Overview > klyn.genai.llm.inference > LocalLLMProvider > complete
Version 0.2.2 Overview Index Tree Deprecated
klyn.genai.llm.inference.LocalLLMProvider.complete
method
public override complete(request as LLMRequest) as LLMResponse throws Exception
Description

Completes a typed text conversation using resident model weights. History is borrowed for this request and is not modified. Only an explicitly enabled system/schema prefix may be retained; user turns are not. Message history is rendered directly; tool schemas are validated as JSON data.

Parameters
ParameterDescription
requestOrdered textual messages ending in a user query or complete typed tool round, with an optional ToolRegistry. Images/attachments remain unsupported. Uses this provider's maxNewTokens bound.
Returns

Assistant message and exact prompt/completion token counts. The completion count includes EOS when generated; response text excludes it.

Throws
  • ValueException for null requests, unsupported content, a closed provider, a checkpoint without a tokenizer or invalid context bounds.
  • Exception for device or generation failures.
Example

response = provider.complete(LLMRequest([LLMThreadMessage.user("Hello")]))