dsh-llm-inspector
统一 LLM 请求/响应检查器:调 reasoning effort、外部思考(think)导出、流量与包分析 —— DeepSeek Harness 插件
107 results
统一 LLM 请求/响应检查器:调 reasoning effort、外部思考(think)导出、流量与包分析 —— DeepSeek Harness 插件
DSH插件:将推理强度选择重构为滑动条组件
Capability-aware reasoning effort slider for the DeepSeek Harness web model selector.
Declare per-model reasoning efforts, image input, and openai-completions dialect on hand-declared llm-pi-ai routes. Writes the official namespace; does not intercept llm/stream or replace the Models page.
DeepSeek Harness 实用工具箱:便签、API 余额与费用、推理等级、删除会话、对话节点导航条、轻拟物皮肤(monorepo,可一键安装全家桶)
Shows the selected model provider's account balance next to the composer model seat once a model and reasoning effort are selected.
DeepSeek Harness web 插件:把一个会话交给本机 Claude Code CLI 驱动。引擎选择器(DSH|Claude)、Claude 权限档替换访问模式、Claude 模型/effort 选择、原生工具卡片与子 agent 嵌套、dsh 原生审批、命令面板并入 Claude 命令(一次性命令带外执行)、订阅用量读数、导入本机 Claude 对话。Claude 进程由独立 broker 持有,插件更新与 dsh 重启都不中断它。
DeepSeek Harness LLM adapter plugin for locally deployed Qwen models behind a vLLM OpenAI-compatible endpoint: per-model multimodal switch, fully configurable reasoning efforts, and a web settings page (client plugin) for editing the deployment from the frontend
在 Settings 面板中为自定义 pi-ai 模型编辑推理等级(reasoningEfforts),支持多选、拖拽、复制粘贴,自动保存。
Volcengine Ark Agent Plan & Coding Plan providers for DeepSeek Harness (DSH), with verified thinking-effort compatibility
DSH plugin: Codex-style side conversations (/side, /btw) in an ephemeral right-workspace panel with Side/Subagents/Goal sections and a ChatGPT-style pinned-notes board (置顶摘要小黑板), plus a best-effort sidebar collapse hotzone.
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.
Per-model reasoning-effort and vision capability editor for DSH pi-ai provider profiles.
DSH plugin (NInfer engine only): fixes compaction failure on local qwen3.8-27b gateways served by NInfer — xhigh thinking burns the entire output token budget, so thinking is off for compaction-only, with the model's non-thinking sampling parameters; the same idea applies to other launch methods
CodeBuddy (copilot.tencent.com) provider bundle for DeepSeek Harness (dsh): 18 models (DeepSeek, GLM, Kimi, MiniMax, Hunyuan, auto) with adjustable reasoning efforts, CodeBuddy web search / web fetch backends for the stock dsh tools, and a settings card in the Web UI.
DSH web plugin: splits the composer model selector into two independent controls — a searchable, favorite-marking model dropdown and a reasoning-effort slider — plus Ctrl+P (cycle favorite models) and Ctrl+T (cycle reasoning effort) shortcuts.
A subagent tool that forces each child onto an explicitly chosen provider/model and reasoning effort instead of inheriting the parent's route
Floating, draggable, resizable video scene player for DeepSeek Harness. Plays scene-per-MP4 clips from a Stash-style scene server, plus best-effort YouTube / Twitch / Jellyfin / custom links.
Animated reasoning-level slider for DeepSeek Harness model selection.
Save and switch model and reasoning-effort presets in the DeepSeek Harness composer.
Provider-folded model selector for the DSH Web composer with catalog-driven reasoning effort choices.
DSH plugin: keyboard shortcuts to switch models and reasoning effort levels
DeepSeek Harness plugin that enhances custom provider setup by auto-discovering models and auto-populating contextWindow, maxTokens, vision, and reasoning capabilities from models.dev
Message review (configurable multi-round cross review + auto summary + thinking-effort) + quote (Q&A chips) for the DSH conversation UI