@maiziman/dsh-model-capabilities
Automatic reasoning and image capability detection for custom DeepSeek Harness models
225 results
Automatic reasoning and image capability detection for custom DeepSeek Harness models
DSH session model selector with search and reasoning-effort pane: shadows the native conversation.input.model seat (independent cordis client plugin).
A minimal reasoning-effort ruler for the DSH composer: one hairline, a sliding marker, per-model memory, optimistic switching — and a streamlined model picker.
DSH 思考增强插件(EAC 定制版):思考与回复强制中文(system-prompt 注入)+ 界面硬编码英文中文化。已移除全部接管对话渲染器的显示功能,不再与折叠类插件(dsh-auto-collapse / dsh-turn-fold)抢占 assistant-step 座位。DSH web plugin (EAC fork): force Chinese agent thinking/replies + localize hard-coded UI labels, with all renderer-hijacking display features removed.
Real-time token usage dashboard for DeepSeek Harness — live totals (input/output/cache/reasoning), filters by model, day, month and project, and a smooth trend chart. · 实时 Token 用量仪表盘:总量/输入/输出/缓存命中按模型、按天、按月、按项目随选,附趋势曲线。
Codex-style /reasoning command for DeepSeek Harness Web: select the reasoning effort of the current model from the slash menu
Auto-adjust model reasoning effort per task: off/low/high/max with task classification, peak-pricing-period capping, countdown notices, and in-turn error escalation
Billing season and account balance chip for the DeepSeek Harness web GUI composer: a native-looking pill after the cache-hit stat, with a panel on click. DeepSeek Harness 计费时段与余额插件
Choose default and per-name provider, model, and reasoning effort for newly created DeepSeek Harness Agent Teams teammates, with live team status and restart-safe route views.
DSH compatibility guard: repairs compaction limits, fail-safe sandbox escalation, and missing tool descriptions for third-party models; capability lookups are cached and the deprecated prompt-injection path is off by default.
Grant official-style reasoning-effort selection to hand-declared pi-ai models (DeepSeek relays and the like) in the chat box
DSH Effort Router(模型分流):按每一轮请求的难度自动选择模型与思考强度——简单问题走便宜快模型,难题才叫强模型。规则判定零 token,灰区交给一次小模型裁定,图片档强度随难度升降(low/high/max)。每轮请求级覆盖,不改你的会话选择与默认模型。
Require and relay concise action summaries before tool calls in DeepSeek Harness
Cost policy for the DeepSeek Harness: per-call metering plus a local fuse that enforces budget, model and reasoning-effort limits before any token is spent.
dsh web Settings section (Custom models): subagent delegation allow-list, plus per-model image-input / reasoning-effort mapping.
Primary-controlled deliberation and opt-in masked review for DeepSeek Harness
Flash-only reasoning-effort routing for DeepSeek Harness: classifies each step by its shape and sets the DeepSeek flash model's reasoning effort to match. The model never changes; effort max is opt-in.
Merge consecutive reasoning blocks in the DeepSeek Harness host stream
Configure Responses reasoning mode and summary per provider/model in DeepSeek Harness
在 dsh 模型目录的每一行直接设置该模型的思考强度范围与支持的输入模态 (Per-model reasoning-effort range and accepted input modalities, right in the dsh Models catalog)
Codex-style folding for the DeepSeek Harness web transcript: a running turn keeps only its last few actions visible, folded behind one row that opens the whole turn.
Command palette for DeepSeek Harness (Cmd+K) — new chat, jump to session, switch model and reasoning effort, permissions, slash commands, theme, language and settings, all from one fuzzy-searchable bar.
Capability metadata for hand-declared gateway models, resolved at runtime by bare model name (context window, output cap, reasoning levels, vision) — matched against the official DeepSeek route's catalog, the installed pi-ai catalog, opencode and models.dev, without writing to your settings.
Project handoff skill: a dated handoff.md per project — current state, decisions with reasons, pitfalls as symptom/cause/fix, actionable next steps — so any fresh session resumes without archaeology. Bilingual EN/CN.