@max-null/dsh-habit

本插件属于 @max-null/* 插件系列——这一系列共同构成 SSID（思灵 · Seek Soul in Darkness） 桌面体验。SSID 是整合它们的盒：dsh-capture · dsh-chat-rail · dsh-chinese-thinking · dsh-draft-polish · dsh-guardian · dsh-habit · dsh-memory · dsh-node-appearance · dsh-plugin-center · dsh-quick-toolbar · dsh-skill-mcp-center · dsh-ssid-panels · dsh-ssid-zh-ui · dsh-achievements。

This plugin belongs to the @max-null/* family — a set of plugins that together form the SSID (思灵 · Seek Soul in Darkness) desktop experience.

Self-learning habit engine for the DeepSeek Harness — observes user-correction signals from session events, judges habits with a low-cost model on threshold, and settles candidates behind a two-level human gate. No new agent role: the judgment is an event-driven plugin, immune to context decay.

The loop

① observe   session/event → correction-signal detection (deterministic, zero-token)
② judge     >=3 signals in one session → one flash call (evidence slices + existing habits)
③ settle    candidate zone → user confirms → dsh-memory remember() (suggested)
            → user confirms again → auto → recall injection

截图

本插件是行为提示类：不新增任何按钮、面板或设置项，也不向会话注入任何内容。

行为效果是后台沉淀，全程静默：同一会话内累积到 3 条命中固定纠正短语（如「再检查一下」「重新做」）的短消息后， 插件在回合结束时做一次低成本模型判断；只有判定为「稳定习惯」才落一条候选（状态 pending，存于 $DSH_HOME/storages/habit）。 候选由宿主 UI 读管——SSiD 侧是侧栏的「习惯」面板，确认后写入 dsh-memory（suggested），再由人放行转 auto。 未达阈值、判断失败或消息不匹配时，插件不做任何可见动作，会话里看不到任何痕迹。

按《SSiD 开发手册》§9 截图规范：截图须回答「装完会多出/变成什么」的入口与面板。 本插件无界面元素（no UI surface），故不适用该项要求，改以上述行为效果说明代替。

Compose

- id: habit
  name: '@max-null/dsh-habit'

Requires storage and llm in the host composition (dsh-base ships both). Installs as a bundle: dsh plugin --profile <name> add @max-null/dsh-habit.

Service

• ctx.habit — the engine:
  • snapshot() → candidates (newest first)
  • confirm(id) / discard(id) → first-level human gate
  • (the second gate is dsh-memory's own suggested→auto confirmation)

Config

Field	Default	Meaning
signalThreshold	3	Correction signals before one judgment call
provider	deepseek-official	Judgment model provider
model	deepseek-v4-flash	Judgment model (cheap, deterministic)
storageRoot	$DSH_HOME/storages/habit	JSON storage root

Design notes

• Deterministic observation, LLM on demand: correction detection is a fixed phrase list + length cap (task descriptions are not corrections); the LLM only runs when a session accumulates enough signals.
• Two-level human gate: candidates must be confirmed in the UI AND then pass dsh-memory's own suggested→auto gate. The model can never promote its own habits.
• Narrow input for quality: the judgment call gets at most 5 evidence texts plus the existing habit list — judgment quality comes from precise context, not volume.

Dependencies

@deepseek-ai/cordis (^4.0.1) plus four kernel packages declared as explicit ranges: @deepseek-ai/dsh-llm and @deepseek-ai/dsh-session as >=0.1.7-rc.2 <0.3.0, @deepseek-ai/dsh-storage and @deepseek-ai/dsh-storage-json as >=0.1.1-rc.1 <0.3.0.

Why a range rather than a caret. A caret's upper bound is the next minor: ^0.1.7-rc.2 expands to >=0.1.7-rc.2 <0.2.0-0, which a kernel one minor ahead fails. An unsatisfied dsh-* peer is not a warning — the plugin is skipped instead of loaded: it never enters the fiber graph and never appears in the did not activate list, so the only symptom is a missing set of features. The explicit range admits both the 0.1 and the 0.2 kernel lines.

The check is app-boot's plugin-compatibility.ts: it walks peerDependencies entries whose name is @deepseek-ai/dsh or starts with @deepseek-ai/dsh-, testing semver.satisfies(runtimeVersion, requirement, { includePrerelease: true }) — which is why @deepseek-ai/cordis and the other non-dsh-* peers sit outside the check. Measured with semver 7.7.4 against runtime 0.2.0-rc.1: ^0.1.7-rc.2 and ^0.1.1-rc.1 are false; >=0.1.7-rc.2 <0.3.0 and >=0.1.1-rc.1 <0.3.0 are true.

Develop

npm install --legacy-peer-deps
npm test
npm run typecheck
npm run build

SSID 系列

本插件是 SSID（思灵 · Seek Soul in Darkness） 全家桶的一员；也可以单独安装到任意 DSH profile——插件自身不依赖其它同系列插件，候选的读管界面由宿主提供（SSiD 侧为 dsh-ssid-panels 的「习惯」面板）。