DeepSeek Harness Plugins · Model Inference
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
A pure Node.js duration formatting tool with no network or external service dependencies, designed for model inference workloads.
A DeepSeek Harness plugin that execs the local Grok Build CLI to offer text/ask, image generation, and video generation tools, supporting SuperGrok login or XAI_API_KEY auth.
A pure Node.js line-level text diff plugin with no network dependency, registered as an in-session callable tool for model inference.
A pure Node.js plugin that converts JSON to SQL locally with no network or external service dependencies.
A pure Node.js text wrap plugin with no network or external service dependencies; invoke its registered tools directly in a session after installation.
A zero-config multimodal router that delegates image perception to local vision models and enables private local realtime voice conversation.
A local OCR plugin for macOS and Windows that exposes system native text recognition as a model-facing tool for DeepSeek Harness.
Calls a vision model to extract normalized bounding boxes from images, giving text-only agents precise object positions, layout, and verifiable spatial geometry.
A DeepSeek Harness plugin for configuring reasoning effort in model inference.
Auto-switch to an alternate model before DeepSeek peak hours and back afterwards, avoiding official peak-hour pricing.
Adds a local microphone button to DeepSeek Harness with offline Parakeet speech-to-text; audio never leaves your computer.
A DSH chat composer plugin that adds one-click and auto-before-send prompt optimization with before/after diff review, dual backend routing, and fail-safe fallback to the original text.