getLLMProvider

workflow.getLLMProvider() is a provider parsing tool. The LLM node is called when it is executed, and the business code can also be used directly in custom nodes or advanced scenarios.

When executor is configured with tokenUsage.enabled, the token usage is automatically reported when the provider parsed by getLLMProvider() is returned by chat() or when the stream finish chunk carries usage. workspace examples that call provider directly in custom nodes will also follow the same statistics link; provider without usage will not write statistics. The report will read the usage field of common provider, including cache read / cache write or cache creation token details.

Signature

workflow.getLLMProvider(
  context: NodeContext,
  nodeName: string,
  source?: LLMProviderSource,
): LLMProvider

Parameter

Parameter	Type	Description
context	NodeContext	Read provider, binding, and default provider from context.providers.
nodeName	string	Current node name; used for node-level binding and error messages.
source	LLMProviderSource	Optional provider source: instance, id, ref, or resolver.

ProviderRegistry

Field	Type	Description
llm	LLMProvider | undefined	A compatible default single provider.
llmProviders	object | undefined	provider id to provider instance mapping.
llmProviderBindings	object | undefined	Node name to provider/model binding mapping.
defaultLLMProviderBinding	LLMProviderBinding | undefined	Default provider/model binding.
defaultLLMProviderId	string | undefined	the default provider id.

LLMProviderSource

Form	Type	Description
provider examples	LLMProvider	Directly back.
provider id	string	Search from llmProviders.
provider ref	LLMProviderRef	Check ref.id first and use ref.fallback if necessary.
resolver	(context, nodeName) => LLMProvider | undefined	Custom parsing.

Parsing order

Order	Source	Description
1	Node Binding	Use context.providers.llmProviderBindings[nodeName] first.
2	source	The provider source declared using the node.
3	Default Binding	Use defaultLLMProviderBinding.
4	default id	Use defaultLLMProviderId to find from llmProviders.
5	Compatible with default provider	Use context.providers.llm directly.
6	ref fallback	Only try when source is ref.

Return value

Return	Type	Description
provider	LLMProvider	contains name, chat(input), stream(input).

Error

Situation	Error Type	Description
Bound provider/model not found	provider_call	Error reporting will include providerId, modelId and nodeName.
Failed to parse all sources	provider_call	The current node requires an LLM provider, but no provider is available in the context.

Example

const provider = workflow.getLLMProvider(
  context,
  "answer",
  workflow.createLLMProviderRef("default"),
);

const output = await provider.chat({ prompt: input.message });