docs: record DeepSeek onboarding credential flow

This commit is contained in:
Yichen Jiang
2026-07-30 12:41:32 +08:00
parent 0b689e0d2c
commit 819a7a6751
9 changed files with 85 additions and 9 deletions
+2 -2
View File
@@ -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 packages/client/ui-models/README.md
README.md: 5bcfdcdbfe31ada89f787cd4d193e9763dba94d3
README.zh.md: 84b4b2187851506697de635d56691ca7e988ea00
README.md: 83a3d8c76ea52acd3a6ac1d76ea60bc8d3f614c0
README.zh.md: 202cfa756265e330cc74bd97e5c40d086073e642
+3 -1
View File
@@ -2,10 +2,12 @@
English | [中文](README.zh.md)
Models settings section plugin: the provider configuration page. It joins three wire domains into one surface — `llm.providers` (the configurable-provider directory with each route's live/dormant state), `settings.describe` (serialized schemas, layered redacted values, secret slots), and `credentials.describe` (value-free configured/source/writable badges) — and renders provider rows with one editor card at a time.
Models settings plugin: the provider configuration page and official-DeepSeek first-run credential overlay. It joins three wire domains into one shared snapshot — `llm.providers` (the configurable-provider directory with each route's live/dormant state), `settings.describe` (serialized schemas, layered redacted values, secret slots), and `credentials.describe` (value-free configured/source/writable badges) — and renders provider rows with one editor card at a time.
Rows are the *configured* providers (their profile resolves in the owning namespace); the add select's vocabulary is every dormant directory entry, so a bare-mounted `llm-pi-ai` offers its whole installed catalog before any route exists. The editor renders the provider's profile subtree through [`@deepseek-ai/dsh-client-schema-form`](../schema-form); the `credential-ref` role mounts the credential control, which shows the reference's live state and stores key values **write-only** through `credentials.set` — no value ever renders back. A row is deletable only when the user layer alone carries it (removal restores the composition base).
The first-run overlay projects `deepseek-official` readiness from that same joined snapshot. A configured literal `apiKey` secret sidecar or configured credential reference suppresses the prompt, including a read-only launch-environment credential. A mounted adapter with a writable missing reference opens the password form and writes only through `credentials.set`; success is accepted only after a fresh describe reports configured. An absent adapter is skipped because a browser form cannot mount Cordis plugins, while a present but unusable settings or credential capability produces a deployment diagnostic and an advanced link opens the Models section.
Apply semantics mirror the settings seam: an edit without removals lands as a minimal `settings.update` merge patch (stored secrets outside the patch survive), while a field reset or row deletion lands through `settings.replace` of the whole user section so removals actually take effect. The page refetches on the pushed invalidations (`settings/changed`, `credentials/changed`, `models/changed`, and `connection/reset`) once it has loaded, so an external `settings.yaml` edit, a second tab, or a settings-born route converges without polling.
## Model Experience
+3 -1
View File
@@ -2,10 +2,12 @@
[English](README.md) | 中文
模型设置分区插件:提供方配置页。它把三个协议领域汇聚为一个界面——`llm.providers`(可配置提供方目录,含每条路由的存活/休眠状态)、`settings.describe`(序列化 schema、分层脱敏值、secret 槽位)与 `credentials.describe`(不含值的 configured/source/writable 徽标)——并渲染提供方行,一次只展开一张编辑卡片。
模型设置插件:提供方配置页和 DeepSeek 官方首次使用凭据浮层。它把三个协议领域汇聚为一个共享快照:`llm.providers`(可配置提供方目录,含每条路由的存活/休眠状态)、`settings.describe`(序列化 schema、分层脱敏值、secret 槽位)与 `credentials.describe`(不含值的 configured/source/writable 徽标);页面据此渲染提供方行,一次只展开一张编辑卡片。
行是*已配置*的提供方(其 profile 在所属 namespace 中解析得出);新增选择框的词汇是全部休眠目录条目,因此裸挂载的 `llm-pi-ai` 在任何路由存在之前就能提供其完整的已安装 catalog。编辑器经 [`@deepseek-ai/dsh-client-schema-form`](../schema-form) 渲染该提供方的 profile 子树;`credential-ref` 角色会挂载凭据控件,它展示该引用的实时状态,并经 `credentials.set` 以**只写**方式存入密钥值——任何值都绝不回显。只有当某行仅由用户层承载时它才可删除(删除会还原组合 base)。
首次使用浮层从同一个联接快照得出 `deepseek-official` 的就绪状态。若 `apiKey` 字面量对应的 secret 槽位标记为已设置,或凭据引用已配置,浮层就不再显示,其中包括来自启动环境且只读的凭据。适配器已挂载、引用可写但尚未配置时,浮层会打开密码表单,且只经 `credentials.set` 写入;只有重新调用 describe 并确认已配置后,才会接受此次提交。适配器缺失时直接跳过,因为浏览器表单无法挂载 Cordis 插件;提供方存在但设置或凭据能力不可用时,则显示部署诊断,并通过高级设置链接打开 Models 分区。
「应用」语义与 settings seam 呈镜像:不含删除的编辑以最小的 `settings.update` 合并 patch 落地(patch 之外已存储的 secret 得以保留),字段重置或整行删除则经对整个用户分节的 `settings.replace` 落地,使删除真正生效。页面加载完成后会在推送的失效事件(`settings/changed`、`credentials/changed`、`models/changed` 与 `connection/reset`)上重拉,因此外部的 `settings.yaml` 编辑、第二个标签页或 settings 新生的路由都无需轮询即可收敛。
## 模型体验