@aiwayds/dsh-dcp 0.10.0 → 0.12.0

package/README.en.md

@@ -3,7 +3,7 @@
 Deterministic context-compaction backend for dsh (DeepSeek Harness): **context
 compaction without an LLM call**, works out of the box.
 
-**Requires dsh >= 0.1.2-rc.1** — this plugin targets the dsh RC/stable line only (CI and releases resolve the newest of the `latest`/`next` dist-tags at runtime). **The alpha line is no longer supported.**
+**Requires dsh >= 0.1.7-rc.1** — this plugin targets the dsh RC/stable line only (CI and releases resolve the newest of the `latest`/`next` dist-tags at runtime). **The alpha line is no longer supported.**
 
 > [简体中文](README.md) · **English**
 
@@ -131,15 +131,25 @@ Removing the package without the reverse step leaves the mount pointing at the v
 
 | Command | Effect |
 |---|---|
-| `/dcp` | status: config, compaction count, tokens saved |
-| `/dcp compact` | compact now (zero LLM) |
+| `/dcp` | compact now (zero LLM); same as `/dcp compact` |
+| `/dcp status` | status: config, compaction count, tokens saved |
+| `/dcp help` / `--help` / `-h` | show the command grammar |
 | `/dcp set <k> <v>` | adjust a knob for this session, with a persist hint |
 
 Settable: `dedup`, `purgeErrors`, `maxItems`, `maxItemChars`,
 `maxSummaryTokens`, `language`, `tokenEstimate`, `thresholdRatio`,
 `roundInterval`, `notice`, `onModelSwitch`, `modelSwitchMinTokens`.
 
-The `/dcp` status also lists every session that has compacted (subagents
+The bare `/dcp` and `/dcp compact` hit the same manual compaction seam:
+the common action takes zero arguments, so nobody has to remember a
+subcommand, while the read-only and tuning verbs stay behind explicit
+arguments (`/dcp status` for the block, `/dcp help` for usage).
+
+> **Behavior change in 0.11.0**: the bare `/dcp` now compacts instead of
+> showing status; the status output moved to `/dcp status`. After
+> upgrading, use `/dcp status` to inspect — the bare command compacts.
+
+The `/dcp status` block also lists every session that has compacted (subagents
 included): `per-session: session-1 (2 compactions, ~444 tokens), child
 (1 compaction, ~22 tokens)`. Compactions count per session; disposed
 sessions (one-shot subagents included) fall out of the overview
@@ -154,11 +164,11 @@ for the rest) so the status stays one line.
 | Overflow recovery | on a provider context-window error | inherited |
 | **Round interval** | every `roundInterval` assistant messages | added by this plugin; one round = one LLM roundtrip (each tool-iteration response counts, so one-shot subagents trigger too). **Default 50**: first compaction after message 50, then every 50 more (100, 150, …); any compaction (pressure included) restarts the clock. Fires at the first idle boundary after the count is reached (below the pressure threshold too). `0` disables; requires the default `auto: true` |
 | **Model switch** | the session's effective provider/model route changes | added by this plugin (`onModelSwitch`). Default `notice`: appends one row suggesting `/dcp compact` first to shadow the old model's history and save tokens; `auto` compacts at the session's next idle boundary; `off` disables. Two gates: switches within 10 assistant messages of the last compaction are ignored (nothing stale to shadow), and switches while the context is below `modelSwitchMinTokens` (default 32768, `0` disables this gate) are ignored (not enough to be worth shadowing); `auto` requires the default `auto: true`, otherwise it degrades to `notice` |
-| Manual | `/dcp compact`, `/compact` | anytime |
+| Manual | `/dcp` (bare), `/dcp compact`, `/compact` | anytime |
 
 - **Subagents are covered**: in-process subagents (including continuable and one-shot children) dispatch through the same events, so pressure/overflow/round/model-switch triggers count and fire per child session independently. The round trigger counts assistant messages, so a one-shot subagent whose whole run is a single turn (many tool iterations) triggers too.
 - **How model-switch detection works**: the per-request `request/context` routing snapshot is folded per session (a provider or model change alone counts as a switch), covering every switch entry point — TUI `/model`, web clients, changed default-model settings. A session's first observed request only seeds the baseline. `notice: false` silences the compaction rows only; the switch row is controlled by `onModelSwitch` independently. The size floor is measured by the host token meter (`measure().surfaceTokens`) and fails open when unavailable.
-- **Visibility**: after every trigger event a one-line notice row (`dcp: compacted N history items (~X tokens, trigger)`) is appended to the session; frontends render it as a collapsed row. Note the row also rides the model request context (~15–25 tokens per compaction), and it is **on by default since 0.4.0** — disable with `notice: false`. `/dcp` stats count every committed region (a pressure retry loop may commit several).
+- **Visibility**: after every trigger event a one-line notice row (`dcp: compacted N history items (~X tokens, trigger)`) is appended to the session; frontends render it as a collapsed row. Note the row also rides the model request context (~15–25 tokens per compaction), and it is **on by default since 0.4.0** — disable with `notice: false`. `/dcp status` stats count every committed region (a pressure retry loop may commit several).
 
 ## Configuration
 
@@ -167,6 +177,7 @@ All optional, defaults work out of the box:
 | Key | Default | Meaning |
 |---|---|---|
 | `thresholdRatio` | 0.8 | pressure trigger (inherited upstream compaction-basic default 0.8; this plugin's bundle patch mounts 0.7, recommended for CJK-heavy sessions) |
+| `headroomTokens` | 65536 | inherited from upstream compaction-basic (new in 0.1.7): extra pressure-budget margin (non-negative integer). The pressure ceiling is context window − routed output reservation (`maxTokens`) − this margin; lower it for small-window models |
 | `roundInterval` | 50 | compact every N assistant messages (one LLM roundtrip) (0 disables). Default 50: 50, 100, 150… — the clock restarts after every compaction |
 | `onModelSwitch` | `notice` | after a model switch: `notice` suggests `/dcp compact` (default); `auto` compacts at the next idle boundary; `off` disables |
 | `modelSwitchMinTokens` | 32768 | minimum priced context size (measured by the host token meter) for a model switch to be announced; `0` disables this gate. 32k default ≈ the post-compaction baseline (~16% of the window) plus some real growth |
@@ -197,7 +208,7 @@ cordis.patch.yml mount block.
 ## Development
 
 ```bash
-npm install && npm test     # 65 tests: extractor/compaction/command/config/triggers/setup
+npm install && npm test     # 109 tests: extractor/compaction/command/config/triggers/recover-seam/setup
 ```
 
 ## License

package/README.md

@@ -2,7 +2,7 @@
 
 dsh（DeepSeek Harness）的确定性压缩后端：**上下文压缩不调 LLM**，开箱即用。
 
-**要求 dsh >= 0.1.2-rc.1** — 本插件只跟随 dsh RC/stable 线（CI 与发版在运行时解析 latest/next 中更新的 dist-tag）。**不再支持 alpha 线。**
+**要求 dsh >= 0.1.7-rc.1** — 本插件只跟随 dsh RC/stable 线（CI 与发版在运行时解析 latest/next 中更新的 dist-tag）。**不再支持 alpha 线。**
 
 > **简体中文** · [English](README.en.md)
 
@@ -110,13 +110,18 @@ npx dsh-dcp-setup --remove /path/to/cordis.patch.yml
 
 | 命令 | 作用 |
 |---|---|
-| `/dcp` | 状态：配置、压缩次数、省下的 token |
-| `/dcp compact` | 立即压缩（零 LLM） |
+| `/dcp` | 立即压缩（零 LLM）；等同于 `/dcp compact` |
+| `/dcp status` | 状态：配置、压缩次数、省下的 token |
+| `/dcp help` / `--help` / `-h` | 显示命令用法 |
 | `/dcp set <k> <v>` | 会话内调参，并提示如何持久化 |
 
 可调键：`dedup`、`purgeErrors`、`maxItems`、`maxItemChars`、`maxSummaryTokens`、`language`、`tokenEstimate`、`thresholdRatio`、`roundInterval`、`notice`、`onModelSwitch`、`modelSwitchMinTokens`。
 
-`/dcp` 状态还会列出每个发生过压缩的会话（per-session 概览，含子代理），例如 `per-session: session-1 (2 compactions, ~444 tokens), child (1 compaction, ~22 tokens)`。压缩按会话独立计数；已销毁的会话（含 one-shot 子代理）自动从概览消失；列表封顶一行（最多前 10 个会话，超出显示 `+N more`）。
+裸 `/dcp` 与 `/dcp compact` 走同一条手动压缩缝——最常用的动作零参数直达，不必记子命令；只读或调参的动作留在显式子命令后面（看状态打 `/dcp status`，用法打 `/dcp help`）。
+
+> **0.11.0 起语义变更**：裸 `/dcp` 由「显示状态」改为「立即压缩」，原来的状态输出移到 `/dcp status`。升级后别再用裸 `/dcp` 查状态——它会直接压一次。
+
+`/dcp status` 还会列出每个发生过压缩的会话（per-session 概览，含子代理），例如 `per-session: session-1 (2 compactions, ~444 tokens), child (1 compaction, ~22 tokens)`。压缩按会话独立计数；已销毁的会话（含 one-shot 子代理）自动从概览消失；列表封顶一行（最多前 10 个会话，超出显示 `+N more`）。
 
 ## 触发条件
 
@@ -126,11 +131,11 @@ npx dsh-dcp-setup --remove /path/to/cordis.patch.yml
 | 溢出恢复 | 模型报 context 超限时 | 继承官方 |
 | **轮数触发** | 会话每收到 `roundInterval` 条 assistant message | 本插件新增；一条 = 一次 LLM 往返（每轮工具迭代各算一条，one-shot 子代理也能触发）。**默认 50**：第 50 条后触发第一次，之后每 50 条一次（100、150……）；任何一次压缩（含压力触发）都会重置轮数时钟。到达条数后的第一个空闲点触发（阈值之下也压）。`0` 关闭；需保持 `auto: true`（默认开） |
 | **模型切换** | 会话实际路由的 provider/model 变化时 | 本插件新增（`onModelSwitch`）。默认 `notice`：追加一行提醒"建议先执行 `/dcp compact` 压缩旧模型历史，节省 token"；`auto` 在该会话下一个空闲点自动压缩；`off` 关闭。两道门控：距上次压缩不足 10 条 assistant message 忽略（没有旧账可甩），上下文不足 `modelSwitchMinTokens`（默认 32768，`0` 关闭该门）忽略（不够甩的量）；`auto` 需保持 `auto: true`（默认开），否则降级为 `notice` |
-| 手动 | `/dcp compact`、`/compact` | 随时可用 |
+| 手动 | `/dcp`（无参数）、`/dcp compact`、`/compact` | 随时可用 |
 
 - **subagent 同样生效**：进程内 subagent（含 continuable 与 one-shot 子代理）走同一套事件分发，压力/溢出/轮数/模型切换对子会话独立计数、独立触发。轮数触发按 assistant message 计数，所以全程只有 1 个 turn 的 one-shot 子代理（多次工具迭代）也能触发。
 - **模型切换检测原理**：折叠每请求的 `request/context` 路由快照（provider 或 model 任一变化即判定切换），覆盖所有切换入口（TUI `/model`、Web 客户端、默认模型设置变更）；会话首个请求只播种基线不告警。`notice: false` 只关压缩通知行，模型切换提醒行由 `onModelSwitch` 独立控制。上下文下限按宿主 tokenMeter 的实测计价（`measure().surfaceTokens`），测不到时 fail-open（只留轮数门控）。
-- **压缩可见性**：每次压缩成功后，会话里追加一行 `dcp: 已压缩 N 条历史（约 X tokens，触发方式）` 通知行（前端渲染为折叠行）。注意该行也会作为上下文随请求发给模型（每次压缩约 15–25 tokens），且 **0.4.0 起默认开启**；`notice: false` 可关闭。`/dcp` 的 stats 持续累计（压力触发的多次 region 提交各计一次）。
+- **压缩可见性**：每次压缩成功后，会话里追加一行 `dcp: 已压缩 N 条历史（约 X tokens，触发方式）` 通知行（前端渲染为折叠行）。注意该行也会作为上下文随请求发给模型（每次压缩约 15–25 tokens），且 **0.4.0 起默认开启**；`notice: false` 可关闭。`/dcp status` 的 stats 持续累计（压力触发的多次 region 提交各计一次）。
 
 ## 配置
 
@@ -139,6 +144,7 @@ npx dsh-dcp-setup --remove /path/to/cordis.patch.yml
 | 键 | 默认 | 说明 |
 |---|---|---|
 | `thresholdRatio` | 0.8 | 压力触发阈值（继承上游 compaction-basic 默认 0.8；本插件 bundle patch 挂载时默认 0.7，中文场景建议 0.7） |
+| `headroomTokens` | 65536 | 继承上游 compaction-basic（0.1.7 新增）：压力预算额外余量（非负整数）。压力上限 = 上下文窗口 − 路由请求预留输出（`maxTokens`）− 本余量；小窗口模型建议下调 |
 | `roundInterval` | 50 | 每 N 条 assistant message（一次 LLM 往返）触发一次压缩（0 关闭）。默认 50：50、100、150……每次压缩后重数 |
 | `onModelSwitch` | `notice` | 模型切换后：`notice` 提醒执行 `/dcp compact`（默认）；`auto` 下一个空闲点自动压缩；`off` 关闭 |
 | `modelSwitchMinTokens` | 32768 | 模型切换提醒/自动压缩的上下文下限（按宿主 tokenMeter 实测计价）：不足则忽略该次切换；`0` 关闭此门。默认 32k ≈ 压缩后基线（~16% 窗口）之上再涨一截才有得甩 |
@@ -164,7 +170,7 @@ npx dsh-dcp-setup --remove /path/to/cordis.patch.yml
 ## 开发
 
 ```bash
-npm install && npm test     # 65 个用例：抽取/压缩/命令/配置/触发/安装脚本
+npm install && npm test     # 109 个用例：抽取/压缩/命令/配置/触发/恢复缝/安装脚本
 ```
 
 ## License

package/icon.svg

@@ -0,0 +1,5 @@
+<svg width="24" height="24" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="1.8" stroke-linecap="round" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg">
+  <path d="M5 5.5h14M8 12h8M5 18.5h14"/>
+  <path d="m10.4 1.8 1.6 1.6 1.6-1.6"/>
+  <path d="m10.4 22.2 1.6-1.6 1.6 1.6"/>
+</svg>

package/lib/command.js

@@ -1,7 +1,11 @@
 /**
- * The `/dcp` slash command: status, manual compaction, and runtime knobs —
- * one entry point, no required arguments (design references opencode-dcp's
- * `/dcp` panel in a dsh-idiomatic, text-only form).
+ * The `/dcp` slash command: manual compaction by default, plus status and
+ * runtime knobs — one entry point, no required arguments (design references
+ * opencode-dcp's `/dcp` panel in a dsh-idiomatic, text-only form).
+ *
+ * The bare command compacts, matching how a user reaches for `/dcp` when the
+ * context feels heavy; the infrequent read-only and tuning verbs stay behind
+ * explicit arguments (`status`, `set`), and `--help` prints the grammar.
  *
  * @module dsh-dcp/command
  */
@@ -10,8 +14,10 @@ import { ManualCompactionError } from '@deepseek-ai/dsh-compaction'
 import { RUNTIME_SETTABLE } from './config.js'
 
 const USAGE = `Usage:
-  /dcp                show status (mode, config, compaction stats)
-  /dcp compact        compact now (deterministic, no LLM call)
+  /dcp                compact now (deterministic, no LLM call)
+  /dcp compact        same as the bare command
+  /dcp status         show status (mode, config, compaction stats)
+  /dcp help           show this help (also --help / -h)
   /dcp set <k> <v>    adjust a knob for this session (dedup, purgeErrors,
                       maxItems, maxItemChars, maxSummaryTokens, language,
                       tokenEstimate, thresholdRatio, roundInterval, notice,
@@ -127,10 +133,10 @@ async function compactNow(ctx, invocation, engine) {
 export async function executeDcp(ctx, invocation, engine, version) {
   const tokens = invocation.rawInput.trim().split(/\s+/).filter((token) => token.length > 0)
   const [subcommand, ...rest] = tokens
-  if (subcommand === undefined) return success(statusText(engine, version))
-  if (subcommand === 'status') return success(statusText(engine, version))
-  if (subcommand === 'help') return success(USAGE)
+  if (subcommand === undefined) return compactNow(ctx, invocation, engine)
   if (subcommand === 'compact') return compactNow(ctx, invocation, engine)
+  if (subcommand === 'status') return success(statusText(engine, version))
+  if (subcommand === 'help' || subcommand === '--help' || subcommand === '-h') return success(USAGE)
   if (subcommand === 'set') {
     if (rest.length < 2) return failure(USAGE)
     const outcome = applySet(engine, rest[0], rest.slice(1).join(' '))
@@ -149,7 +155,8 @@ export async function executeDcp(ctx, invocation, engine, version) {
 export function registerDcpCommand(ctx, engine, version) {
   return ctx.commands.register({
     name: 'dcp',
-    description: 'dsh-dcp: deterministic compaction status and controls',
+    description: 'dsh-dcp: compact now (deterministic, zero LLM) — /dcp --help for status and controls',
+    input: { hint: '[compact|status|help|set <k> <v>]' },
     handler: (invocation) => executeDcp(ctx, invocation, engine, version),
   })
 }

package/lib/config.js

@@ -25,6 +25,7 @@ export const DCP_CONFIG_KEYS = [
 /** compaction-basic policy keys forwarded to the parent engine verbatim. */
 export const BASIC_CONFIG_KEYS = [
   'thresholdRatio',
+  'headroomTokens',
   'retainRatio',
   'retainTokens',
   'summarizationProvider',

package/lib/index.js

@@ -96,6 +96,7 @@ const modelPolicy = z.object({
   provider: z.string().required(),
   model: z.string().required(),
   thresholdRatio: z.number(),
+  headroomTokens: z.number().step(1).min(0),
   retainRatio: z.number(),
   retainTokens: z.number().step(1).min(0),
   summarizationProvider: z.string(),
@@ -111,6 +112,7 @@ export class DcpEngine extends BasicCompactionEngine {
   static Config = z.object({
     // compaction-basic policy keys (forwarded verbatim)
     thresholdRatio: z.number(),
+    headroomTokens: z.number().step(1).min(0),
     retainRatio: z.number(),
     retainTokens: z.number().step(1).min(0),
     summarizationProvider: z.string(),
@@ -371,7 +373,9 @@ export class DcpEngine extends BasicCompactionEngine {
     try {
       session.append('user/message', createUserMessage({
         content: [{ type: 'text', text: summary }],
-        source: { kind: 'plugin', plugin: 'dsh-dcp', form: 'notice', summary },
+        // Producer-owned kind (dsh 0.1.7 removed the 'plugin' catch-all);
+        // declared for checkJs in lib/sources.d.ts.
+        source: { kind: 'dsh-dcp', form: 'notice', summary },
       }))
     } catch (error) {
       const message = error instanceof Error ? error.message : String(error)
@@ -483,7 +487,9 @@ export class DcpEngine extends BasicCompactionEngine {
     try {
       session.append('user/message', createUserMessage({
         content: [{ type: 'text', text: summary }],
-        source: { kind: 'plugin', plugin: 'dsh-dcp', form: 'notice', summary },
+        // Producer-owned kind (dsh 0.1.7 removed the 'plugin' catch-all);
+        // declared for checkJs in lib/sources.d.ts.
+        source: { kind: 'dsh-dcp', form: 'notice', summary },
       }))
     } catch (error) {
       // The compaction already committed; a display-row failure must never

package/lib/skill.js

@@ -39,7 +39,7 @@ const SKILL_RESOURCE_BASE = {
 const SKILL_INVOCATION = { modelInvocable: true, userInvocable: true }
 
 /** Routing description; must stay identical to the SKILL.md frontmatter (asserted in tests). */
-export const SKILL_DESCRIPTION = 'dsh 压缩引擎插件（@aiwayds/dsh-dcp）使用与配置指南。凡涉及上下文压缩、/dcp 命令、压缩调参（阈值/密度/语言/轮数触发/模型切换），或要配置 dcp 时先读本指南：/dcp 状态与 /dcp set 十二个可调键、持久化到 cordis.patch.yml 挂载块 config: 段（dsh-dcp-setup 管理）、ask_user_question 调参向导、五类触发（压力/溢出/轮数/模型切换/手动）、subagent 会话独立计数生效。触发词：dcp、压缩、compaction、上下文超限、摘要、thresholdRatio、roundInterval、onModelSwitch。'
+export const SKILL_DESCRIPTION = 'dsh 压缩引擎插件（@aiwayds/dsh-dcp）使用与配置指南。凡涉及上下文压缩、/dcp 命令、压缩调参（阈值/密度/语言/轮数触发/模型切换），或要配置 dcp 时先读本指南：裸 /dcp 即压缩、/dcp status 看状态与 /dcp set 十二个可调键、持久化到 cordis.patch.yml 挂载块 config: 段（dsh-dcp-setup 管理）、ask_user_question 调参向导、五类触发（压力/溢出/轮数/模型切换/手动）、subagent 会话独立计数生效。触发词：dcp、压缩、compaction、上下文超限、摘要、thresholdRatio、roundInterval、onModelSwitch。'
 
 const SKILL_CANDIDATE = {
   name: SKILL_PROVIDER_NAME,

package/lib/sources.d.ts

@@ -0,0 +1,28 @@
+/**
+ * dsh-dcp's producer-owned message source kinds.
+ *
+ * dsh 0.1.7 removed the shared `kind: 'plugin'` catch-all: MessageSourceMap is
+ * now merge-extensible and every producer declares its own `kind` in its own
+ * module (see `MessageSourceMap` in @deepseek-ai/dsh-llm/message and the
+ * bundled `webhook` producer for the canonical shape). Session-log V4 admission
+ * refuses retired `'plugin'` wrappers, so the declarations below are what lets
+ * dsh-dcp's transcript rows persist.
+ *
+ * Legacy rows: session-format V3→V4 migrates unknown third-party plugin names
+ * to `plugin:<name>` kinds, so pre-migration dsh-dcp notice rows replay as
+ * `plugin:dsh-dcp` — lib/summarizer.js skips both spellings.
+ */
+declare module '@deepseek-ai/dsh-llm' {
+  interface MessageSourceMap {
+    /** dsh-dcp transcript row: a one-line compaction/model-switch account. */
+    'dsh-dcp': {
+      readonly kind: 'dsh-dcp'
+      readonly form: 'notice'
+      /** The bounded one-line account, via `boundContextSummary`. */
+      readonly summary: string
+    }
+  }
+}
+
+/** Module marker: keeps the block above an augmentation (merging), not an ambient module declaration (shadowing). */
+export {}

package/lib/summarizer.js

@@ -16,6 +16,8 @@
  * @module dsh-dcp/summarizer
  */
 
+import { isCompactCheckpointSource } from '@deepseek-ai/dsh-compaction'
+
 /** Section headers shared with compaction-basic's checkpoint instruction. */
 export const SECTIONS = Object.freeze({
   intent: 'Primary Request and Intent',
@@ -101,11 +103,13 @@ export function estimateTextTokens(text, mode = 'cjk') {
 }
 
 /**
- * Token estimate for one message: text blocks, tool-call arguments, and
- * tool-result payloads (the same content the host meter prices, minus its
- * per-block overhead).
+ * Token estimate for one message: text blocks and tool-call arguments. Tool
+ * results are first-class role-`'tool'` messages since dsh 0.1.7 (the old
+ * user-role `'tool-result'` block is gone), so their text blocks price through
+ * the same `'text'` branch — the same content the host meter prices, minus its
+ * per-block overhead.
  *
- * @param {{ content?: Array<Record<string, any>> }} message - any message-shaped object.
+ * @param {{ role?: string, content?: Array<Record<string, any>> }} message - any message-shaped object.
  * @param {'cjk'|'ascii'} [mode] - pricing mode, see {@link estimateTextTokens}.
  * @returns {number} estimated tokens.
  */
@@ -114,11 +118,6 @@ export function estimateMessageTokens(message, mode = 'cjk') {
   for (const block of message?.content ?? []) {
     if (block.type === 'text') parts.push(block.text)
     else if (block.type === 'tool-call') parts.push(block.arguments)
-    else if (block.type === 'tool-result') {
-      for (const inner of block.content ?? []) {
-        if (inner.type === 'text') parts.push(inner.text)
-      }
-    }
   }
   return estimateTextTokens(parts.join('\n'), mode)
 }
@@ -191,15 +190,19 @@ function textOf(message) {
 
 /**
  * Re-injected host context that appears as user-role messages every turn:
- * system-prompt snapshots, skill catalogs, and AGENTS.md-style instructions.
- * The host re-sends these on every request, so copying them into a checkpoint
- * is pure waste — the model never loses them. (A prior compaction checkpoint
- * is handled separately, and plugin notice/relay/recall forms stay.)
+ * skill catalogs, AGENTS.md-style instructions, and any producer's `snapshot`-
+ * form state (runtime context, time, tmux — dsh 0.1.7 has no shared `plugin`
+ * kind, so the snapshot form is matched generically). The host re-sends these
+ * on every request and a later snapshot supersedes an earlier one, so copying
+ * them into a checkpoint is pure waste — the model never loses them. (A prior
+ * compaction checkpoint is handled separately, and notice/relay/recall forms
+ * stay.)
  */
 function isInjectedContext(message) {
-  const kind = message.source?.kind
+  const source = message.source
+  const kind = source?.kind
   if (kind === 'skill-catalog' || kind === 'agent-instructions') return true
-  return kind === 'plugin' && message.source.form === 'snapshot'
+  return source?.form === 'snapshot'
 }
 
 /**
@@ -215,21 +218,17 @@ function noticeSummary(message) {
   return undefined
 }
 
-/** Tool-result blocks flattened as text, first non-empty line kept per block. */
-function resultFirstLines(blocks) {
-  const lines = []
-  for (const block of blocks) {
-    if (block.type !== 'tool-result') continue
-    const text = block.content
-      .filter((inner) => inner.type === 'text')
-      .map((inner) => inner.text)
-      .join('\n')
+/** First non-empty text line of one tool-result message, or undefined. */
+function resultFirstLine(message) {
+  for (const block of message.content) {
+    if (block.type !== 'text') continue
+    const line = block.text
       .split('\n')
-      .map((line) => line.trim())
-      .find((line) => line.length > 0)
-    if (text !== undefined) lines.push({ isError: block.isError === true, line: text })
+      .map((part) => part.trim())
+      .find((part) => part.length > 0)
+    if (line !== undefined) return { isError: message.isError === true, line }
   }
-  return lines
+  return undefined
 }
 
 /**
@@ -376,27 +375,33 @@ export function extractFacts(messages, language = 'en') {
       continue
     }
 
-    // user-role messages: tool results, prior checkpoints, real user text
-    const results = resultFirstLines(message.content)
-    if (results.length > 0) {
-      for (const result of results) {
-        const isError = result.isError || errorRegexes.some((pattern) => pattern.test(result.line))
-        if (!isError) continue
-        const call = message.source?.kind === 'tool' ? callsById.get(message.source.callId) : undefined
-        const who = call === undefined ? 'tool' : call.name
-        facts.errors.push(`${who}: ${result.line}`)
-      }
+    // Tool results are first-class role-'tool' messages since dsh 0.1.7
+    // (required toolCallId, message-level isError): only their error lines
+    // feed the checkpoint.
+    if (message.role === 'tool') {
+      const result = resultFirstLine(message)
+      if (result === undefined) continue
+      const isError = result.isError || errorRegexes.some((pattern) => pattern.test(result.line))
+      if (!isError) continue
+      const call = callsById.get(message.toolCallId ?? message.source?.callId)
+      const who = call === undefined ? 'tool' : call.name
+      facts.errors.push(`${who}: ${result.line}`)
       continue
     }
+
+    // user-role messages: prior checkpoints, real user text, display rows
     const text = textOf(message)
     if (text.trim().length === 0) continue
-    if (message.source?.kind === 'plugin' && message.source.plugin === 'compact') {
+    if (isCompactCheckpointSource(message.source)) {
       facts.carried = parseCheckpointSections(text)
       continue
     }
     // dsh-dcp's own compaction notices are transcript display rows: the model
     // learns nothing from them the checkpoint below does not already carry.
-    if (message.source?.kind === 'plugin' && message.source.plugin === 'dsh-dcp') continue
+    // `plugin:dsh-dcp` is the V3→V4 session-log migration's spelling for
+    // pre-migration rows (cast: no type declares the legacy namespace).
+    const noticeKind = /** @type {string} */ (message.source?.kind)
+    if (noticeKind === 'dsh-dcp' || noticeKind === 'plugin:dsh-dcp') continue
     if (isInjectedContext(message)) continue
     const intentText = noticeSummary(message) ?? text
     facts.lastUserText = intentText

package/locale/en.json

@@ -0,0 +1,6 @@
+{
+  "meta": {
+    "title": "dsh-dcp",
+    "description": "Deterministic zero-LLM context-pruning compaction backend with a /dcp command — works out of the box."
+  }
+}

package/locale/zh.json

@@ -0,0 +1,6 @@
+{
+  "meta": {
+    "title": "dsh-dcp",
+    "description": "确定性、零 LLM 摘要的上下文修剪压缩后端，附 /dcp 命令，开箱即用。"
+  }
+}

package/package.json

@@ -1,9 +1,10 @@
 {
   "name": "@aiwayds/dsh-dcp",
-  "version": "0.10.0",
+  "version": "0.12.0",
   "description": "Deterministic context-pruning compaction backend for dsh (DeepSeek Harness) — zero-LLM summaries, /dcp command, works out of the box. Design references Opencode-DCP/opencode-dynamic-context-pruning.",
   "type": "module",
   "main": "lib/index.js",
+  "icon": "./icon.svg",
   "bin": {
     "dsh-dcp-setup": "scripts/setup.mjs"
   },
@@ -12,7 +13,8 @@
     "./summarizer": "./lib/summarizer.js",
     "./config": "./lib/config.js",
     "./setup": "./lib/setup.js",
-    "./package.json": "./package.json"
+    "./package.json": "./package.json",
+    "./locale/*.json": "./locale/*.json"
   },
   "files": [
     "lib",
@@ -21,7 +23,9 @@
     "README.md",
     "README.en.md",
     "LICENSE",
-    "cordis.patch.yml"
+    "cordis.patch.yml",
+    "icon.svg",
+    "locale/*.json"
   ],
   "scripts": {
     "test": "node --test",
@@ -53,17 +57,17 @@
   "dependencies": {},
   "peerDependencies": {
     "@deepseek-ai/cordis": "^4.0.2",
-    "@deepseek-ai/dsh-agent": ">=0.1.2-rc.1",
-    "@deepseek-ai/dsh-brand": ">=0.1.2-rc.1",
-    "@deepseek-ai/dsh-commands": ">=0.1.2-rc.1",
-    "@deepseek-ai/dsh-compaction": ">=0.1.2-rc.1",
-    "@deepseek-ai/dsh-compaction-basic": ">=0.1.2-rc.1",
-    "@deepseek-ai/dsh-compaction-tool-result-pruner": ">=0.1.2-rc.1",
-    "@deepseek-ai/dsh-invariants": ">=0.1.2-rc.1",
-    "@deepseek-ai/dsh-llm": ">=0.1.2-rc.1",
-    "@deepseek-ai/dsh-session": ">=0.1.2-rc.1",
-    "@deepseek-ai/dsh-skill": ">=0.1.2-rc.1",
-    "@deepseek-ai/dsh-token-meter": ">=0.1.2-rc.1",
+    "@deepseek-ai/dsh-agent": ">=0.1.7-rc.1",
+    "@deepseek-ai/dsh-brand": ">=0.1.7-rc.1",
+    "@deepseek-ai/dsh-commands": ">=0.1.7-rc.1",
+    "@deepseek-ai/dsh-compaction": ">=0.1.7-rc.1",
+    "@deepseek-ai/dsh-compaction-basic": ">=0.1.7-rc.1",
+    "@deepseek-ai/dsh-compaction-tool-result-pruner": ">=0.1.7-rc.1",
+    "@deepseek-ai/dsh-invariants": ">=0.1.7-rc.1",
+    "@deepseek-ai/dsh-llm": ">=0.1.7-rc.1",
+    "@deepseek-ai/dsh-session": ">=0.1.7-rc.1",
+    "@deepseek-ai/dsh-skill": ">=0.1.7-rc.1",
+    "@deepseek-ai/dsh-token-meter": ">=0.1.7-rc.1",
     "@deepseek-ai/schemastery": "^3.18.2"
   },
   "peerDependenciesMeta": {
@@ -109,35 +113,35 @@
   },
   "devDependencies": {
     "@types/node": "^24.0.0",
-    "@deepseek-ai/cordis": "4.0.2",
-    "@deepseek-ai/dsh-agent": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-brand": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-commands": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-compaction": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-compaction-basic": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-compaction-tool-result-pruner": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-invariants": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-llm": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-session": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-token-meter": "0.1.2-rc.1",
+    "@deepseek-ai/cordis": "4.0.4",
+    "@deepseek-ai/dsh-agent": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-brand": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-commands": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-compaction": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-compaction-basic": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-compaction-tool-result-pruner": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-invariants": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-llm": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-session": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-token-meter": "0.1.7-rc.1",
     "@deepseek-ai/schemastery": "3.18.2",
     "typescript": "~5.9.0"
   },
   "overrides": {
-    "@deepseek-ai/dsh-agent": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-brand": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-commands": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-compaction": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-compaction-tool-result-pruner": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-invariants": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-llm": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-session": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-token-meter": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-scope": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-attachment": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-session-projection": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-system-prompt": "0.1.2-rc.1",
-    "@deepseek-ai/dsh-llm-retry": "0.1.2-rc.1"
+    "@deepseek-ai/dsh-agent": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-brand": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-commands": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-compaction": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-compaction-tool-result-pruner": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-invariants": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-llm": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-session": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-token-meter": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-scope": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-attachment": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-session-projection": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-system-prompt": "0.1.7-rc.1",
+    "@deepseek-ai/dsh-llm-retry": "0.1.7-rc.1"
   },
   "dsh": {
     "bundle": {
@@ -146,11 +150,11 @@
   },
   "pnpm": {
     "overrides": {
-      "@deepseek-ai/dsh-scope": "0.1.2-rc.1",
-      "@deepseek-ai/dsh-attachment": "0.1.2-rc.1",
-      "@deepseek-ai/dsh-session-projection": "0.1.2-rc.1",
-      "@deepseek-ai/dsh-system-prompt": "0.1.2-rc.1",
-      "@deepseek-ai/dsh-llm-retry": "0.1.2-rc.1"
+      "@deepseek-ai/dsh-scope": "0.1.7-rc.1",
+      "@deepseek-ai/dsh-attachment": "0.1.7-rc.1",
+      "@deepseek-ai/dsh-session-projection": "0.1.7-rc.1",
+      "@deepseek-ai/dsh-system-prompt": "0.1.7-rc.1",
+      "@deepseek-ai/dsh-llm-retry": "0.1.7-rc.1"
     }
   }
 }

package/scripts/compare.mjs

@@ -79,9 +79,6 @@ function hostTokens(content) {
   return content.reduce((total, block) => {
     if (block.type === 'text') return total + Math.ceil(block.text.length / 4)
     if (block.type === 'tool-call') return total + Math.ceil(block.arguments.length / 4)
-    if (block.type === 'tool-result') {
-      return total + block.content.reduce((sum, inner) => sum + (inner.type === 'text' ? Math.ceil(inner.text.length / 4) : 0), 0)
-    }
     return total
   }, 0)
 }
@@ -145,7 +142,7 @@ function checkpointMessage(summaryText, compactionId) {
       { type: 'text', text: summaryText },
       { type: 'text', text: '</compacted-summary>' },
     ],
-    source: { kind: 'plugin', plugin: 'compact', compactionId },
+    source: { kind: 'compact-checkpoint', compactionId },
   }
 }
 

package/scripts/setup.mjs

@@ -124,4 +124,4 @@ if (existed) {
 fs.mkdirSync(path.dirname(target), { recursive: true })
 fs.appendFileSync(target, plan.block)
 console.log(`${plan.action === 'create' ? 'created' : 'patched'} ${target}`)
-console.log('restart dsh, then run /dcp to verify.')
+console.log('restart dsh, then run /dcp status to verify.')

package/skills/dsh-dcp-config/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: dsh-dcp-config
-description: "dsh 压缩引擎插件（@aiwayds/dsh-dcp）使用与配置指南。凡涉及上下文压缩、/dcp 命令、压缩调参（阈值/密度/语言/轮数触发/模型切换），或要配置 dcp 时先读本指南：/dcp 状态与 /dcp set 十二个可调键、持久化到 cordis.patch.yml 挂载块 config: 段（dsh-dcp-setup 管理）、ask_user_question 调参向导、五类触发（压力/溢出/轮数/模型切换/手动）、subagent 会话独立计数生效。触发词：dcp、压缩、compaction、上下文超限、摘要、thresholdRatio、roundInterval、onModelSwitch。"
+description: "dsh 压缩引擎插件（@aiwayds/dsh-dcp）使用与配置指南。凡涉及上下文压缩、/dcp 命令、压缩调参（阈值/密度/语言/轮数触发/模型切换），或要配置 dcp 时先读本指南：裸 /dcp 即压缩、/dcp status 看状态与 /dcp set 十二个可调键、持久化到 cordis.patch.yml 挂载块 config: 段（dsh-dcp-setup 管理）、ask_user_question 调参向导、五类触发（压力/溢出/轮数/模型切换/手动）、subagent 会话独立计数生效。触发词：dcp、压缩、compaction、上下文超限、摘要、thresholdRatio、roundInterval、onModelSwitch。"
 ---
 
 # dsh-dcp 使用指南（确定性上下文压缩）
@@ -80,7 +80,7 @@ dsh-dcp 自有键（除 `thresholdRatio` 外全部可用 `/dcp set` 调）：
 | 溢出 | 模型报 context 超限 | 继承官方恢复流程 |
 | 轮数 | 每累计 `roundInterval` 条 assistant message | 任何一次压缩（含压力/手动）都重置时钟；`0` 关闭；需保持 `auto: true`（默认开） |
 | 模型切换 | 会话实际路由的 provider/model 变化 | `onModelSwitch` 控制（默认 `notice` 提醒）；`auto` 在下一个空闲点自动压缩；两道门：距上次压缩 <10 条消息、上下文 <`modelSwitchMinTokens` |
-| 手动 | `/dcp compact`、`/compact` | 随时可用 |
+| 手动 | `/dcp`（无参数）、`/dcp compact`、`/compact` | 随时可用 |
 
 - **subagent 同样生效**：进程内子代理（含 continuable 与 one-shot）走同一套事件分发，
   压力/溢出/轮数/模型切换对每个会话独立计数、独立触发。
@@ -89,8 +89,9 @@ dsh-dcp 自有键（除 `thresholdRatio` 外全部可用 `/dcp set` 调）：
 
 ## 排障
 
-1. `/dcp`（无参数）看状态：当前配置、压缩次数、省下的 LLM 调用，以及
+1. `/dcp status` 看状态：当前配置、压缩次数、省下的 LLM 调用，以及
    per-session 概览（含子代理；已销毁的会话自动消失，列表最多前 10 个，超出显示 `+N more`）。
+   手动压缩打裸 `/dcp`（或 `/dcp compact`）；命令用法打 `/dcp help`。
 2. 会话压不动 → 先确认 `auto` 是否为 `true`（自动触发总开关，默认开），再看
    `thresholdRatio` 是否设得过高、`roundInterval` 是否为 `0`。
 3. `npx dsh-dcp-setup --remove` 后出现 `WARN: a compaction-basic entry remains...` →