DeepSeek Harness Plugins · Model Inference
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
A prompt audit plugin for DeepSeek Harness, categorized under model inference.
A DeepSeek Harness plugin for caveman-style model inference output.
A DeepSeek Harness plugin that converts screenshots into code via model inference.
Integrates local Codex CLI as native dsh tools for one-shot tasks, code review, and session resume with a default read-only sandbox.
A floating desktop pet in the Web GUI that reacts to agent actions in real time, with instant image hot-reload—just drop a file and it updates, no restart needed.
LongCat-2.0 adapter for DeepSeek Harness with 1M context, thinking mode, tool calling, and SSE streaming.
A model inference gateway plugin for DeepSeek Harness.
Injects provider routing (only/order) and quantization limits into all OpenRouter requests, configurable via a plugin settings card with persistent storage across restarts.
Cross-stage commercial strategy plugin covering business models, pricing architecture, channel economics, profitability analysis, and business plans from demand to monetization.
A configurable auto-routing tier for DeepSeek Harness that classifies each request and delegates to the optimal model on an existing provider route.
A policy-driven multi-source model router for DeepSeek Harness that selects the first healthy source along a per-task candidate chain, supporting mixed free/subscription/low-price source scheduling with bounded failure recovery.
Installs surgical probes in DeepSeek Harness to capture structured JSONL runtime evidence, letting the agent prove failing branches, apply focused fixes, and clean up all instrumentation.