Choose a provider
Presets fill in the base URL for popular providers, and any OpenAI-compatible service works.
In development. This part of Papeliz is still being built. This guide describes how it is designed to work, and details may change before launch.
Recommended: Muse Spark on the Meta Model API#
The first preset, selected out of the box, is Muse Spark on the Meta Model API. Its long context lets the assistant work across many notes at once, and the API accepts calls straight from the browser.
- Base URL
https://api.meta.ai/v1- Model
muse-spark-1.3- API key
- From the Model API dashboard at
dev.meta.ai
Other presets#
Papeliz speaks the OpenAI-compatible format, so any other compatible provider works too. The presets:
- OpenAI.
- OpenRouter: many hosted models behind one key.
- Groq, Together and Fireworks: hosted open-weights models.
- Google Gemini, through Google’s OpenAI-compatible endpoint.
- Ollama Cloud, OpenCode Zen and OpenCode Go: hosted models with a key from those services.
- Ollama and LM Studio running on your own computer, with no key.
- Custom: any other OpenAI-compatible base URL.
Calls from the browser#
Papeliz calls the provider from your browser, so the provider has to accept calls from a web page. That has been checked for the recommended preset and OpenAI. For the others, Test connection tells you.
Models on your own computer#
To use Ollama on your computer, add Papeliz’s web address to Ollama’s OLLAMA_ORIGINS setting so it accepts calls from the page. For LM Studio, start its local server with CORS turned on in the server settings. Then pick the matching preset.
Custom#
Choose Custom for any other service that speaks the OpenAI Chat Completions format. Enter its base URL, which usually ends in /v1, and a key if it needs one.