Discussion on QueryExpander Component

[bilgeyucel]

This is the discussion board for QueryExpander component.

We introduced the QueryExpander component, which generates a list of semantically similar queries to the user query to improve retrieval recall in RAG.

Usage example

from haystack.components.generators.chat.openai import OpenAIChatGenerator
from haystack_experimental.components.query import QueryExpander

expander = QueryExpander(
    chat_generator=OpenAIChatGenerator(model="gpt-4.1-mini"),
    n_expansions=3
)

result = expander.run(query="green energy sources")
print(result["queries"])
>>> ['alternative query 1', 'alternative query 2', 'alternative query 3', 'green energy sources']

Try it out and let us know what you think in the comments 👇

Note: Experimental features live in this repository for a fixed period of time. We don't guarantee that we will continue maintaining experimental features. But if they are successful, stable and if you like it, we will move the feature to the core Haystack package.

[onestardao]

🔥 This is a super neat and clean implementation. Query expansion is usually treated like the “quiet middle child” of RAG, but honestly—it’s often the key to stability in weak-context setups.

We’ve been experimenting with something a bit sideways to this: instead of just generating parallel queries, we use semantic shifts to simulate contradiction, missing premise, or reframing — basically like a Socratic query wiggler.

For example, from "green energy sources" we’d generate:

• "Why might green energy be insufficient?"
• "What if fossil fuels are still needed?"
• "Compare carbon offset and carbon capture as policies."

It acts almost like a hallucination vaccine — you prime the retriever to defend, not just explain.

If you're interested I can share the logic layer — we just open-sourced it in this PDF:
📄 WFGY Semantic Reasoning Engine

Would love to know if you’ve tried adversarial-style expansions in this QueryExpander pipeline!