Plugins

Configure the contextWindow for models registered in dsh, so long sessions are no longer misjudged as context overflow.

OpenCues for DeepSeek Harness — LLM word alternatives and `_`-gated blank fill-ins in the dsh composer.

Installs four multi-agent math-research presets for DSH: v2 (probability-driven pipeline with multi-verifier debate), v3 (paper-style Markdown knowledge base with a planner agent and a reusable method library), v4 (persistent self-organizing residents that message and meet), and v5 (a research institute with an academician who decomposes and assigns work, voting researchers, temp workers, group chat and a compare-and-set task board); all four support checkpoint resume, human intervention, and an optional Lean formal-verification switch (off/encourage/require) whose passing proof turns the vote into a fidelity check of the Lean statements.

在deepseek harness中使用workbuddy api，因为公司只提供workbuddy积分

Select text in a committed DSH response to open an independent, persistent, read-only multi-turn Topic bound to source-session evidence and project files while the source Agent keeps running.

Feishu (Lark) UI for DeepSeek Harness — a panel-driven control console: every slash command is a button on the ⚙️ control-panel card, in-card approvals and questions, live streaming cards, and one-QR app setup.

Block-addressed document editing: an MCP server exposing geml_get / geml_set / geml_check and friends, so an agent reads or rewrites one addressed block of a Markdown or GEML document instead of the whole file. Ships the GEML authoring skill and a code-graph skill that builds and navigates a project call graph as GEML codemaps.

Auto mode for DeepSeek Harness (dsh): an Auto entry in the Permissions selector whose approval prompts are answered by OpenGuardrails policy instead of a human — plus the full OGR guard engine underneath. No core changes.

Converts session, agent-loop, LLM, and tool lifecycle events into OpenTelemetry GenAI traces and metrics, exported over standard OTLP/HTTP to any compatible backend, with content capture off by default.

Agent-Native document handling for DSH — addressable blocks let an agent read and edit one section instead of the whole file. Ships the geml MCP server and the authoring and code-graph skills.

Classifies pending tool calls with Jev (TypeSafe AI's non-generative decision model) on tools/pre-execute — confident high-risk calls are denied, ambiguous ones escalated to approval — and auto-approves clearly-granted reversible calls on approval/request behind argument-evidence gating. Also subclasses the session-reference resolver so snapshots keep Jev-scored messages instead of dropping oldest-first. Degrades to stock behavior on any provider failure; shadow mode with a /jev-stats command; TypeSafe or OpenRouter endpoints; 64 tests.

Voice input plugin for DeepSeek Harness (dsh): a microphone button in the composer turns speech into a draft transcript, with a choice of speech-recognition backends, optional polish through dsh own LLM routes, and a native settings page.

Detects prompt-injection, jailbreak, and secret-leak patterns on the agent/pre-step, tools/pre-execute, and tools/post-execute seams with allow/ask/block tiers, sanitized defend/detection audit events, a defend_report tool, and a destructive-delete command guard.

DeepSeek Harness bundle: the AgentRouter relay as one pi-ai route with a domestic/international endpoint switch in Settings, plus the outbound request compatibility the relay requires.

Role-based LLM retry & fallback strategies.

Role-based LLM retry & fallback strategies.

Role-based LLM retry/fallback strategy.

A drawer panel for the DSH web UI with a paginated list of free LLMs; one-click or batch writes each as a custom model provider into Settings, plus a local multi-key rotation proxy.

Read any page as clean main content: charset auto-detect (GBK/GB2312/UTF-8/Big5), noise stripping, compact text/Markdown, offset continuation, optional SPA rendering, batch reads and site crawling.

StrataGate gives DeepSeek Harness evolving, local-first long-term memory for user preferences, project decisions, conversations, and tool results across sessions; it layers history with user-tunable time decay, reinforces memories through real answer use, shows workspace-aware usage counts, and keeps recalled evidence traceable to its original messages.

Native Rapid-MLX provider for DeepSeek Harness (dsh) — dsh reads model facts from the server instead of your settings.yaml.

AI 原生智能爬虫与 JavaScript 逆向工程平台，基于 DSH、Patchright/CDP 与独立语义运行时，从浏览器证据恢复参数生成逻辑并交付可验证 Solver。

Long-term memory for DSH: conversations auto-distilled into atomic facts, scene summaries and a persona profile, with relevant memories injected before every model step. Zero-config out of the box — hybrid BM25 + vector retrieval, optional offline local embeddings, chat/work family separation.