dsh-work-continuity
Work Continuity for DSH — goal/decisions/checkpoints/next-steps as explicit WorkState, decoupled from generic memory
98 results
Work Continuity for DSH — goal/decisions/checkpoints/next-steps as explicit WorkState, decoupled from generic memory
DeepSeek Harness plugin: learning agentic router — rule + UCB + LinUCB + k-NN experts under an EXP3 meta-selector, with a durable quality-reward flywheel, REAL model switching via the session request-header precedence chain, and STEP-LEVEL routing (v1.5.0)
Dual-model routing policy: premium plans and rescues, local model implements ordinary steps
A configurable DeepSeek Harness host plugin that silently appends route-specific system-prompt guidance for GPT and DeepSeek steps.
DSH Web plugin: collapse consecutive tool-use and thinking steps between messages into a single bracketed, expandable item
Hot-switch agent presets (working modes) inside a running DSH session: mode changes are armed immediately and applied at the next step boundary through agentPresets.recompose, via the /mode slash command.
Fail-closed per-turn step, tool-call, and provider-token budgets for DeepSeek Harness
Plug-and-play proteomics analysis bundle for DeepSeek Harness: auto-managed R 4.4 runtime, step-wise traceable pipeline (normalize / PCA / batch / DEA / enrichment / GSEA), cached GO-KEGG annotation backgrounds.
DSH 插件安检机 — install-time static security auditor for DeepSeek Harness plugin bundles. Zero dependencies, zero install scripts, zero build step.
Pre-step skill router for DeepSeek Harness: rule-first task-to-skill matching that pours matched skill bodies only on high-confidence hits and stays silent otherwise. Companion to skill-bartender.
Token usage dashboard for DeepSeek Harness: cache, context, efficiency, per-step details, and cross-session comparison
DeepSeek Harness 原生插件:让纯文本模型(如 DeepSeek)经 agent/pre-step 劫持 + resolveModelInfo 包装自动识别上传图片(qwen-vl),使纯文本对话也能“看见”图片。
Token usage analytics tab for the dsh web GUI: whole-log projections plus a per-turn/per-step breakdown
Git-backed long-term memory for DeepSeek Harness: a local .mem git repo (single main branch, immutable entries, structured commit messages) with mem_search / mem_read / mem_write / mem_delete / mem_replace tools and fully automatic legacy migration (no CLI needed; the CLI remains for explicit control), auto-managed by the root agent with zero manual steps.
First-run health check for DeepSeek Harness (dsh): verifies Node/pnpm/dsh toolchain, profile, API key, workspace and registry config, then prints actionable next steps. Zero runtime dependencies, read-only. CLI + agent-callable quickstart tool.
dsh-experience-flywheel · DSH 经验飞轮插件 — agent cross-session experience memory + dual-acceptance gates (auto query/inject via agent/pre-step middleware; exit-1 gate scripts)
DSH web plugin: a draggable ring-gauge dashboard pinned to the bottom-right that shows live model output speed (tok/s), with an expandable detail panel (per-step exact speed, TTFT, cache-hit rate, cache read/write, reasoning and input tokens, model, cumulative tokens).
全局提示注入 / 上下文病毒(context virus)防御插件:tools/pre-execute 扫工具参数、tools/post-execute 扫工具结果、agent/pre-step 扫进入模型前的消息并注入金丝雀守卫、llm/stream 出站消毒;支持 block / quarantine / monitor 三模式、磁盘化可演进病毒库(学习/导入/导出)与审计日志。纯 JS 无原生依赖,profile 层挂载,对所有预设与子代理全局生效。
长任务编排引擎。接收任务→模块装配(steps.json 装配表 + minds.json)→两层审计→机械化执行。触发词:长任务、编排、拆解任务、多步骤执行、任务流水线、编排引擎。
Independent-model advisory review for DeepSeek Harness: after each tool step, a reviewer model audits the agent's operations and injects concerns/guidance into the next step.
DSH web plugin: two-step inline archive on workspace sidebar rows — one click arms a red confirm, the second archives; archive is no longer buried in the ⋮ menu. · 工作区侧边栏行内两击归档:点一次变红确认,再点即归档,归档不再藏在 ⋮ 菜单里。
Wide stats footer: removes the max-width clamp on the composer stats line (turns/steps | LLM time | TTFT | cache hit | token counts) so it spans the full composer width, centered
dsh plugin: auto-open the default browser as soon as the dsh web GUI server is listening — one-step startup for `dsh web`
DSH plugin: view OBJ/STL/STEP models in the web UI and translate them into a standardized, LLM-readable JSON