Skip to content

Setting Up a Connection

This guide walks you through creating your first connection to an AI provider.

Creating a Connection

  1. Open the Connection Manager panel (plug icon)
  2. Click New Connection
  3. Fill in the fields:

Required Fields

Field Description
Name A label for your reference (e.g., "Claude Sonnet 4.6," "GPT-5 Main")
Provider Select your AI provider from the dropdown

Optional Fields

Field Description
Model The model to use — you can type it or select from the model list
API URL Override the default endpoint (useful for proxies or self-hosted models)
API Key Your provider's API key
Preset A default preset to use with this connection
Default Whether this is your primary connection
  1. Click Save

Setting Your API Key

API keys are stored encrypted and never displayed after saving. To set or update a key:

  1. Open the connection in the Connection Manager
  2. Click Set API Key (or Update API Key if one already exists)
  3. Paste your key
  4. Save

The has_api_key indicator shows whether a key is stored without revealing the value.

Per-connection keys

Each connection has its own API key. There's no shared "provider key" — this lets you use different keys for different connections to the same provider (e.g., separate keys for personal and work use).

Testing Your Connection

After setting up a connection, verify it works:

  1. Click the Test button on the connection
  2. Lumiverse sends a simple request to your provider
  3. You'll see either a success message or an error describing what went wrong

Common issues:

  • Invalid API key — Double-check the key
  • Wrong API URL — Make sure the endpoint matches your provider
  • Model not found — Verify the model name is correct for your provider

Listing Available Models

Click the Models button on a connection to fetch the list of models available with your API key. This queries your provider directly, so the list is always up to date.

Multiple Connections

You can create as many connections as you want. Common setups:

  • Different providers — One OpenAI connection, one Anthropic connection
  • Different models — A fast model for casual chat, a powerful model for important scenes
  • Different keys — Separate billing or quota management
  • Self-hosted — A connection to your local LLM alongside cloud connections

Switch between connections by setting a different one as default, or select a specific connection when starting a generation.

Binding Reasoning Settings

Reasoning-capable models (Claude with extended thinking, OpenAI o-series, DeepSeek R1, Gemini thinking models) often want different reasoning depth on different connections — heavy thinking for your hero model, light thinking for a sidecar.

Each connection form has a Bind reasoning settings toggle. Turn it on, configure the reasoning options the way you want them for this connection, and Lumiverse stores a snapshot on the connection's metadata. Whenever you switch to that connection, the bound reasoning settings are restored automatically — even if your global reasoning settings were set differently.

Bindings are per-connection and survive across sessions. Switching to a connection with no binding leaves your current reasoning settings unchanged, so you don't lose tuning work on connections that aren't bound.

Bind a sidecar to low reasoning

Sidecar tasks (scene parsing, expression detection, council tools) usually don't need deep thinking. Bind your sidecar connection to a minimal reasoning level and your main connection to whatever you prefer for prose — Lumiverse switches automatically as features call the right connection.

Sidecar Connection

Some features (expression detection, council tools, image gen scene analysis) use a separate sidecar connection — a lightweight model for background tasks. Configure the sidecar in Settings > Advanced or in the Sidecar Settings section.

The sidecar keeps background processing costs low by using a smaller, cheaper model for auxiliary tasks while your main connection handles the primary generation.

Deleting a Connection

Deleting a connection also removes its stored API key from the encrypted secrets. This cannot be undone.