@deepseek-ai/dsh-client-connection 0.1.1-rc.2 → 0.1.2-alpha.2

package/README.i18n.yaml

@@ -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/connection/README.md
-README.md: 71ef204a589bb67c15ccab58d3cac5a13782ce27
-README.zh.md: 6d33ac3c13cdfceeba6e7472b618084267d09bbc
+README.md: ee0566bf811a0bad32f7d59b4c8849e26efea613
+README.zh.md: 48425a428247b7749bf0dd75f592b5f7e4088ef2

package/README.md

@@ -1,17 +1,51 @@
+---
+description: "Browser-host wire layer for the web GUI: Remote RPC, event-stream delivery with reconnect, exact Fetch routes, the /api HTTP bridge, and the browser-trust fence."
+kind: "package-reference"
+---
+
 # @deepseek-ai/dsh-client-connection
 
 English | [中文](README.zh.md)
 
-Wire consumer layer: the client plugin's apply mounts `ctx.connection` (shared api client + current-page loopback state + observable generation-scoped `hostDescription` + single-consumer stream-loop starter); the export face carries the wire contract types, the `AbstractApiClient` abstraction, and the loop's sink/config types. Each successful readiness handshake publishes the exact `host.describe` value before `onConnected`; generation loss and explicit stop clear it, so native-capability consumers never retain a disconnected answer. The browser carrier uses HTTP POST for unary and respond operations and opens one downlink-only WebSocket each for `events.mux` and `events.host`; the in-process carrier satisfies the same two-stream abstraction. The exported `ClientTransportHooks` names the page global `__DSH_TRANSPORT__` that replaces the browser carrier wholesale: the served web app leaves it unset and gets HTTP + WebSocket, while a shell owning a different physical transport (the worker preview's postMessage tunnel) provides `createApiClient` and `fetch` — plus `loadBundle` when it also owns bundle bytes — instead of forking the plugin. The Host half owns the single `/api` route and its Fetch bridge; a registered Typert interceptor claims its Remote endpoints before the API Proxy fallback. Loopback hostname classification stays package-internal: the `/api` Host fence and WebSocket upgrades use it directly, while other client plugins consume the derived `ctx.connection.isLoopback` state. The node half's `/api` route pins the privileged method set (`host.pickDirectory`, `host.openPath`, and the whole configuration plane — `settings.describe`/`openDocument`/`update`/`replace`/`mutate` and `credentials.describe`/`set`/`unset`; reads and native actions included, since describing returns the exposed configuration, opening acts on the Host desktop, and probing an arbitrary reference reports where a credential comes from — and the agent-preset authoring plane, `agentPreset.read`/`copy`/`openDocument`/`remove`, since a composition names the plugins a session runs, so reading one is reconnaissance, and copy/remove/openDocument manage the roster and drive the host desktop (authoring is copy-only, so none of them accepts composition text or a path); `agentPreset.list` and `agentPreset.select` stay out — the roster carries only ids and trust, and choosing a preset grants nothing `session.create`'s own `agentPreset` did not, over a default that already carries bash) to loopback by passing the trust fence with an empty trust list — a declared `trustedHosts` authority reaches every other method, while these stay loopback-local until a real authentication layer exists. The platform carriers and ConnectionController loop are package-internal; apply selects and drives them. The downlink boundary is documented in the [WebSocket downlink carrier Agent Note](../../../.agents/notes/implemented/architecture/2026-08-04-websocket-downlink-carrier.md).
+## Summary
+
+The package carries browser-to-Host Remote calls, exact Fetch responses, and connection generations. The Client plugin mounts `ctx.connection` with current-page loopback state, a generic RPC carrier, the active generation and its Host facts, observable recovery state, an immediate reconnect command, and the registration point for one generation source. A generation becomes visible when its source reports ready; source completion, failure, withdrawal, or an explicit stop clears it before `ConnectionController` applies its retry policy.
+
+## Table of Contents
+
+- [Use this package](#use-this-package)
+- [Browser authentication and request trust](#browser-authentication-and-request-trust)
+- [Connection generation](#connection-generation)
+- [Model Experience](#model-experience)
+- [Known Limitations and Deferred Work](#known-limitations-and-deferred-work)
+- [Dev Note](#dev-note)
+
+-----
+
+<a id="use-this-package"></a>
+## Use this package
+
+The browser uses HTTP POST for Remote unary calls. API Gateway owns the `/api/remote.mux` WebSocket and its logical streams; in-process compositions provide equivalent Remote streams through `connection.rpc.open` without opening a WebSocket. The Host half owns the sole `/api` route, Fetch bridge, browser authentication, Host/Origin checks, and exact `GET`/`HEAD` route registry. Typert Gateway claims generated Remote endpoints, feature packages register non-JSON responses such as Session-log downloads, and unclaimed requests return 404. Loopback hostname classification remains package-internal to the browser-facing Client state.
+
+-----
+
+<a id="browser-authentication-and-request-trust"></a>
+## Browser authentication and request trust
 
-## /api browser-trust fence
+Every Host RPC method and WebSocket stream requires one browser session; there is no method-specific loopback tier. Each process mints a random launch token. `dsh-web-app` prints and opens the ordinary root URL with `?token=...`; `frontend-static` delegates root and index requests to `ctx.connection.authorizeIndex`, which accepts that token only on `GET /`, writes an authority-bound signed cookie, and redirects to clean `/`. A missing, expired, malformed, or wrong-authority cookie returns 401 before RPC dispatch. Static assets remain public. The HTTP carrier accepts no query token outside the root exchange and no Authorization-header token.
 
-The node half guards every entry under `/api` before bridging or upgrading (`src/api-request-trust.ts`). Every request — browser-marked or not — must present a `Host` that is a loopback authority or matches a `trustedHosts` entry: exact on `host:port` entries, any port on port-less entries, both sides compared through WHATWG normalization (DNS-rebinding defense). There is deliberately no shortcut for unmarked HTTP requests: over plain HTTP a browser attaches neither `Origin` nor Fetch-Metadata to image and navigation reads, so an unmarked request may still be a rebound browser read with a readable response, and Host is the one header rebinding cannot forge; a browser WebSocket handshake carries `Origin` and passes the same comparison. Non-browser clients pass the same fence via loopback, deployment-derived LAN IP literals, or a declared authority. When markers are present, an attached `Origin` must equal the Host authority, and an explicit `sec-fetch-site: cross-site` marker is refused. A `trustedHosts` entry that is not a bare, canonical `host[:port]` authority — one WHATWG parsing reads back exactly as written — fails the plugin load loudly: parsing would otherwise quietly authorize the hostname inside `harness.internal/path`, or broaden a dangling-colon or zero-padded port to an any-port grant. HTTP failures answer plain 403 before any RPC dispatch; upgrade failures reject the handshake before any event stream starts. Non-loopback compositions must trust their serving authorities explicitly: the Web runtime derives LAN IP literals from an all-interfaces server config, while `trustedHosts` in cordis.yml and the CLI's `--trusted-host` flag declare named authorities. `dsh web --host 0.0.0.0` is intentionally unsupported until remote access has an authentication layer. The fence is a reachability policy, not authentication; the Web carrier provides no authentication layer. Decision record: [the api browser-trust boundary Agent Note](../../../.agents/notes/implemented/architecture/2026-07-28-api-browser-trust-boundary.md).
+The cookie signing secret is the owner-scoped `client-connection/browser-session` grant record in `ctx.credentials`. The local provider persists it in `$DSH_HOME/.credentials.yaml`; `BrowserAuth` loads or creates the record during Connection activation and retains the secret in memory, so request authentication is synchronous. Deleting or replacing the record takes effect on the next Connection activation. Cookies carry an absolute issue/expiry interval, defaulting to 30 days through `cookieMaxAgeDays`, and bind the normalized hostname plus port in both their deterministic name and signed payload. They are host-only, `Path=/`, `HttpOnly`, and `SameSite=Strict`; they deliberately omit `Secure` because the shipped server uses loopback HTTP.
 
-## `/api` WebSocket downlinks
+Before authentication, every request still passes `src/api-request-trust.ts`. Its `Host` must be loopback or match a `trustedHosts` entry: exact on `host:port`, any port on port-less entries, both sides WHATWG-normalized. An attached `Origin` must equal that Host and `sec-fetch-site: cross-site` is refused. Malformed configured authorities fail plugin load. These checks defend DNS rebinding and cross-site browser requests; they never establish identity. A failed Host/Origin check returns 403, while a trusted but unauthenticated request returns 401. `dsh web --host 0.0.0.0` remains unsupported. Decision records: [browser request trust](../../../.agents/notes/implemented/architecture/2026-07-28-api-browser-trust-boundary.md) and [browser token authentication](../../../.agents/notes/implemented/architecture/2026-08-24-browser-token-authentication.md).
 
-`/api/events.mux` and `/api/events.host` each accept a WebSocket upgrade and send only the corresponding `ServerRequest` text messages to the browser; the client sends no application data over these sockets. If either socket ends, the current connection generation fails and rebuilds both streams; readiness still requires both sockets to be open and the `host.describe` HTTP call to succeed. Host teardown terminates both sockets, aborts their sources, and waits for source cleanup before returning. Ordinary network GETs to these paths return 426 with no SSE fallback; `toFetchHandler`'s SSE codec serves only the isomorphic in-process carrier.
+<a id="connection-generation"></a>
+## Connection generation
 
+API Gateway Client registers the internal `$events` logical stream as the sole generation source, independently of whether any `$on` listener exists. The Host attaches all incremental listeners in the API Remotes source factory, then sends one `{ type: 'ready', clientId, host: { home } }` item before events. `ConnectionController` publishes that generation and calls `onConnected` only after the ready item arrives, so baseline acquisition cannot race ahead of incremental observation.
+
+An ended `$events` stream, a Remote stream error, a non-ready opening item, or a malformed event item invalidates the current generation. While the browser reports network availability, the controller publishes `connecting` and retries with 50%–100% jitter under caps of 500ms, 1s, 2s, 4s, 8s, and 10s. It logs each attempt, asks Gateway to replace the physical WebSocket, and reopens `$events`; failure in the 10s tier publishes terminal `disconnected`. `ctx.connection.reconnect()` interrupts active work, resets the sequence, and starts retry 1 immediately. Browser `offline` aborts active work, publishes `disconnected`, and suspends automatic attempts; the next `online` transition resets the sequence and starts at the 500ms tier. A ready item publishes `connected`. The Gateway mux performs one physical connection attempt per request rather than running an independent retry schedule. The [connection recovery decision](../../../.agents/notes/implemented/feature/2026-08-28-web-connection-recovery-control.md) owns the cadence and manual recovery behavior.
+
+<a id="model-experience"></a>
 ## Model Experience
 
 None, as the wire consumer layer moves already-composed messages between browser and host; nothing here reaches a model request.
@@ -22,5 +56,19 @@ None; this package neither assembles nor sends a provider request.
 
 ## Known Limitations and Deferred Work
 
-- **History resumes an unattached session** — opening history may create the host-side agent and add latency to the first open; there is no persistence-only read path.
+<a id="known-limitations-and-deferred-work"></a>
+
 - **The `/api` bridge buffers each request body in memory** — `maxRequestBodyBytes` (default 300 MiB, sized for the default 200 MiB aggregate image limit after base64 expansion plus envelope headroom) is therefore also the per-request resident bound; a streaming body path would be needed to lower it without shrinking the image limits.
+- **The browser cookie is not marked `Secure`** — loopback HTTP is the shipped transport, so exposing the same authority over plaintext networking can expose the bearer cookie in transit.
+- **There is no logout operation** — clearing the browser cookie ends one browser session; deleting the owner credential record and restarting `dsh` revokes every session.
+
+
+<a id="dev-note"></a>
+### Dev Note
+
+<details>
+<summary>Working context for maintainers — click to expand</summary>
+
+None.
+
+</details>

package/README.zh.md

@@ -1,17 +1,51 @@
+---
+description: "Web GUI 的浏览器-Host 线层：Remote RPC、带重连的事件流投递、精确 Fetch 路由、/api HTTP 桥与浏览器信任栅栏。"
+kind: "package-reference"
+---
+
 # @deepseek-ai/dsh-client-connection
 
 [English](README.md) | 中文
 
-协议消费层：客户端插件的 apply 会挂载 `ctx.connection`（共享 API 客户端 + 当前页面的 loopback 状态 + 可观察且按 generation 生效的 `hostDescription` + 单消费方流循环启动器）；导出表层携带协议约定类型、`AbstractApiClient` 抽象，以及循环的 sink／配置类型。每次就绪握手成功后，都会在 `onConnected` 之前发布完整的 `host.describe` 值；generation 失效或显式 stop 会清空它，因此原生能力消费者不会保留已经断线的判断。浏览器载体以 HTTP POST 发送 unary／respond，并为 `events.mux` 与 `events.host` 各开一条只下行的 WebSocket；进程内载体满足同一双流抽象。导出的 `ClientTransportHooks` 命名了整体替换浏览器载体的页面全局量 `__DSH_TRANSPORT__`：served web app 不设置它、走 HTTP + WebSocket；拥有另一种物理传输的壳（worker 预览的 postMessage 隧道）则在此提供 `createApiClient` 与 `fetch`——当它同时持有 bundle 字节时再加 `loadBundle`——而不必 fork 本插件。Host half 持有唯一 `/api` route 及其 Fetch bridge；已注册的 Typert interceptor 会先认领自己的 Remote endpoint，未认领请求再回退 API Proxy。Loopback hostname 判定逻辑留在包内部：`/api` Host fence 与 WebSocket upgrade 会直接使用它，其他客户端插件则消费派生的 `ctx.connection.isLoopback` 状态。node 半侧的 `/api` 路由让特权方法集（`host.pickDirectory`、`host.openPath`，以及整个配置面——`settings.describe`/`openDocument`/`update`/`replace`/`mutate` 与 `credentials.describe`/`set`/`unset`；读取与原生操作也在内，因为 describe 会返回已暴露的配置、打开操作会作用于 Host 桌面，而探测任意引用会报出某条凭据来自何处——以及 agent（智能体） preset 的创作面 `agentPreset.read`/`copy`/`openDocument`/`remove`，因为组装指明了一个会话所运行的插件，读取它是侦察，而 copy/remove/openDocument 管理名单并驱动宿主桌面（创作只有复制一种写入，因此这些方法都不接收组装文本或路径）；`agentPreset.list` 与 `agentPreset.select` 不在其中——名单只携带 id 与信任级别，而选择一个 preset 并不比 `session.create` 自带的 `agentPreset` 多给任何能力，何况默认 preset 本就带着 bash）以空信任表过信任 fence，从而钉在回环——已声明的 `trustedHosts` 授权可达其余全部方法，而这些方法在真正的认证层出现之前仍只限回环本机。平台载体与 ConnectionController 循环属于包内部；apply 负责选择并驱动它们。下行边界见 [WebSocket 下行载体 Agent Note](../../../.agents/notes/implemented/architecture/2026-08-04-websocket-downlink-carrier.zh.md)。
+## 概述
+
+本包承载浏览器到 Host 的 Remote 调用、精确 Fetch 响应与 connection generation。Client 插件挂载 `ctx.connection`，其中包含当前页面的 loopback 状态、通用 RPC carrier、当前 generation 及其 Host 信息、可观察的恢复状态、立即重连命令，以及单一 generation source 的注册点。source 报告 ready 后 generation 才可见；source 结束、失败、被撤回或显式 stop 都会清空它，再由 `ConnectionController` 执行重试策略。
+
+## 目录
+
+- [使用本包](#use-this-package)
+- [浏览器认证与请求信任](#browser-authentication-and-request-trust)
+- [Connection generation](#connection-generation)
+- [模型体验](#model-experience)
+- [已知限制与暂缓事项](#known-limitations-and-deferred-work)
+- [开发备注](#dev-note)
+
+-----
+
+<a id="use-this-package"></a>
+## 使用本包
+
+浏览器通过 HTTP POST 执行 Remote 一元调用；API Gateway 自己拥有 `/api/remote.mux` WebSocket 及其逻辑流。进程内组合通过 `connection.rpc.open` 提供等价的 Remote 流，不打开 WebSocket。Host half 拥有唯一 `/api` route、Fetch bridge、浏览器认证、Host/Origin 校验与精确 `GET`/`HEAD` 路由注册表。Typert Gateway 认领生成的 Remote endpoint，功能包注册 Session 日志下载等非 JSON 响应，未认领的请求返回 404。Loopback hostname 判定只供浏览器侧当前页面状态使用，留在包内。
+
+-----
+
+<a id="browser-authentication-and-request-trust"></a>
+## 浏览器认证与请求信任
 
-## /api 浏览器信任栅栏
+每个 Host RPC 方法和 WebSocket stream 都要求同一个浏览器会话，不存在按方法区分的 loopback 层。每个进程生成一个随机启动令牌。`dsh-web-app` 打印并打开带 `?token=...` 的普通根 URL；`frontend-static` 把根路径和 index 请求交给 `ctx.connection.authorizeIndex`，后者只在 `GET /` 接受该令牌，写入绑定 authority 的签名 cookie，再重定向到干净的 `/`。缺失、过期、畸形或 authority 不匹配的 cookie 会在 RPC 分发前得到 401。静态资源保持公开。HTTP 载体不在根路径交换之外接受 query token，也不接受 Authorization header token。
 
-node 半侧在桥接或 upgrade 前守卫 `/api` 下的每个入口（`src/api-request-trust.ts`）。每个请求——无论是否带浏览器标记——`Host` 都必须是回环地址权威，或与某个 `trustedHosts` 条目匹配：带端口的 `host:port` 条目精确匹配，不带端口的条目匹配任意端口，两侧均经 WHATWG 归一化后比较（DNS rebinding 防御）。刻意不为无浏览器标记的 HTTP 请求开捷径：明文 HTTP 下浏览器的图片与导航读取既不带 `Origin` 也不带 Fetch-Metadata，因此无标记请求仍可能是被重绑页面发起的、响应可被读走的读取，而 Host 是重绑唯一伪造不了的请求头；WebSocket 浏览器握手会带 `Origin` 并通过同一道比较。非浏览器客户端经由回环地址、部署推导的 LAN IP 字面量或已声明的权威通过同一道栅栏。当标记存在时，如附带 `Origin`，则它必须与 Host 权威完全一致；显式的 `sec-fetch-site: cross-site` 标记一律拒绝。不是纯的、规范形 `host[:port]` 权威的 `trustedHosts` 条目——即 WHATWG 解析读回后与原文不完全一致的——会让插件加载明确报错：否则解析会悄悄授权 `harness.internal/path` 这类笔误里的 hostname，或把悬空冒号、补零端口放大成任意端口授权。HTTP 失败在任何 RPC 分发之前以纯 403 应答，upgrade 失败在启动任何事件流前拒绝握手。非回环组合必须显式信任其服务权威：Web 运行时从全接口服务器配置推导 LAN IP 字面量，cordis.yml 中的 `trustedHosts` 与 CLI（命令行界面）的 `--trusted-host` flag 则声明具名权威。`dsh web --host 0.0.0.0` 在远程访问具备认证层之前有意不受支持。这道栅栏是可达性策略，而不是认证；Web 载体不提供认证层。决策记录：[api 浏览器信任边界 Agent Note](../../../.agents/notes/implemented/architecture/2026-07-28-api-browser-trust-boundary.zh.md)。
+cookie 签名密钥是 `ctx.credentials` 中由 `client-connection/browser-session` 拥有的 grant 记录。本地提供方把它持久化到 `$DSH_HOME/.credentials.yaml`；`BrowserAuth` 在 Connection 激活期间加载或创建该记录，并把密钥留在内存中，因此请求认证同步执行。删除或替换该记录会在下一次 Connection 激活时生效。cookie 携带绝对签发与过期区间，`cookieMaxAgeDays` 默认设为 30 天，并在确定性名称与签名 payload 中同时绑定规范化 hostname 和 port。它是 host-only、`Path=/`、`HttpOnly`、`SameSite=Strict`；随附服务器使用 loopback HTTP，因此刻意不设置 `Secure`。
 
-## `/api` WebSocket 下行
+认证之前，每个请求仍经过 `src/api-request-trust.ts`。其 `Host` 必须是 loopback，或与 `trustedHosts` 条目匹配：带端口的 `host:port` 精确匹配，不带端口的条目匹配任意端口，两侧均经 WHATWG 归一化。若附带 `Origin`，它必须等于该 Host；`sec-fetch-site: cross-site` 一律拒绝。畸形配置 authority 会让插件加载失败。这些检查防御 DNS rebinding 与跨站浏览器请求，绝不建立身份。Host/Origin 校验失败返回 403；Host 可信但未认证的请求返回 401。`dsh web --host 0.0.0.0` 仍不受支持。决策记录：[浏览器请求信任](../../../.agents/notes/implemented/architecture/2026-07-28-api-browser-trust-boundary.zh.md)与[浏览器令牌认证](../../../.agents/notes/implemented/architecture/2026-08-24-browser-token-authentication.zh.md)。
 
-`/api/events.mux` 与 `/api/events.host` 各接受一条 WebSocket upgrade，并只向浏览器发送对应的 `ServerRequest` 文本消息；客户端不会在这些 socket 上发送业务数据。任一 socket 结束都会使当前 connection generation 失败并重建两条流，连接就绪仍要求两条 socket 均已打开且 `host.describe` HTTP 调用成功。Host teardown 会终止两条 socket、中止各自的 source，并等待 source 清理完成后再返回。普通网络 GET 这些路径会返回 426，不保留 SSE（Server-Sent Events）回退；`toFetchHandler` 的 SSE 编解码只服务进程内同构载体。
+<a id="connection-generation"></a>
+## Connection generation
 
+API Gateway Client 把内部 `$events` logical stream 注册为唯一 generation source，与有无 `$on` 订阅无关。Host 在 API Remotes source factory 同步挂好所有增量 listener 后，先发送唯一 `{ type: 'ready', clientId, host: { home } }` 项，再发送事件。`ConnectionController` 仅在收到该 ready 项后发布 generation 并调用 `onConnected`，因此 baseline 不会跑在增量 listener 前面。
+
+`$events` 结束、返回 Remote stream error、收到非 ready 首项或畸形事件项，都会使当前 generation 失效。浏览器报告网络可用时，Controller 发布 `connecting`，并在 500ms、1s、2s、4s、8s 与 10s 上限内采用 50%–100% 抖动重试。它记录每次尝试、要求 Gateway 替换物理 WebSocket，再重开 `$events`；10s 档失败后发布终态 `disconnected`。`ctx.connection.reconnect()` 会中断活动工作、重置序列，并立即开始 retry 1。浏览器 `offline` 会中断活动工作、发布 `disconnected` 并暂停自动尝试；下一次 `online` 转换会重置序列并从 500ms 档开始。ready 项会发布 `connected`。Gateway mux 每次收到请求只做一次物理连接尝试，不再运行另一套重试调度。[连接恢复决策](../../../.agents/notes/implemented/feature/2026-08-28-web-connection-recovery-control.zh.md)规定重试节奏和手动恢复行为。
+
+<a id="model-experience"></a>
 ## 模型体验
 
 无。协议消费层只在浏览器与主机之间搬运已经组合好的消息；这里没有任何内容进入模型请求。
@@ -22,5 +56,19 @@ node 半侧在桥接或 upgrade 前守卫 `/api` 下的每个入口（`src/api-r
 
 ## 已知限制与暂缓事项
 
-- **History 会恢复未附加的会话**：打开 history 可能创建宿主侧 agent，并增加首次打开的延迟；没有仅从持久化读取的路径。
+<a id="known-limitations-and-deferred-work"></a>
+
 - **`/api` 桥把每个请求体整体缓冲在内存里**：`maxRequestBodyBytes`（默认 300 MiB，按默认 200 MiB 图片总量上限经 base64 膨胀加信封余量得出）因此同时是单请求的驻留内存上界；要降低它而不缩小图片限额，需要流式请求体路径。
+- **浏览器 cookie 不带 `Secure`**：随附载体是 loopback HTTP；若部署经明文网络暴露同一 authority，bearer cookie 可能在传输中泄露。
+- **没有 logout 操作**：清除浏览器 cookie 会结束单个浏览器会话；删除 owner 凭据记录并重启 `dsh` 会撤销全部会话。
+
+
+<a id="dev-note"></a>
+### 开发备注
+
+<details>
+<summary>维护者工作上下文——点击展开</summary>
+
+无。
+
+</details>