feat(llm): add provider-agnostic parameter mapping system

[Pouyanpi]

Implements flexible LLM parameter transformation to support provider-specific naming conventions (e.g., max_tokens -> max_new_tokens for HuggingFace).

Key features:

• Automatic provider inference from LangChain module names
• Built-in mappings for HuggingFace and Google Vertex AI
• Custom parameter mapping support via config
• Parameter dropping capability (map to null)
• Integration with llm_call function

TODO:

• see if we can reduce the parameter transformation overhead by only doing it when parameter_mapping is given or provider is in the default ones. (nothing critical)

[Copilot]

Pull Request Overview

This PR implements a flexible parameter mapping system for LLM providers to handle parameter naming inconsistencies across different providers (e.g., transforming max_tokens to max_new_tokens for HuggingFace).

• Provider-agnostic parameter transformation with automatic provider inference from LangChain module names
• Built-in mappings for HuggingFace and Google Vertex AI with support for custom mappings
• Integration with the llm_call function and LLMRails configuration system

Reviewed Changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
nemoguardrails/llm/parameter_mapping.py	Core parameter mapping module with provider inference and transformation logic
nemoguardrails/rails/llm/config.py	Adds parameter_mapping field to Model configuration
nemoguardrails/rails/llm/llmrails.py	Registers parameter mappings during LLM initialization
nemoguardrails/actions/llm/utils.py	Integrates parameter transformation into llm_call function
tests/test_parameter_mapping.py	Comprehensive tests for parameter mapping functionality
tests/test_llmrails.py	Tests for parameter mapping registration in LLMRails
tests/test_llm_call_parameter_mapping.py	Tests for parameter mapping integration in llm_call

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}