llm-tiyuvta

LLM plugin for tiyuvta — prepaid open-weight-model inference with one price per model, no subscription.

Every model on the endpoint speaks the identical OpenAI-compatible API: streaming, tool calling, structured output, and vision where the model has eyes. The plugin reads the live model catalog, so new models appear without a plugin update.

Installation

Install this plugin in the same environment as LLM:

llm install llm-tiyuvta

Configuration

Get an API key from inference.tiyuvta.ai, then:

llm keys set tiyuvta
# Paste key here

You can also set it as the TIYUVTA_KEY environment variable — note that a key stored with llm keys set takes precedence over the environment variable.

Models appear in llm models once a key is configured.

Usage

Run a prompt (short aliases work — qwen3.8-27b for tiyuvta/qwen/qwen3.8-27b):

llm -m qwen3.8-27b "Three reasons the sky looks blue"

Chat interactively:

llm chat -m qwen3.8-27b

Vision — attach an image:

llm -m qwen3.8-27b "describe this" -a photo.jpg

Tools and schemas work the way they do for any OpenAI-compatible LLM model:

llm -m qwen3.8-27b --schema 'name, bullet_points: three key points' "summarize: ..."

List the models the endpoint currently serves:

llm tiyuvta models

Refresh the cached catalog after a new model launches:

llm tiyuvta refresh

Development

To set up this plugin locally, first checkout the code. Then create a new virtual environment:

cd llm-tiyuvta
python -m venv venv
source venv/bin/activate

Now install the dependencies and test dependencies:

llm install -e '.[test]'

To run the tests:

pytest