Skip to content

Interface ZIF_LLM_OPTIONS

AI Generated documentation.

Overview

The zif_llm_options interface defines configuration methods for Large Language Model (LLM) call parameters. It provides setter methods for various model generation parameters with specific value ranges and validation:

  • set_temperature: Controls randomness (range 0-2)
  • set_max_tokens: Sets maximum token output (> 0)
  • set_top_p: Nucleus sampling parameter (range 0-1)
  • set_top_k: Top-k sampling parameter (range 1-infinity)
  • set_seed: Enables reproducible output generation
  • set_frequency_penalty: Reduces repetition of tokens (range -2 to 2)
  • set_presence_penalty: Encourages topic diversity (range -2 to 2)
  • set_min_p: Minimum probability threshold (range 0-1)
  • set_top_a: Alternative sampling parameter (range 0-1)
  • set_custom_parameters: Allows adding custom key-value parameters
  • get_paramters: Retrieves all configured parameters

Each setter method raises a zcx_llm_validation exception for invalid input values.

Dependencies

  • Custom exception class: zcx_llm_validation
  • Custom type: zllm_keyvalues (key-value list for parameters)