DeepSeek Harness Plugins · Model Inference
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
Use your ChatGPT Codex subscription directly in DeepSeek Harness for model access, search, quota management, and image generation without an OpenAI API key.
An OpenCode Go inference provider for DeepSeek Harness that registers the zen-go route, unifying chat, responses, and messages surfaces with multimodal support, reasoning tiers, and per-call usage tracking.
Adds multi-provider ASR, TTS, and real-time transcription WebSocket services to DSH hosts—install with one command, no source changes.
SSH/SFTP workspace plugin for DeepSeek Harness featuring terminal, file management, and AI-assisted command troubleshooting.
Adds a per-model capabilities fold to third-party provider cards on the Models settings page, declaring accepted input modalities (e.g. images) and reasoning levels.
Registers the generate_image tool for DSH agents, enabling image generation and reference-based redrawing via Doubao Seedream or Qwen models with in-session embedding.
Bridges WeChat, QQ, Email, 5G messaging, Feishu, and HTTP webhooks into persistent Harness Agent sessions with multi-channel isolation and rpcId reply claiming.
Advisor plugin for DSH: a lightweight model executes the main path while a stronger reviewer injects plan/correction/stop signals at key checkpoints to curb self-validation blind spots and long-horizon drift.
Displays peak, off-peak, and campaign rate badges per provider and model in the selector and composer, with a live switch countdown so you know the real cost before switching.
An education-mode plugin for DeepSeek Harness that gathers authoritative sources, teaches systematically, writes persistent course notes, and quizzes you interactively until the material sticks.
Upload vLLM-Ascend NPU profiling artifacts to get auto-parsed operator swim lane charts, latency attribution, and structured performance optimization reports.
Integrates MiniMax ASR and TTS into DeepSeek Harness as a global plugin, providing an audio transcription tool, voice input button, automatic spoken summaries after each task, and hands-free real-time conversation.