DeepSeek Harness Plugins · Model Inference
Plugins that extend DeepSeek Harness (DSH) with vision, workflows, memory, web tools and more.
Pins agent thinking and reply language to Chinese across any DeepSeek Harness deployment via one stable system-prompt section, with no tool or model changes.
Adds image understanding and generation tools to a text-only DeepSeek Harness main model via configurable OpenAI-compatible or DashScope endpoints.
Built-in short-video templates that call China-based LLMs to generate a full content package and platform-specific publishing plans for Douyin, Bilibili, and more.
Automatically routes requests to tiered models by task complexity with built-in vision support, balancing performance and cost on your API bill.
Two-phase session auto-captioning for dsh: instant keyword-based titles on every message, refined by the cheapest available model when idle with a one-line summary, all running in the background.
A DeepSeek Harness add-on that prunes oversized tool output before each model request to control token growth in long sessions, while preserving history for targeted recall.
An academic-behaviour layer for DeepSeek Harness: research persona, anti-fabrication integrity rules, and bilingual conventions — no tools registered, it simply makes the agent read like a research assistant.
A native in-process DeepSeek Harness plugin providing a local-first HTML review, editing, and annotation workspace with six html_review tools and no MCP layer or external server.
Auto-discovers model listings, pricing, and capability flags from any OpenAI-compatible host, normalizes them, and generates ready-to-use dsh config fragments.
Enables text-only DeepSeek models to understand images via a local Ollama vision engine, providing native paste support and nine analysis tools without cloud API keys.
A one-click prompt enhancement button beside the input that rewrites your draft via LLM or offline fallback, with a confirm dialog and no auto-send.
Scientific computing plugin that enables agents to perform numeric evaluation and symbolic algebra (differentiation, integration, factoring, and solving) in-conversation.