Hybrid Search

1. Plain-Language Definition

Hybrid search combines dense vector search (which understands semantic meaning) with sparse keyword search (which looks for exact words like BM25).

2. Why It Exists

Different queries require different search strategies. A query like 'How do we handle financial loss?' requires semantic understanding, while 'Find ID #8492-B' requires exact keyword matching.

3. What Breaks Without It

Pure vector search often fails at exact keyword matches (acronyms, serial numbers, names). Pure keyword search fails if the user uses synonyms instead of the exact words in the text.

4. How StarkLLM Implements It

StarkLLM queries the local database using both methods simultaneously, retrieves the top candidates from both, and merges them to provide the LLM with the most comprehensive context.

5. User-Visible Benefit

You don't have to compromise. You get the deep conceptual understanding of AI combined with the precise, exact recall of a traditional search engine.

6. Practical Example

Searching for 'NASA budget' will retrieve documents mentioning 'National Aeronautics and Space Administration funding' (semantic) as well as exact mentions of the acronym 'NASA' (keyword).