Embabel Agent Documentation/com.embabel.common.ai.model/ConfigurableModelProviderProperties

ConfigurableModelProviderProperties

data class ConfigurableModelProviderProperties(var llms: Map<String, String> = emptyMap(), var embeddingServices: Map<String, String> = emptyMap(), var defaultLlm: String = "gpt-4.1-mini", var defaultEmbeddingModel: String? = null, var roles: Map<String, Map<String, <Error class: unknown class>>> = emptyMap(), var credentialServiceCacheSize: Int = 500)

Configuration properties for the model provider

Constructors

constructor(llms: Map<String, String> = emptyMap(), embeddingServices: Map<String, String> = emptyMap(), defaultLlm: String = "gpt-4.1-mini", defaultEmbeddingModel: String? = null, roles: Map<String, Map<String, <Error class: unknown class>>> = emptyMap(), credentialServiceCacheSize: Int = 500)

Properties

credentialServiceCacheSize

var credentialServiceCacheSize: Int

Upper bound on LLM services built from user-supplied keys and held for reuse.

defaultEmbeddingModel

var defaultEmbeddingModel: String?

Default embedding model name. Must be an embedding model name. Need not be set, in which case it defaults to null.

defaultLlm

var defaultLlm: String

Default LLM name. Must be an LLM name. It's good practice to override this in configuration.

embeddingServices

var embeddingServices: Map<String, String>

Map of role to embedding service name. May not include the default embedding service.

llms

var llms: Map<String, String>

Map of role to LLM name. Each entry will require an LLM to be registered with the same name. May not include the default LLM.

roles

var roles: Map<String, Map<String, <Error class: unknown class>>>

Map of role to provider to options, for deployments whose provider is not fixed - a bring-your-own-key application, or one configured for failover across providers.

Functions

allWellKnownEmbeddingServiceNames

fun allWellKnownEmbeddingServiceNames(): Set<String>

The embedding counterpart of allWellKnownLlmNames. Shorter because embedding roles have no nested per-provider shape: a role maps to one name, and there is default-embedding-model.

allWellKnownLlmNames

fun allWellKnownLlmNames(): Set<String>

Every LLM name this configuration mentions, across all three places one can appear: the flat llms role map, the nested roles map (one model per provider), and default-llm.