This page answers: which models RelaySolo can use, whether it can use a local model, how your own API key works, and who receives what the AI processes.
AI models and keys
RelaySolo works with models from the major AI providers, using your own API key. It can also connect to a local model you installed yourself. With a provider, each request sends only the text being processed and the few passages it needs, straight from your computer to the provider you picked.
Which models you can use
RelaySolo connects to models from the major AI providers. You choose which one to use.
You can also use a local model; see below.
The AI does the work behind translating buyer emails and drafting replies, turning inquiries into cards and drafting quotes from your price list, summing up what each client said and building the morning briefing, and answering from your own files with the source cited. The full list is on features.
Local models
RelaySolo can connect to Ollama or llama.cpp that you have installed yourself.
RelaySolo does not install them for you. You set up Ollama or llama.cpp and download the model you want first.
With a local model, your questions are processed on your own computer and are not sent to a model provider.
Sending and receiving email and checking for updates still need an internet connection. How fast and how good a local model is depends on your computer and the model you pick.
Your own API key
Create an API key with the provider and add it to RelaySolo. Requests go straight from your computer to the provider, without passing through our servers.
Usage and charges follow your agreement with that provider. The key is stored encrypted on your computer; we don't have it.
What is sent when the AI is used, and to whom
-
Leaves your computer Each time the AI is used
- What
- That one question, or the text being translated or drafted, plus the parts of your emails and files it needs. Not your whole database.
- To
- The AI provider you chose, straight from your computer with your key.
- You control
- Which provider, whether to use a local model, and whether to ask at all.
Say you ask for a draft quote. The request carries that inquiry and the rows of your price list it uses. Other clients and other prices stay where they are.
With a local model you installed yourself, this step is not sent to a model provider.
This is one of four cases where data leaves your computer. The others, such as email, are on where your data lives. When your own Claude or ChatGPT looks things up in RelaySolo, that is a separate case; see use it in Claude or ChatGPT.
Read the provider's own data terms
What happens to your request after it arrives is up to the provider, and each one has its own rules. Before you choose, read their data terms and look for three things:
- Whether what you send can be used to train their models.
- How long it is kept, and whether you can ask for deletion.
- Whether API use follows different rules from their chat website.
What stays on your computer
- Your AI chat history
- What the AI remembers about clients
- Your key settings
These are stored on your computer and not on our servers. Your API key is stored encrypted. That is also why you should set up a backup.
More answers: FAQ and the privacy policy.