Query Decomposition FAQs

Question 1

What is query decomposition in the context of RAG?

Accepted Answer

Query decomposition is the process of breaking a single, complex user query into multiple simpler search terms. This ensures that the retrieval system finds relevant documents for every part of a question rather than just the most prominent keywords.

Question 2

Can this skill be combined with HyDE?

Accepted Answer

Yes, this skill is designed to work with Hypothetical Document Embeddings (HyDE) by generating a hypothetical answer for each decomposed concept, further bridging the vocabulary gap between queries and documents.

Question 3

When should I use the Query Decomposition skill?

Accepted Answer

This skill is best used when your AI application needs to handle 'and/or' queries, comparisons (X vs Y), or impact analysis questions where the relevant information might be spread across different documents.

Question 4

Does query decomposition increase application latency?

Accepted Answer

While it adds a small overhead for the initial decomposition step, the use of fast-path heuristics and parallelized sub-query searches ensures that the impact on total response time is minimized.

Question 5

What is Reciprocal Rank Fusion (RRF)?

Accepted Answer

RRF is a parameter-free algorithm used to combine the results from multiple search lists into a single ranked list, prioritizing documents that appear consistently and highly across different searches.

Query Decomposition

Query Decomposition

Key Features

Use Cases

Key Features

Use Cases