Allow setting query transformers in BaseRAGQA (#67)

[Ch-Abhinav-Chowdary]

Summary

• Adds an optional query_transformer_prompt parameter to BaseRAGQuestionAnswerer to support query rewriting before retrieval.
• Accepts any pw.UDF (e.g., prompts.prompt_query_rewrite , prompts.prompt_query_rewrite_hyde ).
• Defaults to None (no transformation) to preserve current behavior.

Motivation

• Query rewriting can significantly improve recall and relevance in RAG systems.
• The codebase already includes query-rewrite prompts ( prompt_query_rewrite , prompt_query_rewrite_hyde ); this change enables using them directly in the base RAG pipeline without custom plumbing.
• Keeps the default path unchanged for users who do not need rewriting.

Key Changes

• python\pathway\xpacks\llm\question_answering.py

  • BaseRAGQuestionAnswerer.init : added query_transformer_prompt: pw.UDF | None = None .
  • Stored self.query_transformer_prompt .
  • answer_query : if query_transformer_prompt is set, constructs a rewrite prompt and calls the configured LLM to produce search_query ; otherwise, uses the original prompt as search_query . Retrieval now uses search_query .
  • Updated the class docstring to document the new parameter and behavior.

• python\pathway\xpacks\llm\tests\test_rag.py

  • Added test_base_rag_with_query_transformer_identity to validate compatibility by using an identity rewrite UDF and confirming behavior remains unchanged.

[CLAassistant]

All committers have signed the CLA.

[szymondudycz]

python\pathway\xpacks\llm\tests\test_rag.py

Added test_base_rag_with_query_transformer_identity to validate compatibility by >using an identity rewrite UDF and confirming behavior remains unchanged.

I don't see this file in the changed files

[szymondudycz]

Please update the docstring of this class with the description of the newly added parameter

[Ch-Abhinav-Chowdary]

Okay i will look upon it

[zxqfd555]

@Ch-Abhinav-Chowdary, could you please address the comments so that we can proceed with merging your PR? Thanks!