@whatsmore-nf/dsh-context-steward
DeepSeek Harness 插件：固定上下文容量下的认知资源调度与智能压缩系统（支持 dsh plugin CLI 安装）
插件会安装到这里；不确定时保持 web。
npx -y @deepseek-ai/dsh plugin --profile web add @whatsmore-nf/dsh-context-steward@0.20.0Context Steward 以 @whatsmore-nf/dsh-context-steward 发布，当前版本为 0.20.0。Plugin Hub 会校验它的 manifest，并保存精确安装来源，便于复现安装结果。
English | 中文
A DeepSeek Harness plugin that treats the fixed-capacity context window as a scarce cognitive resource and schedules it: within a hard token budget it protects attention bandwidth for the agent's current key decisions, tiers compression by value density, consolidates completed phases, and keeps a structured fact vault so evicted source text stays recallable.
It ships as a Cordis Service registered at ctx.contextSteward, with a
schemastery-declared static Config and static inject, following the same
plugin shape as the official Harness plugins (e.g. @deepseek-ai/dsh-compaction-basic).
Official dsh-compaction-* compacts the conversation body itself. This plugin
is the complementary layer: it schedules injected memory — deduplicating
repeated observations, tiered compression, phase consolidation, key-decision
bandwidth guarantees, and fact recall — before the compacted snapshot is
injected into the model-visible context. Both can run side by side.
Via Harness plugin CLI (same as other official plugins):
dsh plugin --profile web add @whatsmore-nf/dsh-context-steward@latest
Or directly via npm:
npm install @whatsmore-nf/dsh-context-steward
The plugin needs the Harness runtime (@deepseek-ai/cordis, @deepseek-ai/dsh-agent,
@deepseek-ai/dsh-llm, @deepseek-ai/dsh-session) as peer dependencies; a
Harness profile already provides them.
Add a row to a cordis.yml / cordis.patch.yml bundle:
- id: dsh-context-steward
name: '@whatsmore-nf/dsh-context-steward'
config:
capacity: 8000
enabled: true
inject: true
Loading registers the ctx.contextSteward Service and wires it automatically
to the Harness lifecycle events (see Events).
ContextStewardConfig)Every key is optional; missing keys fall back to the resolved defaults below.
Unknown keys, wrong types, and out-of-range ratios fail plugin load
(resolveConfig rejects them, mirroring official plugins).
| Key | Default | Meaning |
|---|---|---|
capacity | 4000 | Token budget of the injected compressed context (the fixed bandwidth cap). |
reserved | 0 | System tokens excluded from the compressible region. |
decisionGuarantee | floor(capacity × 0.35) | Attention bandwidth floor: max tokens the working set may occupy while a key decision is protected. |
halfLifeMs | 600000 | Time-decay half-life (10 min) for recency weighting. |
adaptiveRecency | true | Adapt the half-life to observed decision cadence. |
minAdaptiveHalfLifeMs | 10000 | Lower bound of the adaptive half-life. |
maxAdaptiveHalfLifeMs | 3600000 | Upper bound of the adaptive half-life. |
adaptiveThresholds | true | Self-tune demote/promote thresholds from churn. |
churnWindowMs | 60000 | Churn observation window for threshold tuning. |
tuneStep | 0.03 | Threshold tuning step. |
demoteThreshold | 0.35 | Score below which working items are demoted to the cold pool. |
promoteThreshold | 0.55 | Score above which cold items are promoted back to the working set. |
workingRenderRatio | 0.5 | Share of the render budget given to the working set. |
maxProtectedDecisions | 4 | Max protected decision snapshots kept verbatim (older ones age out). |
coldCompactScore | 0.4 | Minimum score for cold-pool tiered compression. |
rehydrateThreshold | 0.6 |
import type { Context } from '@deepseek-ai/cordis'
import ContextSteward from '@whatsmore-nf/dsh-context-steward'
export const name = 'context-steward'
export const inject = ['sessions']
export function apply(ctx: Context): void {
const plugin = ctx.plugin(ContextSteward, { capacity: 8000 })
// Optional: plug in an LLM semantic summarizer (async prewarm during pre-step idle time,
// upgrading compression from heuristics to structured summaries)
// plugin.asyncSummarize = async (content, depth) => await llm.complete(
// buildCompactionPrompt({ context: [content] }), { maxTokens: depth >= 3 ? 160 : 80 },
// )
}
The Service is available as ctx.contextSteward; per-session schedulers are
obtained with ctx.contextSteward.scheduler(session).
import { createContextStewardPlugin } from '@whatsmore-nf/dsh-context-steward'
const plugin = createContextStewardPlugin({ capacity: 8000 })
plugin.hooks.onAppend?.({ id: 'u1', kind: 'user', content: '目标是部署服务', timestamp: 0 })
plugin.hooks.onDecision?.({ goal: '部署服务', currentStep: '选型', attentionFocus: ['部署', '服务'] })
const prompt = plugin.hooks.onBeforePrompt?.() // 注入压缩后的上下文
CognitiveResourceScheduler is exported standalone: ingest(), checkpoint(),
setPhase(), consolidatePhase(), compiledContext(), query(),
exportArchive(), exportState() / restoreState() — see the type
declarations for the full surface.
When enabled is true, the plugin subscribes to:
| Event | Purpose |
|---|---|
session/event | Feed model-visible user/assistant/tool events into the scheduler (dedupe, tiered compression, fact extraction, bandwidth accounting). Self-injected compressed context is skipped. |
agent/pre-step | Treat the upcoming step as a key decision: attention re-ranking, then inject the compressed snapshot. |
agent/request-error | On CONTEXT_WINDOW_EXCEEDED failure, feed the overflow as an observation so the next snapshot perceives it. |
agent/disposed | Export the archive, log closing metrics, and release the session state. |
MIT
| Archive score threshold for rehydrating facts back into the working set. |
dedupe | true | Aggregate repeated observations into repeated ×N records. |
consolidate | true | Consolidate a finished phase into one structured summary. |
renderBudget | available capacity | Token cap of the rendered injected context. |
maxItemChars | 6000 | Source-text truncation cap per tool/observation item (full text stays in the vault). |
reclaimPeekLimit | 16 | Candidate peek count for value-density reclamation. |
enabled | true | false registers the Service only, without event wiring. |
inject | true | Enable injecting the compressed snapshot (still gated by injectThresholdRatio). |
injectThresholdRatio | 0.8 | Injection engages only when measured session pressure totalTokens >= floor(window × ratio) — the same trigger semantics as the official compaction's thresholdRatio. Below it the plugin only keeps invisible in-memory bookkeeping, so normal reasoning steps are not disturbed. |
injectContextWindow | 0 | Explicit context-window override (tokens). 0 = resolve from the routed model adapter's contextWindow; when the window cannot be resolved and no override is set, injection is conservatively skipped. |