Model provider connection
Use this checklist when a model provider doesn’t answer. It applies to Rhizome’s own engine. Prime handles its own providers; see Chat doesn’t answer.
Local providers
Section titled “Local providers”For Ollama or LM Studio:
- Start the local model server.
- Check that the base URL in Settings → AI Agents → Local model matches the server.
- Check that the model ID is installed and loaded.
- Click Test model again.
Hosted providers
Section titled “Hosted providers”- Check the provider kind and endpoint.
- Check that the model ID exists for your account.
- Check the key. With Save locally in Rhizome, the key is in the system keychain. With Read from env var, the variable must be set for the app, not only in your terminal.
- Click Test model. It sends one short request with no tools.
Never store keys in the vault. Vault notes are committed and can be pushed.
Keychain problems
Section titled “Keychain problems”If a Free tier provider row says Keychain error, the system keychain is locked or Rhizome isn’t allowed to use it. Unlock it or allow access, then reopen Settings. Keys from older versions move to the keychain the first time they’re read; don’t edit a key file by hand.
Taproot can’t route
Section titled “Taproot can’t route”If no free provider has a key, Settings says “Add a key to at least one provider to use Taproot.” Turn on a provider and paste its key. With Strict mode on, only Groq and your own endpoint are used. See Free models and routing.
No file tools
Section titled “No file tools”Rhizome’s engine can read, search, edit, and create notes in the open vault. See Agents, engines and models.