LLM Openrouter Decisions

Install it in the same environment as LLM using `llm install llm-openrouter-decisions`. It requires LLM 0.36 or later. After installing, set an OpenRouter API key, otherwise no models are registered and nothing appears in the model list.

It uses a single OpenRouter API key, the same one used by llm-openrouter. Set it with `llm keys set openrouter`, or supply it through the OPENROUTER_KEY environment variable. Models are only registered when a key is available.

The default is noul: the probability from 0 to 1 that a statement is true. Passing `-o answer_type choice` returns a category with a probability per option, `-o answer_type score` returns a probability-weighted position on a scale of 2 to 10 levels, and image-capable models also answer box questions.

Yes. The plugin registers every model in the OpenRouter decision models catalog, 30 from 13 makers as of October 2026, so you can run the same state and question against different models by changing only the `-m` flag and compare how their probabilities differ.

The content you supply as the prompt or on stdin, plus the question passed with `-s/--system`, is sent to the OpenRouter Decisions API for evaluation. The model catalog itself is downloaded from OpenRouter and cached for an hour in `openrouter_decisions_models.json` in your LLM user directory.

Use `llm aliases set jev14 openrouter-decisions/typesafe/jev-1.14` and then pass `-m jev14` on every call. The examples in the project use Jev 1.14 through such an alias.

Decision models do not write text; they return numbers such as probabilities, categories, scale positions or image boxes. If you need open-ended generation or chat output, use a text model instead. Decision models instead suit classifying, filtering and triaging many items from the command line.