PerryLink/dsh-fast6

dsh-fast

Read-only performance diagnostics for DeepSeek Harness: session load (open/restore) timing, spill-hit counts, compaction count and trigger, context-injection volume (AGENTS.md/skills/tool-schema token share), and LLM cache hit rate — surfaced via the /fast command and the fast_report tool, persisted as reconstructable fast/* session events with async sampling off the model path.

AI Analysis

适合需要监控会话加载、缓存命中率等性能指标的开发者。

Package
dsh-fast
Version
0.2.12
License
Apache-2.0
Last updated
Sep 12, 2026

Install

$npx -p @deepseek-ai/dsh dsh plugin --profile web add github:PerryLink/dsh-fast

Configuration

All tunables are Schemastery Config fields; invalid values fail the profile load loudly.

KeyDefaultDescription
enabledtrueMaster switch; false mounts nothing.
privacy.includeCwdfalseInclude the sanitized session working directory in reports.
sampling.snapshotIntervalMs60000How often active sessions are sampled (ms).
sampling.maxHistorySamples20Samples retained per session in the durable history.
thresholds.systemPromptTokens20000Warn when the system prompt exceeds this many tokens.
thresholds.toolSchemaTokens8000Warn when the tool schema exceeds this many tokens.
thresholds.surfaceTokens60000Warn when the conversation surface exceeds this many tokens.
thresholds.cacheHitRateFloor0.1Warn when the cache hit rate falls below this (0..1).
thresholds.compactionCountWarn10Warn once this many compactions have triggered.
thresholds.compactionShadowTokens40000Warn when the average shadowed token count per summary exceeds this.
spill.detectSpilledResultstrueDetect spilled tool results from the durable notice marker.