From zero to your first chat — four steps.
Solme doesn't bundle a model — conversations are powered by the provider you choose. Common choices: OpenAI, Anthropic, and various compatible endpoints or relays. Your key is stored only on your phone and never touches any third party.
Settings → API → Models → + Pick an API type (OpenAI-compatible / Anthropic-compatible), paste your key, and save — leaving Base URL and model name empty fills in the defaults. Use "Test connection" to verify before saving.
+ at the top right of the TA page Fill in a nickname and name. The persona description can be left empty, pasted from the community personas, or imported from an .md file. Avatars can be cropped in-app.
Send your first message and it begins. The plus panel at the bottom left of the chat page has tools, chat settings, and more — explore as you go.
The memory system's retrieval is powered by an Embedding model — a separate configuration from your chat model. Add it under Settings → API → Vector Search; test connection is supported there too.
Supports Tavily (key only), SearXNG (self-hosted), and custom endpoints. Configure under Settings → API → Web Search; after enabling, also turn on the web_search tool in the chat's plus panel.
Finish these four steps and you're on your way — the rest is yours to explore, one chat at a time.