路由默认值，最后回退为文本，打开时不会生成覆盖值。Pi-ai 目录发现将可选的 inputModalities 经 LLM 服务和 Remote（远程调用）响应传递给界面；采纳候选时将其复制到 input。已知提供方的编辑器读取已安装目录，不向端点发送请求，因此已有的仅容量模型行也能显示继承的视觉能力。上一个提供方的目录响应会被丢弃，输入控件会等待当前读取完成。首次修改复选框会将确切的非空选择写入 DeepSeek 的 inputModalities 或 pi-ai 的 input。已有显式选择保持可见，包括仅图片输入。编辑保留无关元数据；禁用 DeepSeek 图片时会移除图片请求限制，因为适配器在没有图片输入时拒绝这些限制。保存使用现有的模型目录数组变更和适配器校验。
考虑过的替代方案
独立的模型行渲染器。 适配器专属默认值和模型发现仍由各自编辑器负责，但重复的展示实现会造成可避免的视觉差异。共享模型行将这些行为与字段布局分离。
保留图片支持选择器。 仅通过设置文件编辑能力，会让自定义视觉模型实际保持仅文本。带有显式「默认」选项的选择器会清楚展示 pi-ai 模型目录继承；两态图片开关则会将继承的视觉能力误表示为禁用。复选框界面直接展示两种输入类型，并在允许编辑前解析继承的能力。仅打开模型行仍保留继承；修改复选框后，要恢复继承需在设置中移除输入字段，或重置整个模型目录覆盖。
允许全部取消。 DeepSeek 拒绝空列表，pi-ai 则将其视为继承。保留非空选择，避免同一个操作在两个适配器中具有不同含义。
禁用 DeepSeek 图片时保留图片限制。 这会留下适配器拒绝保存的配置。清除图片专属限制，可以让选中的仅文本状态有效，同时保留模型的无关字段。
影响
用户通过相同布局编辑仅文本、仅图片和组合声明。配置声明上游能力，不会为仅文本模型增加图片处理能力。移除 pi-ai 输入声明后，有效能力可能随模型目录或提供方默认值变化。禁用图片后，DeepSeek 图片限制需要重新配置。提供方路由和模型目录恢复规则保留现有责任方。组件检查覆盖默认值、确切选择、元数据保留和禁用控件；浏览器场景覆盖保存后的适配器能力、重新打开后的选择，独占一行的输入类型，以及发现、采纳和重新打开过程中传递的内置视觉元数据。