createLLMProviderRef

steamharness.createLLMProviderRef() Creates a provider reference object. When the LLM node is executed, the actual provider is resolved by the getLLMProvider node name, provider id, binding table, and fallback.

Signature

steamharness.createLLMProviderRef(
  id: string,
  fallback?: LLMProviderSource,
): LLMProviderRef

Parameter

Parameter	Type	Description
id	string	provider id, used to find provider from context.providers.llmProviders.
fallback	LLMProviderSource	Optional fallback; continue parsing when id is not found.

Return value

Field	Type	Description
kind	"llm-provider-ref"	Identifies that this is a provider ref.
id	string	Provider ID.
fallback	LLMProviderSource | undefined	Fallback provider source.

Provider Parse Order

Order	Source	Description
1	context.providers.llmProviderBindings[nodeName]	Node-level binding takes precedence.
2	source itself	If source is a provider, string, ref, or resolver, it is parsed first.
3	defaultLLMProviderBinding / defaultLLMProviderId	Use the default binding or default provider id.
4	context.providers.llm	The default provider of compatible orders.
5	ref fallback	The last attempt to fallback when ref cannot find id.

Analysis process

Phase	Input	Result
Workflow definition	id, fallback	LLMProviderRef
LLM node execution	LLMProviderRef, NodeContext, nodeName	LLMProvider
Provider call	LLMProvider, LLMChatInput	LLM output

Example

const defaultProvider = steamharness.createLLMProviderRef("default", () => {
  return steamharness.createLLMProviderFromConfig();
});

const llmNode = steamharness.createLLMNode({
  provider: defaultProvider,
  mapInput(input) {
    return { prompt: input.message };
  },
});

Common Scenarios

Scene	Method
Local Default provider	createLLMProviderRef("default", () => steamharness.createLLMProviderFromConfig())
server per-node binding model	Use the node name to configure in llmProviderBindings.
Multi-provider switching	Pass a different provider id, or use resolver.