AI Agent Hub
Back to plugins
dsh-model-fix preview

dsh-model-fix

Web Tools Updated 2026.09.12

Run the following command in DeepSeek Harness:

dsh plugin install TikaFlow/dsh-model-fix

Paste the following prompt into your AI chat to install this plugin:

In DeepSeek Harness, open the Plugin page, search for dsh-model-fix in the plugin marketplace, or manually install the package from https://github.com/TikaFlow/dsh-model-fix; the plugin takes effect immediately after installation.

About this plugin

Adding non-official model providers in DSH is rarely the hard part; the tedious part is filling in every missing metadata field by hand — reasoning efforts, context window, max tokens, image input support. dsh-model-fix eliminates that friction by pulling authoritative data from models.dev and auto-populating every absent field the moment a custom model appears, while leaving your existing configurations completely untouched.

Beyond basic fill-in, the plugin bundles a set of practical features: per-model memory of your last manually chosen reasoning level that restores automatically when you switch back; a default developer-role compatibility flag written to every openai-completions route; a provider exclusion list that makes the plugin skip specific providers entirely; and a built-in offline cache so everything works the instant DSH starts, with data refreshing in the background. Every action can be performed through a visual settings card — no YAML editing required.

If you are constantly adding custom models from various vendors in DSH or are tired of manually typing contextWindow and maxTokens every single time, this plugin is made for you. Install it and it simply works — zero configuration needed.

Screenshots

Use Cases

  • Auto-populate reasoning efforts, context window, and max tokens when adding custom models
  • Restore the last manually chosen reasoning level when switching between models
  • Exclude a specific provider so the plugin skips it entirely

Best For

  • DSH users who frequently add non-official model providers
  • Developers tired of manually filling in contextWindow and maxTokens
  • Power users managing multiple custom endpoints who want to cut repetitive config