modelspoke
Run the following command in DeepSeek Harness:
dsh plugin install americanjeff/modelspoke
Paste the following prompt into your AI chat to install this plugin:
Run dsh plugin install americanjeff/modelspoke in the DeepSeek Harness terminal to install the plugin. The source code is available at https://github.com/americanjeff/modelspoke . After installation, restart dsh once and open the modelspoke settings page in the web UI to start configuring providers.
About this plugin
Connecting local or OpenAI-compatible model servers through DeepSeek Harness leaves several gaps in the built-in custom-provider flow: no automatic capability discovery, no image-input support, no reasoning-effort controls, and no unified management UI. modelspoke is a purpose-built plugin that fills exactly those holes.
It offers first-class capability discovery for llama-swap and Ollama, ships preset capability tables for common backends such as llama-server, vLLM, and sglang, and renders a full settings page inside the dsh web UI. There you can add providers, edit per-model fields (context window, max output tokens, reasoning-effort mapping, image input, nothink), and maintain multiple configuration profiles for the same underlying model. The plugin also patches the missing upstream support for inline images in session chat.
It is designed for developers who run several local OpenAI-compatible servers, manage large local model collections via llama-swap or Ollama, and want a single unified entry point instead of hand-editing provider configs for every endpoint.
Screenshots
Use Cases
- Managing several local OpenAI-compatible servers and switching between them from one place
- Reading capability metadata automatically from a llama-swap or Ollama model collection
- Tuning per-model fields such as reasoning effort, image input, and context window
Best For
- Local model enthusiasts and solo developers
- Managers maintaining large model collections via llama-swap or Ollama
- Harness users who want a single UI entry point instead of hand-editing provider configs for each endpoint
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.