DeepSeek Harness Plugins · Model Inference
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
Enables text-only models to read images and generate descriptions via a dedicated vision route, eliminating the need for third-party tools.
Convert Markdown or text into playable HTML webpages or editable PPTX presentations. Zero dependencies, 5 built-in themes, and animations.
Provides local vision "eyes" for DeepSeek Harness, enabling screen capture and OCR text extraction.
Performance plugin for DSH large sessions featuring zero-copy fork, chunked warmup, and memory trimming to resolve blocking and OOM.
DeepSeek Harness plugin for Volcengine Ark Agent Plan and Coding Plan inference with verified thinking-effort compatibility.
Minimal PTC Agent aligned with RL training, featuring complete toolchain and Windows Git Bash support.
A minimalist DSH prompt-activation plugin offering multiple modes, managing tool visibility and lifecycle.
Integrates context-aware voice input into Composer, supporting local SenseVoice transcription and editing.
Migrates AiToEarn content suite to DeepSeek Harness, supporting creative direction, script generation, image/video creation, and Douyin publishing.
Enables text models to recognize images via MMX, automatically injecting visual context.
This plugin allows DeepSeek Harness agents to use a locally installed CloakBrowser for web automation and visual understanding.
A DSH plugin that enables image recognition for text-only DeepSeek models via local CLI.