DeepSeek Harness Plugins · Model Inference
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
AI-driven Partial Wave Analysis (PWA) plugin for DeepSeek Harness, covering PDG lookup, J^PC gatekeeping, config editing, ctpwa fitting, and goal-driven iterative convergence.
A hands-free voice assistant for dsh web featuring keyword wake-up, real-time speech dictation, spoken editing commands, and Chinese TTS, powered by local sherpa-onnx WASM inference with streaming responsive results.
Command Code Provider API chat plugin for DeepSeek Harness with model discovery, credential management, and account quota reporting.
Adds an invert selection button to the auto-fetched model list so you can quickly check only the models you already have.
Raises Node global fetch timeouts to 30 minutes so slow local models like Ollama are not cut off after 5 minutes of silence.
Visual GUI for managing per-model thinking levels, context windows, gateway compatibility fields, and one-click presets in DSH Web settings.
Use ChatGPT Codex, xAI Grok, Kimi, Copilot, and other AI subscriptions inside DeepSeek Harness via OAuth or API-key auth with quota display and multi-protocol routing.
Bridges a GitHub Copilot subscription into dsh as a streaming LLM provider with model listing, reasoning passthrough, and built-in login management commands.
A native DSH vision plugin that bridges image understanding to non-vision models via mimo-v2.5, with multi-format auto-transcoding support.
Adds a vision route to the text-only DeepSeek agent by delegating image analysis to an external provider and returning text descriptions to the model.
Adds Gemini vision reading and Flash/Pro smart routing to DeepSeek Harness, with image paste support, multi-key rotation, and pre-delivery visual self-checks.
Cross-disciplinary deep structural analysis skill featuring 16 lenses, an attack loop protocol, and deduction checks to produce confidence-graded, fact-bound analysis of complex systemic problems.