05

Quick reference

Bring your own API key🔗

Onboarding Step 4, then "Bring your own API key":

  1. Pick your backend and a model preset, or Custom.
  2. Paste your API key. Add the API base if the backend requires one.
  3. Pick at least one provider group.
  4. Click "Add provider".

Full details: Bring your own API key.

Bring your own hardware🔗

Onboarding Step 4, then "Your own LLM hardware":

  1. Create a tailnet at tailscale.com and join your model machine to it.
  2. Generate a reusable, expiring auth key with a dedicated tag such as tag:scirix-proxy.
  3. Enter the tailnet name and auth key, then click "Register tailnet & provision pool".
  4. Add the endpoint: name, 100.x tailnet address, port (11434 for Ollama).
  5. Describe the model: name, backend, model ID (the ollama_chat/ route prefix is added automatically if you leave it off), capabilities, and at least one provider group. Click "Add provider".

Full details: Bring your own hardware.

Capabilities🔗

  • Presets set them for you. Hardware and Custom providers need them checked by hand.
  • tool_use is required for agent roles, json_mode for audits, embedding for indexing.
  • Only check what the model truly supports.

Full details: Capabilities.

If things stop working🔗

  • Pages dead after idling: sign out and back in.
  • Timeouts: reload and retry.
  • Tasks failing: check capabilities and role assignments.
  • Nothing starting at all: check the health surfaces, then report it.

Full details: Troubleshooting.