Connect SillyTavern to UnsavedAI
Use a hosted AI model for character conversations without running a GPU yourself.
Choose the connection
Click the plug icon to open API Connections. Set API to Chat Completion and Chat Completion Source to Custom (OpenAI-compatible). Paste https://api.unsavedai.com/v1 into Custom Endpoint (Base URL) and your key into Custom API Key. Use a key made just for SillyTavern so you can revoke it on its own.
Choose a model and context
Click Connect, then pick the model under Available Models or type it into Enter a Model ID: gemma-4-12b in every plan, gemma-4-31b with Priority. In AI Response Configuration (the sliders icon) set Context Size to 56,000 or less on either plan for gemma-4-12b or gemma-4-31b (61,440 including the reply). Character cards, lore, instructions and history all count toward it. Raise Max Response Length from the default 300 to 600–800, or longer replies stop mid-sentence.
Test a short conversation
Start with basic sampling settings and a short reply. Check streaming and stopping a response before loading a long conversation. Free writes one reply at a time and Priority two; extra messages wait their turn, and a 429 only appears when four more are already waiting.
Client reference: official documentation. Client settings can change between versions.
Next step
Get your values on Setup, check the models, or make a key in your account.