dsh-mnemon
Composable three-tier memory control plane for DeepSeek Harness: persistent runtime context, searchable project documents, pluggable long-term memory, guarded strategies, WebUI, and headless tools.
52 results
Composable three-tier memory control plane for DeepSeek Harness: persistent runtime context, searchable project documents, pluggable long-term memory, guarded strategies, WebUI, and headless tools.
Reflect2Evolve memory plugin: layered L1/L2/L3 memory, reflection-weighted value backprop, cross-task policy induction, skill crystallization, and three-tier retrieval for OpenClaw, Hermes Agent, and DeepSeek Harness.
Claude Science-style research workbench for DeepSeek Harness: ReAct research-loop engine (research_* tools), versioned artifacts with provenance (artifact_* tools), SSH remote-compute engine for long bioinformatics jobs on workstations/HPC (remote_* tools), and 11 science skills for genomics / pathogens / bioinformatics. Ships the tiered cross-provider model router via the companion dsh-model-tier bundle.
OpenCode Zen free-tier models for DeepSeek Harness: zero-config public-key provider with quota-aware pacing.
DSH ad-suppression plugin: writes dsh-ads's own settings key to all-false, broadcasts its retire event, scrubs its portal from the DOM, and short-circuits its dynamic-tier fetch. Four layers, all client-side, all removable from a settings page of its own.
Self-evolving memory + skill lifecycle for DeepSeek Harness. Cross-session memory with zero-token deterministic recall (bigram-Jaccard fused with FTS5 BM25 via RRF), a tiered approval gate, and reinforcement that strengthens what you repeat. Procedural knowledge crystallizes into SKILL.md files that refine in place and are curated through an active-stale-archived lifecycle (reversible archive, pre-op backups, rollback, never deletes). Includes background per-turn review, anti-bloat convergence for both skills and memory, an auto-grown user profile, and a web settings page.
Model router & cost optimizer for DeepSeek Harness: heuristic tier routing between cheap and strong models, automatic fallback on transient failures, and a per-session token/cache/cost panel in the composer dock.
Free vision plugin for DeepSeek Harness (dsh): image understanding for text-only models with free-tier providers (Qwen3-VL-Flash / DeepSeek-OCR / Doubao). 免费视觉插件:纯文本模型看图能力,优先免费模型(通义千问 / 硅基流动 / 豆包)。
OpenCode Go quota ring for DSH Web: shows 5-hour / weekly / monthly usage as a click-to-cycle progress ring left of the model selector, with tiered in-prompt quota warnings (Codex-CLI style).
dsh-skillport — every skill you already have (Claude Code, Codex, Cursor, Gemini CLI) works in DSH: Agent Skills SKILL.md discovery across .claude/.agents/.dsh/.gemini + extraPaths, Tier-2 conversions (cursor rules, claude commands, context files), find_skill search, and a skills doctor
佩丽卡终端 (Perlica Terminal):明日方舟终末地佩丽卡主题的分级任务提示音插件。计划出方案、任务完成、需要你回应、出错时播放不同提示音,普通问答静音。Perlica-themed tiered sound notifications for DeepSeek Harness: plan ready, task done, needs your input, error tones.
Hook DeepSeek Harness into your Hermes pipeline: dispatch-spec template, model-tier routing, quality gates, git single-writer rule — as an installable skill pack.
L1-L2-L3 three-tier memory agent harness: gate approval, zero-judgement event log, weekly distillation, TDD execution loop. Claude Code plugin and DeepSeek Harness (dsh) bundle.
Busy-time input queue for the DeepSeek Harness web GUI: three-tier (now/next/later) planning, queue editing with drag-to-reorder and concurrency protection, interject (steer) and interrupt, batch clear with confirm, session freeze - a cordis client plugin, no dsh source changes, no PR required
dsh-polyglot — the model switch for DSH: one generic OpenAI-compatible ctx.llm adapter, curated free/cheap DeepSeek provider presets, and automatic provider fallback when a free tier rate-limits you
Manage OpenCode Zen free-tier and Go-tier models in settings.yaml: live model listings from opencode.ai, configured-versus-live drift per route, add/remove through four agent tools plus an OpenCode Models settings section.
An elevated Plugin list for DeepSeek Harness Web Settings — trust tiers, collapsible groups, and complete per-plugin detail
Two-tier model router for DeepSeek Harness — automatic execution/judgment routing with LLM Judge, multi-model fallback chains, and exponential-backoff runtime failover. A DSH adaptation of pi-shift-router.
A project-agnostic multi-agent collaboration protocol for the DeepSeek Harness: loop guards that stop runaway agent loops, a six-field handoff format, risk-tiered review routing, and a fixed verify-commit-push-review-close sequence.
Unified ModelID routing for DeepSeek Harness: one logical model id over multiple providers, first-token failover with cooldown, health-aware ranking, fallback reasoning efforts, purpose-based simple/complex task split, per-tier reasoning effort, and a settings-page management panel.
Model capability radar for the DeepSeek Harness web GUI: a Settings tab that charts codexradar crowd-benchmark scores per model tier — task-composition bars, 7-day IQ trend, efficiency badges
A middle permission tier for DeepSeek Harness between workspace-write and danger-full-access: auto-approves harmless commands and operations targeting configured trusted areas, beyond the current workspace.
DSH Web client plugin: chat content width tiers (standard / medium / wide / ultra / full) with a floating picker button. Frees wide tables from the app's hard-coded 748px content cap. UI copy follows the app language (zh / en).
Automatic model routing for DeepSeek Harness: a virtual `smart` model classifies every request by difficulty (hard / normal / easy) and by vision need, then delegates to the models you already configured in Settings → Models. 内置智能路由:三级难度 + 视觉自动路由,模型从「设置-模型」已配置模型中选取,虚拟 smart 模型零配置接入。