dsh-model-router
Run the following command in DeepSeek Harness:
dsh plugin install tianji-qingtian/dsh-model-router
Paste the following prompt into your AI chat to install this plugin:
To install this plugin in DeepSeek Harness, run the command: dsh plugin install tianji-qingtian/dsh-model-router, with the full source address at https://github.com/tianji-qingtian/dsh-model-router.
About this plugin
Balancing cost and reliability is a core challenge when working with large language models. dsh-model-router provides an intelligent routing and optimization solution for DeepSeek Harness, automatically identifying simple questions and directing them to a cheap model for immediate answers, which substantially reduces ongoing usage costs. It also features graceful degradation for transient provider failures, automatically falling back to a cheaper model on errors like rate limits or timeouts to keep conversations running smoothly.
The plugin's core capabilities revolve around precise routing decisions: using zero-prefix flash judge calls, it distinguishes simple queries from complex tasks with minimal latency, processing the former quickly. It includes vision-aware routing to handle messages with images correctly, avoiding blind text-only responses. Live statistics are displayed via a Composer dock panel, showing per-session token usage, cache-hit rates, and estimated costs for transparent expense tracking. In auto mode, it prompts users for preferences, blending automation with control.
dsh-model-router is ideal for developers, researchers, or enterprise users seeking to optimize AI model costs. It suits high-frequency simple query scenarios like customer support or cost-effective daily development, where smart routing boosts efficiency. For users prioritizing service stability, its automatic fallback mechanisms add robustness, making model inference more resilient and reliable.
Screenshots
Use Cases
- When users ask simple queries, the plugin uses a cheap model to answer directly, saving costs.
- During temporary service errors, it automatically falls back to a cheap model, preventing conversation interruptions.
- Real-time viewing of token usage and costs via the Composer panel helps optimize model expenses.
Best For
- Developers using DeepSeek Harness, aiming to reduce model inference costs.
- Enterprise users handling high volumes of simple queries, seeking to control operational expenses.
- AI researchers looking for cost-effective solutions during experiments.
Related Plugins
ModLens is a vision plugin for DeepSeek Harness that gives text-only models sight by reading images pasted directly into chat, with zero-config setup and multiple vision engines.
On-demand vision for text-only DeepSeek Harness agents: built-in free keyless vision chain and 14 vision tools, routing image turns as tool calls to vision models with pixel fidelity, no Python needed, one-command install.
Give text-only models in DeepSeek Harness eyes, enabling image Q&A, long-screenshot OCR, UI restoration, and GUI visual tasks.
An armor-breaking plugin for DeepSeek: rewrites instruction priority, expands refusal-blocked phrases, persists across sessions, and shows a green active indicator.