dsh-cherry-bridge 0.1.8 → 0.1.10

package/README.en.md

@@ -38,13 +38,13 @@ flowchart TB
 
 ```bash
 # From npm (recommended, pinned)
-npx @deepseek-ai/dsh plugin --profile web add dsh-cherry-bridge@0.1.8
+npx @deepseek-ai/dsh plugin --profile web add dsh-cherry-bridge@0.1.10
 
 # Install the latest npm release
 npx @deepseek-ai/dsh plugin --profile web add dsh-cherry-bridge
 
 # From GitHub source (same commit as the npm release)
-npx @deepseek-ai/dsh plugin --profile web add github:SAXEM1997/dsh-cherry-bridge#v0.1.8
+npx @deepseek-ai/dsh plugin --profile web add github:SAXEM1997/dsh-cherry-bridge#v0.1.10
 ```
 
 The package ships `cordis.patch.yml` (`package.json`'s `dsh.bundle.patch` points
@@ -92,7 +92,7 @@ cherry_interactions → see pending question/approval cards; cherry_respond answ
 | --- | --- | --- | --- |
 | `dataDir` | string | `$DSH_HOME/cherry-remote` (fallback `~/.dsh/cherry-remote`) | Plugin data directory holding `identity.key` and `devices.json` (both `0600`). |
 | `perSessionCacheBytes` | number | `2097152` (2 MiB) | Per-session cap on the **local** subscription event cache; beyond it the oldest events are dropped LRU-style with a `warn`. |
-| `rateBudgetPerSecond` | number | `32` | Per-device smoothed request budget (req/s). The desktop's hard limit is 64 req/s with no negotiation channel; exceeding it terminates the session with close `1011` — hence the conservative value. |
+| `rateBudgetPerSecond` | number | `32` | Per-device smoothed request budget (req/s). The desktop's shared allowance is 64 req/s through 2.1.6 and 256 req/s from 2.1.7, and only the newer generation downgrades an overrun to a single `RATE_LIMITED` refusal (the older one terminates the session with close `1011`) — hence a value sized for the harsher generation. Content reads (`agent.content.read`) derive their own tier at 8:1, **capped at 512/s** (that allowance is the same in both generations). |
 
 > ⚠️ `cordis.patch.yml`'s `config` is **whole-replace**, not merged with the
 > defaults. Removing a key gets you the plugin default; a key that exists but is
@@ -106,7 +106,7 @@ cherry_interactions → see pending question/approval cards; cherry_respond answ
 | `cherry_agents` | List the agents available on a device (paginated). A truncated answer always carries `truncated` + `truncationReason` (`limit` or `safety-bound`, the latter the 1000-item internal ceiling); when the stop point can be named it also carries a `resumeCursor`: replay it as `cursor` with the same device and filters to continue from that stop point (`limit` then bounds the next answer). A token from another tool, device or filter set is refused **locally**, never forwarded. |
 | `cherry_workspaces` | List an agent's **registered** workspaces, and whether a system workspace exists. A truncated answer always carries `truncated` + `truncationReason`; when the stop point can be named it also carries a `resumeCursor` — replay it as `cursor` with the same filters to continue. |
 | `cherry_sessions` | List an agent's sessions, optionally filtered by workspace. A truncated answer always carries `truncated` + `truncationReason` (`limit` or `safety-bound`); when the stop point can be named it also carries a `resumeCursor`: replay it as `cursor` with the same device and **identical filters** to continue (`limit` then bounds the next answer); a token from another tool, device or filter set is refused **locally**. |
-| `cherry_read` | Read a session's message history snapshot, **newest-first** (`limit` drops the **oldest** messages). Attachment parts come back as `name`/`mediaType`/`byteLength` with an empty `content` — the bytes are never fetched. A truncated walk always says so (`truncated` + `truncationReason`): `limit` means your own ceiling was reached and raising `limit` helps; `aggregate-bytes` means the answer hit its **output size limit** — independent of `limit` (a larger `limit` only costs more round trips and stops at the same budget). **The older messages are reached by the resume token**: pass the `messagesResumeCursor` a truncated answer returned as this tool's own `messagesResumeCursor` argument to continue from that stop point (`limit` then bounds how many messages come back **from there**), walking a long history in successive answers. When the stop point cannot be named the token is **absent** (`truncated` stays `true`, and the absence does **not** mean the history ended); a wider `limit` is not a general remedy — it moves a `limit` stop only, while an `aggregate-bytes` stop leaves the older messages unreachable in that answer (see [Errors and limits](#errors-and-limits)). Two token forms: `cursor:<peer position>` (the peer page position after a whole window was consumed) and `after:<messageId>` (the walk stopped inside its window; the resume scans from the newest end to that message and returns the older ones). A token this tool did not emit is **refused locally** (never forwarded — the desktop silently falls back to the newest page for one), and an `after:` anchor that is no longer in the history is an **error**, not an empty result. |
+| `cherry_read` | Read a session's message history snapshot, **newest-first** (`limit` drops the **oldest** messages). A failed message carries `failure: {reasonCode, retryable, message}` when the peer sent one (its classification, whether a retry is worthwhile, and its own bounded summary; the peer's `source`/`context` are not projected). Attachment parts come back as `name`/`mediaType`/`byteLength` with an empty `content` — the bytes are never fetched. A truncated walk always says so (`truncated` + `truncationReason`): `limit` means your own ceiling was reached and raising `limit` helps; `aggregate-bytes` means the answer hit its **output size limit** — independent of `limit` (a larger `limit` only costs more round trips and stops at the same budget). **The older messages are reached by the resume token**: pass the `messagesResumeCursor` a truncated answer returned as this tool's own `messagesResumeCursor` argument to continue from that stop point (`limit` then bounds how many messages come back **from there**), walking a long history in successive answers. When the stop point cannot be named the token is **absent** (`truncated` stays `true`, and the absence does **not** mean the history ended); a wider `limit` is not a general remedy — it moves a `limit` stop only, while an `aggregate-bytes` stop leaves the older messages unreachable in that answer (see [Errors and limits](#errors-and-limits)). Two token forms: `cursor:<peer position>` (the peer page position after a whole window was consumed) and `after:<messageId>` (the walk stopped inside its window; the resume scans from the newest end to that message and returns the older ones). A token this tool did not emit is **refused locally** (never forwarded — the desktop silently falls back to the newest page for one), and an `after:` anchor that is no longer in the history is an **error**, not an empty result. |
 | `cherry_events` | Read the session's buffered events newer than `cursor`, acking them after delivery. The answer is `{events, nextCursor}`; pass that `nextCursor` back as `cursor` to continue from it. |
 | `cherry_create` | Create a session in an agent workspace (workspace selection is **required**; identity is read back). |
 | `cherry_send` | Send a message, by default waiting for a terminal state and returning the assistant text; `wait: false` returns the command receipt immediately. `messagesTruncated`/`messagesTruncationReason` report a **message**-level cut (older messages dropped); `deadline` means the **answer's own assembly budget** ran out (see below), with the content read so far still returned. |
@@ -133,9 +133,15 @@ subscription was lost and could not be rebuilt, so the outcome is
 
 ## Errors and limits
 
-- Read methods are replayed **once** after a reconnect; **commands are never
-  replayed** — a write that loses its reply is recovered by retrying with the
-  **original `commandId`**, and the desktop's command ledger replays the receipt.
+- Read methods are replayed **once** after a reconnect, and **once** after a
+  desktop >= 2.1.7 rate-limit refusal (`RATE_LIMITED` + `retryAfterMs`, waited
+  out within the caller's budget); **commands are never replayed** — a write
+  that loses its reply is recovered by retrying with the **original
+  `commandId`**, and the desktop's command ledger replays the receipt. A
+  rate-limited COMMAND is the other case: the desktop refuses it **before
+  dispatch**, so this attempt was never applied — the error text says so, names
+  the wait the peer asked for (`retryAfterMs`, omitted when it sent none), and
+  points at `rateBudgetPerSecond` (the default budget is far below the limit).
 - One externalized content is capped at 32 MiB; a larger part is marked
   `unavailable` rather than returning unverifiable partial content.
 - **Externalized-content assembly is gated process-wide**: at most 4 `{ref}`
@@ -203,7 +209,7 @@ subscription was lost and could not be rebuilt, so the outcome is
 
 The plugin is also a **web client bundle** (`package.json`'s `dsh.client`,
 exporting `./client`). Through the `settings.section` slot it registers a
-**`cherry-remote`** page (nav label “Cherry Studio Bridge”): device cards
+**`cherry-remote`** page (nav label “Cherry Bridge”): device cards
 (enable/disable/remove), a pairing wizard (paste → 6-digit code → verdict), a
 scope panel, and live state pushed over the same-origin SSE feed
 (`/cherry-remote/api/events`).

package/README.md

@@ -36,13 +36,13 @@ flowchart TB
 
 ```bash
 # 从 npm 安装（推荐，钉住版本）
-npx @deepseek-ai/dsh plugin --profile web add dsh-cherry-bridge@0.1.8
+npx @deepseek-ai/dsh plugin --profile web add dsh-cherry-bridge@0.1.10
 
 # 安装 npm 最新发布版
 npx @deepseek-ai/dsh plugin --profile web add dsh-cherry-bridge
 
 # 从 GitHub 源码安装（与 npm 发布同一提交）
-npx @deepseek-ai/dsh plugin --profile web add github:SAXEM1997/dsh-cherry-bridge#v0.1.8
+npx @deepseek-ai/dsh plugin --profile web add github:SAXEM1997/dsh-cherry-bridge#v0.1.10
 ```
 
 本包自带 `cordis.patch.yml`（`package.json` 的 `dsh.bundle.patch` 指向它），
@@ -84,7 +84,7 @@ cherry_interactions → 看待答的提问/审批卡；cherry_respond 作答
 | --- | --- | --- | --- |
 | `dataDir` | string | `$DSH_HOME/cherry-remote`（回退 `~/.dsh/cherry-remote`） | 插件数据目录，存放 `identity.key` 与 `devices.json`（均 `0600`）。 |
 | `perSessionCacheBytes` | number | `2097152`（2 MiB） | 每个会话的**本地**订阅事件缓存上限，超出按 LRU 丢弃最旧事件并 `warn`。 |
-| `rateBudgetPerSecond` | number | `32` | 每设备平滑请求预算（请求/秒）。桌面端硬上限是 64 req/s 且无协商通道，超限会以 close `1011` 终止会话，故取保守值。 |
+| `rateBudgetPerSecond` | number | `32` | 每设备平滑请求预算（请求/秒）。桌面端共享额度在 ≤2.1.6 是 64 req/s、≥2.1.7 是 256 req/s，且只有新版本才把超限降级为单请求 `RATE_LIMITED`（旧版本直接 close `1011` 终止会话），故按更严格的一代取保守值。内容读取（`agent.content.read`）按 8:1 派生独立一档，且**封顶 512/s**（该额度两代一致）。 |
 
 > ⚠️ `cordis.patch.yml` 的 `config` 是**整体替换**，不与默认值合并。删掉一个键得到的是
 > 插件默认值；键存在但值非法（`0`、负数、字符串）会按名 `warn` 后回落默认值。
@@ -97,7 +97,7 @@ cherry_interactions → 看待答的提问/审批卡；cherry_respond 作答
 | `cherry_agents` | 列出某设备上可用的 agent（分页）。被截断时必带 `truncated`+`truncationReason`（`limit` 或 `safety-bound`，后者是 1000 条内部上限）；停点可命名时还给出 `resumeCursor`——用同一 device 与过滤条件把它回传为 `cursor` 即可从停点继续（此时 `limit` 约束的是下一次回答）；跨工具/设备/过滤条件的令牌会被**本地拒绝**，不会转发给对端。 |
 | `cherry_workspaces` | 列出某 agent 的**已注册**工作区，以及是否存在系统工作区。被截断时必带 `truncated`+`truncationReason`；停点可命名时还给出 `resumeCursor`——用相同过滤条件回传为 `cursor` 即可续读。 |
 | `cherry_sessions` | 列出某 agent 的会话，可按工作区过滤。被截断时必带 `truncated`+`truncationReason`（`limit` 或 `safety-bound`），停点可命名时还给出 `resumeCursor`——用同一 device 与**相同过滤条件**把它回传为 `cursor` 即可从停点继续（此时 `limit` 约束的是下一次回答）；跨工具/设备/过滤条件的令牌会被**本地拒绝**。 |
-| `cherry_read` | 读会话消息历史快照，**从新到旧**返回（`limit` 丢掉的是**最旧**的消息）。附件 part 只回 `name`/`mediaType`/`byteLength`，`content` 为空——字节不取。被截断时 `truncated` + `truncationReason` 一定说明：`limit` 表示你设的条数到了，提高 `limit` 即可；`aggregate-bytes` 表示单次回答达到了**输出大小上限**——与 `limit` 无关，提高 `limit` 只会多付往返、仍停在同一预算。**被挡下的更旧消息用 `messagesResumeCursor` 续读**：把本次返回的 `messagesResumeCursor` 原样传给下一次调用的 `messagesResumeCursor` 参数即可从停点继续（`limit` 此时约束的是**从停点起**本次最多返回多少条），长历史分几次回答读完。停点不可命名时**没有**这个令牌（`truncated` 仍为 `true`，**缺失本身不代表历史已到底**；补救按 `truncationReason` 分：`limit` 可以提高 `limit`，`aggregate-bytes` 不行——被挡住的旧消息在这一回答里拿不到，详见[错误与限制](#错误与限制)）。两种令牌：`cursor:<peer position>`（整窗读完后的 peer 分页位置）与 `after:<messageId>`（停在窗内，续读从最新端扫到该消息、返回它更旧的那些）。不是本工具发出的令牌会被**本地拒绝**（不转发给桌面端——桌面端会静默回退到最新页）；`after:` 锚点在历史里找不到则**报错**，不会返回空结果。 |
+| `cherry_read` | 读会话消息历史快照，**从新到旧**返回（`limit` 丢掉的是**最旧**的消息）。失败消息在桌面端给出时附带 `failure: {reasonCode, retryable, message}`（对端分类、是否值得重试、对端自己的有界摘要；其 `source`/`context` 不投影到回答里）。附件 part 只回 `name`/`mediaType`/`byteLength`，`content` 为空——字节不取。被截断时 `truncated` + `truncationReason` 一定说明：`limit` 表示你设的条数到了，提高 `limit` 即可；`aggregate-bytes` 表示单次回答达到了**输出大小上限**——与 `limit` 无关，提高 `limit` 只会多付往返、仍停在同一预算。**被挡下的更旧消息用 `messagesResumeCursor` 续读**：把本次返回的 `messagesResumeCursor` 原样传给下一次调用的 `messagesResumeCursor` 参数即可从停点继续（`limit` 此时约束的是**从停点起**本次最多返回多少条），长历史分几次回答读完。停点不可命名时**没有**这个令牌（`truncated` 仍为 `true`，**缺失本身不代表历史已到底**；补救按 `truncationReason` 分：`limit` 可以提高 `limit`，`aggregate-bytes` 不行——被挡住的旧消息在这一回答里拿不到，详见[错误与限制](#错误与限制)）。两种令牌：`cursor:<peer position>`（整窗读完后的 peer 分页位置）与 `after:<messageId>`（停在窗内，续读从最新端扫到该消息、返回它更旧的那些）。不是本工具发出的令牌会被**本地拒绝**（不转发给桌面端——桌面端会静默回退到最新页）；`after:` 锚点在历史里找不到则**报错**，不会返回空结果。 |
 | `cherry_events` | 读取该会话缓冲区内比 `cursor` 更新的事件，并在投递后 ack。返回 `{events, nextCursor}`；把 `nextCursor` 作为下次的 `cursor` 传回即可续读。 |
 | `cherry_create` | 在 agent 工作区创建会话（工作区选择**必填**；创建后回读身份）。 |
 | `cherry_send` | 向会话发消息，默认等待运行到终态并返回助手文本；`wait: false` 立即返回命令回执。内容被截断时 `messagesTruncated`/`messagesTruncationReason` 说的是**消息**被截断（丢更旧的）；`deadline` 表示**回答自身的装配时限**到了（见下），已读到的内容照常返回。 |
@@ -119,8 +119,11 @@ cherry_interactions → 看待答的提问/审批卡；cherry_respond 作答
 
 ## 错误与限制
 
-- 读方法在重连后**单次**重放；**命令永不重放**——写操作丢失应答时，携**原 `commandId`**
-  重试即可，桌面端的命令账本会回放原回执。
+- 读方法在重连后**单次**重放；桌面端 ≥2.1.7 的限流拒绝（`RATE_LIMITED` + `retryAfterMs`）
+  也在调用方预算内等待后**单次**重放。**命令永不重放**——写操作丢失应答时，携**原 `commandId`**
+  重试即可，桌面端的命令账本会回放原回执。命令被限流拒绝是另一回事：桌面端在**派发前**就拒绝，
+  命令根本没执行，错误文本会说明这一点、给出对端要求的等待时间（`retryAfterMs`，对端没给就不写），
+  并提示该检查的 `rateBudgetPerSecond`（默认预算远到不了限流）。
 - 单个外置内容上限 32 MiB，超过的 part 标 `unavailable`，不回无法校验的部分内容。
 - **外置内容装配是进程级限流的**：同时最多 4 个 `{ref}` 在装配（官方客户端的读并发策略）。
   单个 ref 的装配峰值约 3× 它的大小，所以**装配峰值**上界是 4×3×32 MiB = 384 MiB，日常远低于此
@@ -159,7 +162,7 @@ cherry_interactions → 看待答的提问/审批卡；cherry_respond 作答
 ## 设置页与启用范围
 
 插件同时是 **web client bundle**（`package.json` 的 `dsh.client`，导出 `./client`），通过
-`settings.section` 槽位注册 **`cherry-remote`**（导航标签「Cherry Studio 互联」）页面：
+`settings.section` 槽位注册 **`cherry-remote`**（导航标签「Cherry 互联」）页面：
 设备卡片（启用/停用/删除）、配对向导（粘贴 → 6 位验证码 → 终态）、启用范围面板，以及经
 `/cherry-remote/api/events` SSE 推送的实时状态。
 

package/cordis.patch.yml

@@ -12,8 +12,9 @@
 # 键值与 spec §8 一致：
 #   - perSessionCacheBytes：每个会话的本地订阅事件缓存上限（字节）。2 MiB。
 #   - rateBudgetPerSecond：每设备二级平滑请求预算（请求/秒），与
-#     `DeviceConnection` 默认值一致，远低于桌面端实现的 64 req/s 硬上限
-#     （该上限无协商通道，超限会以 close 1011 终止整条会话）。
+#     `DeviceConnection` 默认值一致。桌面端共享额度在 ≤2.1.6 是 64 req/s，
+#     ≥2.1.7 是 256 req/s；只有新版本把超限降级为单请求 RATE_LIMITED 拒绝，
+#     旧版本直接以 close 1011 终止整条会话，故按更严格的一代取保守值。
 #   - dataDir 故意不写：留空时装配层按 $DSH_HOME/cherry-remote 解析，
 #     写死会让所有 profile 共用同一份设备档案。
 #

package/dist/index.d.mts

@@ -136,18 +136,21 @@ interface CherryHttpContext {
  * This is the SHARED tier: it covers every business RPC EXCEPT the content-read
  * tier below (see {@link DEFAULT_CONTENT_READ_BUDGET_PER_SECOND}).
  *
- * WHY 32: the desktop's own allowance for "other" RPCs is 64/s
- * (`RemoteAccessService.ts:223-225`; the official client paces them at 40/s,
- * `DesktopSession.ts:242-244`), and exceeding the allowance is not a refusal
- * but a whole-session close (1011). Half the peer's line leaves room for the
- * housekeeping traffic that shares the socket (hello/authenticate/ping/refresh
- * bypass pacing by design) and for arrival compression between our admission
- * times and the peer's arrival times.
+ * WHY 32: the desktop's allowance for "other" RPCs is 64/s through 2.1.6 and
+ * 256/s from 2.1.7 (`RemoteAccessService.ts`; the official client paces them at
+ * 40/s, `DesktopSession.ts:36-37`). What an over-budget request does also
+ * changed: through 2.1.6 it closes the WHOLE session (1011), while 2.1.7 (PR
+ * #21687) refuses that one request (`RATE_LIMITED` + `retryAfterMs`) and keeps
+ * the session. The generations cannot be told apart before one answers, so the
+ * default is sized for the older, harsher peer: 32 is half of 64 and an eighth
+ * of 256, leaving room for the housekeeping traffic that shares the socket
+ * (hello/authenticate/ping/refresh bypass pacing by design) and for arrival
+ * compression between our admission times and the peer's arrival times.
  *
  * SOURCES (neither is in this repository): the desktop lines are
- * `_research/cherry-studio` @ v2.1.6 (90b234e); the official-client lines are
- * CherryHQ/cherry-studio-app @ main 18329fe (app 0.3.0), read on 2026-10-18 —
- * recorded in `.superpowers/review-v013-r1/context.md`.
+ * `_research/cherry-studio` @ v2.1.6 (90b234e) and CherryHQ/cherry-studio @
+ * v2.1.7; the official-client lines are CherryHQ/cherry-studio-app @ v0.4.0,
+ * both read on 2026-10-22 (issue #102).
  */
 export declare const DEFAULT_RATE_BUDGET_PER_SECOND = 32;
 /**
@@ -163,9 +166,10 @@ export declare const DEFAULT_RATE_BUDGET_PER_SECOND = 32;
  * (`RemoteAccessService.ts:223-225`) and the official client issues them at
  * 480/s (`DesktopSession.ts:33-37`). Pacing them with the 32/s shared tier
  * bound a multi-slice walk to ~670 KiB/s and queued it behind unrelated RPCs;
- * 256 is half the peer's own kill line — the same 2× margin the shared tier
- * keeps on 64/s — and lifts the bound to the multi-MiB/s range of the official
- * client while leaving the desktop a session-sized headroom.
+ * 256 is half the 512/s content allowance in BOTH desktop generations (the
+ * shared-tier allowance changed in 2.1.7; this one did not) and lifts the bound
+ * to the multi-MiB/s range of the official client while leaving the desktop a
+ * session-sized headroom.
  *
  * NOT a plugin-config key: the two tiers derive from ONE injected shared budget
  * at the ratio below, so an operator who lowers `rateBudgetPerSecond` lowers
@@ -174,12 +178,18 @@ export declare const DEFAULT_RATE_BUDGET_PER_SECOND = 32;
  *
  * THE MARGINS ARE A DEFAULT-INJECTION CLAIM: the 2× headroom above holds at
  * the default injection (32/s), where this tier is exactly 256/s. An injected
- * budget consumes BOTH tiers' headroom in proportion — at 64/s the shared tier
- * is at the peer's 64/s line and this tier derives to 512/s, i.e. nominal
- * saturation of both allowances. The limiter's integer gap means the ACHIEVED
- * rate stays below the nominal one in that case (≤ 334/s on this tier, 55.6/s
- * on the shared one; see `SecondWindowRateLimiter` in `connection.ts`), which
- * is margin by accident rather than by design.
+ * budget consumes BOTH tiers' headroom up to the point where this tier reaches
+ * the PEER'S CEILING — at 64/s the shared tier sits on the ≤2.1.6 line (the
+ * harsher generation, where overrun still closes the session) and this tier
+ * derives to exactly 512/s, the content allowance BOTH generations share. Above
+ * 64/s the clamp holds it at 512/s: `contentBudgetFor` (issue #111) still
+ * derives the ratio, then bounds it — because a knob copied from 2.1.7's shared
+ * figure (256)
+ * would otherwise derive 2,048/s and spend the tier being refused. The
+ * limiter's integer gap means the ACHIEVED rate stays below the nominal one in
+ * that case (≤ 334/s on this tier, 55.6/s on the shared one; see
+ * `SecondWindowRateLimiter` in `connection.ts`), which is margin by accident
+ * rather than by design.
  */
 export declare const DEFAULT_CONTENT_READ_BUDGET_PER_SECOND = 256;
 /** 2 MiB of event text cached per session (spec §3.5). */
@@ -240,9 +250,13 @@ export interface CherryBridgeConfig {
    * The CONTENT tier is not configurable on its own: it DERIVES from this value
    * at the 8 : 1 production ratio (`DEFAULT_CONTENT_READ_BUDGET_PER_SECOND` /
    * `DEFAULT_RATE_BUDGET_PER_SECOND`), so the default 32 yields the 256/s
-   * content tier and any other value scales both tiers together. Raising this
-   * toward the peer's own lines therefore spends BOTH tiers' safety margin —
-   * see `DEFAULT_CONTENT_READ_BUDGET_PER_SECOND` for the numbers.
+   * content tier and lowering this lowers both. UPWARD the derivation stops at
+   * the peer's 512/s content allowance (issue #111): a value above 64 no longer
+   * scales that tier, it only spends the shared tier's margin. Raising this
+   * toward the peer's own lines spends margin: on a <=2.1.6 desktop that is
+   * BOTH tiers (64 shared / 512 content), while on 2.1.7 only the shared line
+   * moves (256 there). See `DEFAULT_CONTENT_READ_BUDGET_PER_SECOND` for the
+   * numbers.
    */
   rateBudgetPerSecond?: number;
   /**