dsh-relay-fast
Run the following command in DeepSeek Harness:
dsh plugin install Rock-ql/dsh-relay-fast
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install Rock-ql/dsh-relay-fast (source: https://github.com/Rock-ql/dsh-relay-fast) in your terminal, then restart DSH Desktop to activate the plugin.
About this plugin
When routing DeepSeek Harness traffic through third-party relays, newly added models frequently lack a reasoningEfforts declaration, the local model list drifts out of sync with the relay, and turning on the Fast priority tier requires manual, model-by-model configuration. dsh-relay-fast bundles these scattered tasks into a single lightweight plugin, bringing DSH Desktop/Web much closer to a clean, official-API experience out of the box.
The plugin delivers three core capabilities. First, it autofills missing reasoningEfforts for new llm-pi-ai models while staying protocol-aware for both openai-completions and anthropic-messages, and it never overwrites an existing explicit declaration. Second, it lets you import missing models straight from the relay /models endpoint in Settings, eliminating the tedious add-one-by-one workflow. Third, it exposes a dynamic Fast toggle that appears only when the endpoint explicitly advertises Fast support in /models or when the user has listed the model in fastModels; when enabled, requests carry service_tier: "priority" with zero name-based guessing.
It is built for DSH Desktop/Web users who route DeepSeek traffic through third-party relays, juggle multiple providers, see models appear and disappear frequently, and want the leanest possible inference setup without touching plugin source code. All overrides live in cordis.patch.yml, keeping the existing profile completely intact.
Use Cases
- One-click sync of missing models from a newly added relay
- Autofill reasoningEfforts for new llm-pi-ai models
- Dynamically enable the Fast priority tier based on endpoint metadata
Best For
- DSH users routing DeepSeek traffic through third-party relays
- Developers switching between multiple inference endpoints frequently
- DSH Desktop/Web users who prefer zero-code inference configuration
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.