DeepSeek Harness Plugins · Model Inference
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
Prunes the toolset to a configurable whitelist on turn 1 with an exploration-framing preamble, then fully restores all tools from turn 2 onward.
A DeepSeek Harness plugin that leverages model inference to assist with contract clause review and analysis.
Adds a vision_analyze tool to DeepSeek Harness that sends an image to an OpenAI-compatible vision model and returns a text answer.
A visual configuration plugin for DeepSeek Harness context compaction, supporting percentage or absolute-token thresholds with a live context-usage ring indicator.
A DeepSeek Harness plugin for Feishu Meego that lets agents search projects, query work items, check todos, and with user confirmation create, update, or transition work items via the local meegle CLI.
Injects an ambiguity-handling rule into the system prompt so the model clarifies unclear user intent with listed options instead of guessing.
Use Tencent CodeBuddy chat models in DeepSeek Harness with browser OAuth login, automatic token renewal, and model-list sync.
DeepSeek Harness model-inference adapter that exposes 325+ AI providers from the aimux unified LLM layer as hot-swappable dsh model routes.
Real-time conversation cost tracker for DeepSeek Harness that calculates per-call fees from true token usage with peak and off-peak pricing, displayed live in the sidebar.
Access Kimi K3, Claude, GPT, Grok, GLM, DeepSeek, and MiniMax-H3 through one endpoint and one API key via AllRouter, with direct token pricing and real per-call cost.
Routes Codex, Grok, Claude, Cursor, Antigravity, and Kiro official subscriptions into the DeepSeek Harness account pool with a provider selector on Windows.
Rewrites the composer draft via an auxiliary LLM call; replaces the text only without sending a message or starting a turn.