llm 98

KoboldCpp for DeepSeek Harness - a tool plugin that lets the harness online model hand repetitive text and vision (OCR) labor to a local KoboldCpp (llama.cpp) server.

为 opencode go 设计的 API 池:根据余额热管理切换不同 API key,官方用量/余额查询 + 设置页图形化管理。

Provider discovery, matching, health checks, and pre-output failover for DeepSeek Harness.

Provider failover for DeepSeek Harness: on rate-limit (429) or quota exhaustion, automatically switch between providers with configurable per-provider models, fall back to the last one (e.g. DeepSeek official), and show a visible notice in the UI.

OpenAI-compatible gateway adapter and dialect fixes for DeepSeek Harness

Transport-layer data masking for deepseek-harness (dsh): sensitive values never leave the process — the model sees placeholders, you see real values restored live in the stream.

DeepSeek Harness plugin: a model-callable local_vision tool that describes images with a local vision model through Ollama. No cloud, the image never leaves the machine. Two tiers: fast (text/colors/summary) and detailed (full description).

DSH 双模型路由插件：主对话留在 DeepSeek，图片任务自动分发到显式配置的视觉模型，再把分析结果交回 DeepSeek 继续。

DeepSeek Harness plugin: per-request thinking-mode switch (enable_thinking) for Qwen3-family models on OpenAI-compatible endpoints

DeepSeek Harness (dsh) plugin: /refine command + settings panel for the dsh-continual-harness self-refinement engine — entry browsing, refinement history timeline, one-click rollback, auto-gate audit; engine optional at runtime.

The LLM debug console inside DeepSeek Harness — capture every model call, see everything, replay anything.

DSH plugin: expose Trae Work CN credits as a local OpenAI/Anthropic compatible API for Claude Code, Cursor, Cline and other agents

Unsloth for DeepSeek Harness - a pure-client tool plugin that connects the harness online model to a locally running Unsloth Desktop (Unsloth Studio) server for repetitive text and vision (OCR) labor.

AgentTeams for DeepSeek Harness: multi-agent team collaboration (captain, members, tasks with dependencies, messaging) driven by natural language, with a tree monitor in the web GUI

DeepSeek Harness plugin: LLM model selection & deployment analysis (deployability, VRAM, TTFT, latency, throughput, power for 38 models × 20 GPUs/NPUs)

脑珊瑚 Coral Memory — 面向 LLM Agent 的热度感知记忆层（DeepSeek Harness 插件）：MCP 记忆工具（memory_search/insert/thread_* 等 16 个）+ 设置面板（统计/容量/检索/配额）。作者：Mr. Code Muggle (@Ne)。

A searchable, sortable model picker for the DeepSeek Harness Web UI — provider groups, favorites, live per-task cost, prices and context windows.

Active Context Pruning (ACP) for the DeepSeek Harness — model-driven context management as a CompactionEngine backend.

DeepSeek Harness plugin: role-based model routing — the planner agent runs on deepseek-v4-pro, delegated executor subagents run on deepseek-v4-flash.

Configurable model-request retry limit for DeepSeek Harness (Settings → General).

Modellix Design, LLM, and Web providers for DeepSeek Harness

Per-model reasoning effort declarations for GPT, Gemini and OpenAI-compatible providers in DeepSeek Harness

Bridge OpenCode CLI free models into dsh (no OpenCode login needed) via an llm adapter

OpenCode plugin — define custom AI agent roles via YAML with per-role prompts, models, skills, and permissions

Use ChatGPT (Codex) subscription as a DeepSeek Harness LLM provider via OAuth — standalone plugin, no upstream patching, proxy-aware.

Background auto-fill of llm-pi-ai model capability fields from overrides → pi-ai catalogs → models.dev

Custom request headers settings page for DeepSeek Harness: per-route HTTP headers for custom model (llm-pi-ai) providers, written into the llm-pi-ai user settings section

DeepSeek Harness compatibility plugin for third-party OpenAI-style gateways: tolerates SSE streams that close without the [DONE] sentinel after a complete reply.

LLMTR AI gateway provider for the DeepSeek Harness LLM seam: 170+ chat models and Türkiye-hosted LLMs behind one OpenAI-compatible key

DeepSeek Harness plugin for Readproof: stable readproof:// identity, freshness policy, snapshots, per-run manifests, diff, byte-exact replay, and evidence bundles as model-callable tools.

OpenAI-compatible LLM adapter plugin for DeepSeek Harness with configurable default sampling parameters (temperature / topP / topK / penalties / seed) over a providers dict.

Token usage, estimated cost, and a 52-week activity dashboard for DeepSeek Harness Web

DSH prompt 工具箱：预制 + 自定义 prompt 模板，点击即插入当前对话输入框（v1 常规模式 + v1.1 智能模式悬浮卡） | Prompt toolbox for DeepSeek Harness: preset + custom prompt templates, one-click insert into the conversation.

TimeWeb Cloud AI provider for DeepSeek Harness — a "Time Web" section with "Добавить агента Time Web": auto-fetch models from the agent, one-click create of a correct provider, success toast, auto-close to Models.

Calendar todo plugin for the dsh web GUI: a sidebar entry plus a week/month calendar where you drag-select time ranges to create tasks, with Eisenhower urgency/importance quadrants, subtasks, pinned execution session + LLM provider, and host-authoritative

DeepSeek Harness plugin: LLM cost metering with budget enforcement (deny/warn), peak-valley pricing and a live usage dashboard

dsh plugin: bounded retry for gateway network_error failures the stock retry policy cannot classify (e.g. OpenCode Zen finish_reason: network_error)

Content-free request and stream diagnostics for DeepSeek Harness

对抗式多视角代码审查插件：多透镜并行攻击式审查、确定性静态哨兵、跨视角合并去重、严重度定级、抑制规则与审查历史。支持 dsh（DeepSeek Harness）工具接入与独立 CLI 两种入口。

Model-routed subagent delegation for DeepSeek Harness: a `subagent_model` tool with per-call provider/model/max_tokens overrides and a built-in `model: "auto"` routing policy (anchored to the parent's own model by default; task-tier upgrades, failure esca

Brave Search API-backed web search provider for the DeepSeek Harness web capability seam (ctx.web)

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

NewAPI (OpenAI-compatible gateway) chat-completions adapter for the DeepSeek Harness LLM seam. Provider route id 'newapi', display name 'NewAPI'. Dual-face: host half registers the LLM adapter and chat-only model discovery; browser half adds a NewAPI sett

LLM API rate limiting, concurrency control, queuing, and adaptive cooldown for DeepSeek Harness

DSH plugin: expose Trae Work CN credits as a local OpenAI/Anthropic compatible API for Claude Code, Cursor, Cline and other agents. Ships a DSH web settings page (status, stats, token management, hot-reload config).

DeepSeek Harness 插件：极简智能体浏览器（naked-fetch）→ web_fetch / web_search_bing 工具，自带引擎零外部依赖

Request image optimization, intelligent retention, and observability for DeepSeek Harness (DSH)

Independent team coordination, dynamic routing, usage ledger, and Web panel for DeepSeek Harness