Sunset Mesa Software

AI setup

Getting startedAI setup

The deterministic transforms need no setup. The AI transforms are bring-your-own-key: you connect a provider account you already have, or make a free one, and Rewire Text uses it directly. There is no Sunset Mesa Software subscription and no markup on top of what you pay your provider. Pick one provider to start. You can add others later.

How it works in the app

Open Settings, choose a provider, and paste your API key. The local providers (Ollama, LM Studio, and llama.cpp) are the exception and need no key. For Google Gemini, Anthropic, OpenAI, and xAI Grok, Rewire Text pre-selects sensible models, so you can start right away. For OpenRouter and the local providers it lists the models it finds, and you can type a model ID by hand.

Each transform asks for a capability tier rather than a specific model: Light (cheapest and fastest), Standard (balanced), or Advanced (most capable). You decide which model fills each tier for a given provider. The defaults are a fine place to begin.

Once you have entered a key, use the Test connection button in provider settings to confirm the key and endpoint work before you rely on them.

Getting a key, by provider

  • Google Gemini. Create a key at Google AI Studio. The free tier is enough to try everything, and it is the cheapest option for everyday use.
  • OpenAI. Create a key at platform.openai.com under API keys, and add a little credit to your account.
  • Anthropic (Claude). Create a key in the Anthropic Console, and add a little credit to your account.
  • xAI (Grok). Create a key in the xAI Console, and add a little credit to your account.
  • OpenRouter. Create a key at openrouter.ai. One key reaches many models from different providers, including some free ones.

Once you have a key, paste it into Settings for that provider and you are ready to go.

Run a model locally

If you would rather keep everything on your own machine, Rewire Text works with three local runners: Ollama, LM Studio, and llama.cpp. None of them use an API key, and nothing leaves your computer. Pick whichever you already run, or the one that fits your setup. Each is configured the same way: point Rewire Text at the local server in Settings, then pick the model from the list.

  • Ollama. Install Ollama and pull a model from its library (a Gemma 4-class model is a good default). Rewire Text talks to Ollama at http://localhost:11434 by default.
  • LM Studio. In LM Studio, download a model and start its local server. Rewire Text connects at http://localhost:1234 by default.
  • llama.cpp. Run the llama.cpp server (llama-server) with the model you want. Rewire Text connects at http://localhost:8080 by default.

If you changed the port for any of these, update the server URL in Settings to match.

Hardware matters for all three. Small models run on most recent laptops. A Gemma 4-class model around 12B parameters wants a capable GPU, or a recent Apple Silicon machine with roughly 16 GB of memory or more. Smaller models are lighter, larger ones need more.

Connect any OpenAI-compatible endpoint

Beyond the named providers, Rewire Text can talk to any service that exposes an OpenAI-compatible API. Add it as a custom provider, enter the base URL and the model ID you want, and add an API key if the service requires one. Remote gateways and proxies usually need a key. A server running on your own machine often does not. This covers self-hosted setups, company proxies, and providers that are not in the built-in list.

Cost and usage

Because it is bring-your-own-key, you pay your provider directly. For everyday text on the Light and Standard tiers that is usually pennies. Rewire Text estimates the cost of each AI transform and shows a running total in Settings, so the bill never surprises you. Deterministic transforms and local models cost nothing to run.

Recommendations

  • Easiest free start: a free-tier Google AI Studio key, or a free model on OpenRouter. Either one lets you try the AI transforms at no cost.
  • Fully local: run a Gemma 4-class model through Ollama, LM Studio, or llama.cpp on capable hardware, so no text ever leaves your machine.

New here? Start with Getting started.