API Reference

All types use the ERP_ / UERP_ / FERP_ prefix and are Blueprint-accessible.

UERP_LLMComponent

The Blueprint-friendly entry point. Add it to an NPC or controller.

Member	Kind	Description
SystemPrompt	property FString	Persona / instructions
GenerationConfig	property FERP_LLMGenerationConfig	Sampling params
bAutoInitialize	property bool	Create context on BeginPlay
SendMessage(UserMessage)	function → bool	Add user message + generate
PushText(Text)	function	Pipeline sink (same as SendMessage)
GenerateContinuation()	function → bool	Continue without new input
ClearHistory()	function	Reset conversation (keeps system prompt)
AddMessage(Message)	function	Pre-load a FERP_LLMMessage
SetSystemPromptRuntime(Text)	function	Change persona at runtime
IsLLMReady()	pure → bool	Backend loaded?
OnGenerationComplete	event FERP_LLMResult	Full result + metrics
OnText	event FString	Generated reply text — for chaining
OnTokenGenerated	event FString	Per-token (when streaming)

FERP_LLMGenerationConfig

Field	Type	Default
MaxTokens	int32	256
Temperature	float	0.7
TopP	float	0.9
TopK	int32	40
RepetitionPenalty	float	1.1
StopSequences	TArray<FString>	—
bEnableStreaming	bool	false

FERP_LLMMessage / FERP_LLMResult

FERP_LLMMessage = { Role (System/User/Assistant), Content, Metadata }. FERP_LLMResult = { GeneratedText, TokensGenerated, bCompleted, StopReason, ProcessingTimeMs, TokensPerSecond }.

UERP_LLMSubsystem

GameInstanceSubsystem owning the shared model.

Member	Description
InitializeBackend(Config) → bool	Load the GGUF model
IsBackendInitialized() → bool	Ready state
GetModelInfo() → FString	Loaded model description
GetMaxContextLength() → int32	Context window size
CountTokens(Text) → int32	Tokenize for budgeting

IERP_LLMBackend

Implement to add a custom inference engine, then select Custom in settings. UERP_LlamaCppBackend is the bundled implementation.