DeepSeek Harness Plugins · Model Inference
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
Integrates the local GSV-TTS-Lite engine into DeepSeek Harness for voice cloning, streaming synthesis, and auto-reading, running entirely locally.
A DeepSeek Harness plugin that adds an artifact sidebar and bookmarking for agent conversations, with detection, preview, and persistence of artifacts.
dsh-kiro is a plugin for DeepSeek Harness that provides Kiro model inference support with multiple login methods, automatic token refresh, live model discovery, streaming, and reasoning effort controls.
A plugin that injects private, per-project prompt rules into DeepSeek Harness sessions, matched by git remote, repo path, or cwd prefix.
A plugin for managing OpenCode Zen free-tier and Go-tier models in DeepSeek Harness, with tools and a settings page to fetch, add, remove, and sync model listings.
A real-time repetition guard for DSH that safely halts generation the moment a string repeats past a configurable threshold, preserving all output already produced.
Browse, edit, and lint a Karpathy-style interlinked markdown wiki in the DSH sidebar with clickable wikilinks, provenance links, backlinks, and a built-in llm-wiki agent skill.
A zero-dependency chemistry arithmetic plugin with four deterministic tools for molar mass, mass-mole-particle conversions, solution concentration, and pH/buffer calculations.
Hover over any model in the selector to instantly view its input types, context window, max output tokens, reasoning level, and auth environment variable.
Automatically describes images for text-only models via a local Ollama VL model, injecting the result into the same inference step with no cloud and no API key.
A zero-dependency deep research plugin that mechanically anchors citations, counts independent origins via a lineage DAG, runs adversarial evidence search, and degrades to a deterministic fallback without an LLM.
Integrates Grok subscription as an LLM provider in DeepSeek Harness, supporting native multimodal inference with grok-4.6/4.5 via the chat-completions endpoint.