dsh-plugin-gpt-load
Run the following command in DeepSeek Harness:
dsh plugin install wesleyel/dsh-plugin-gpt-load
Paste the following prompt into your AI chat to install this plugin:
Fetch the plugin from https://github.com/wesleyel/dsh-plugin-gpt-load, run dsh plugin install wesleyel/dsh-plugin-gpt-load in your target profile directory, then restart Harness and configure the gateway URL and AccessKey in Settings to complete installation.
About this plugin
The gpt-load gateway serves models that speak both OpenAI and Anthropic protocols from a single pool, but DeepSeek Harness pins protocol selection to the provider route level and ignores per-model protocol declarations. Users end up either splitting into two groups or hand-writing protocol overrides in settings.yaml, which becomes unwieldy as the model list grows.
dsh-plugin-gpt-load bridges the gap. It syncs the model pool from /api/models at startup and on a schedule, registers a single GPT-Load group in the model picker, and assigns each model a protocol based on the upstream protocols field. You can also pin any individual model to OpenAI or Anthropic from the settings UI. Context window, max output tokens, multimodal input, and Thinking levels are all auto-populated. The AccessKey is stored via Harness credentials, and sync writes only touch its own config section, leaving the rest of your settings and comments intact.
If you manage multiple models through gpt-load or a similar unified gateway and want a clean, auto-syncing, per-model protocol entry point inside DeepSeek Harness, this plugin is built for you.
Use Cases
- Manage a multi-protocol gpt-load model pool without manually splitting provider groups
- Periodically auto-pull the model list and capability metadata instead of hand-writing settings
- Pin specific models to OpenAI or Anthropic while letting the rest resolve automatically
Best For
- Operators managing multi-protocol LLM fleets behind gpt-load or similar unified gateways
- DeepSeek Harness users who want a single clean entry point for model inference
- LLM platform admins who require credential vault storage and non-destructive config writes
Related Plugins
A unified suite combining hot runtime injection, task-aware thinking-mode routing, and a graded session protocol with red-team gates to sustain model diligence across long-horizon inference.
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.