Hybrid Search

Software development skill, available on Zeplik

Hybrid Search is a ready-to-run software development skill on Zeplik. Not for full RAG pipelines (use rag-implementation) or model/chunking choice (use embedding-strategies). Ask in plain language and Zeplik applies the skill's method for you inside the conversation, on whichever AI model you prefer.

The Hybrid Search skill loads automatically when your request matches it, or you can invoke it directly by typing /hybrid-search-implementation in any chat. It works with attachments, connectors, and any model that supports the task, so you get the same expert method every time without setting anything up.

What the Hybrid Search skill can do

Try these prompts on Zeplik

Pick a prompt to open it in the Zeplik app. If you are not signed in yet, your prompt is waiting for you the moment you do.

How the Hybrid Search skill works

/hybrid-search-implementation

Combine vector similarity and keyword-based search for better retrieval than either alone. The user describes their stack and failure cases (or pastes queries with bad results and their current retrieval code); deliver a fusion design and working implementation code as chat artifacts. For the end-to-end RAG system, use rag-implementation; for picking the embedding model or chunking, use embedding-strategies.

When to Use

  • Building retrieval with improved recall
  • Combining semantic understanding with exact matching
  • Handling queries with specific terms (names, codes, SKUs)
  • Improving search for domain-specific vocabulary
  • When pure vector search misses keyword matches

Core Concepts

1. Hybrid Search Architecture

Query --+--> Vector Search  --> Candidates --+
        |                                    |
        +--> Keyword Search --> Candidates --+--> Fusion --> Results

2. Fusion Methods

MethodDescriptionBest For
RRFReciprocal Rank FusionGeneral purpose
LinearWeighted sum of scoresTunable balance
Cross-encoderRerank with neural modelHighest quality
CascadeFilter then rerankEfficiency

RRF score for a document: sum over each ranker of 1 / (k + rank) with k typically 60. It needs no score normalization, which is why it is the default recommendation.

Implementation Workflow

  1. Confirm the stack: vector store (pgvector, Qdrant, Weaviate, Elasticsearch/OpenSearch) and keyword engine (BM25 in the same store or separate).
  2. Start with RRF over top-N (e.g., 50) candidates from each side; it works well without tuning.
  3. If quality demands more, add a cross-encoder reranker over the fused top 20-30.
  4. Tune empirically: build a small query set from the user's real failure cases, measure recall@k and MRR before and after.
  5. Deliver the retrieval function, fusion code, and an evaluation snippet as chat artifacts, in the user's language and client library.

Best Practices

Do's

  • Tune weights empirically -- test on your data
  • Use RRF for simplicity -- works well without tuning
  • Add reranking -- significant quality improvement
  • Log both scores -- helps with debugging
  • A/B test -- measure real user impact

Don'ts

  • Don't assume one size fits all -- different queries need different weights
  • Don't skip keyword search -- handles exact matches better
  • Don't over-fetch -- balance recall vs latency
  • Don't ignore edge cases -- empty results, single word queries

Templates and Detailed Worked Examples

The full template library (RRF and weighted fusion implementations, store-specific queries, reranker integration) lives in references/details.md. Read that file when you need the concrete templates.

Usage

/hybrid-search-implementation $ARGUMENTS

How to use the Hybrid Search skill

  1. Sign in to Zeplik

    Create a free Zeplik account or sign in. New accounts start with free credits, so you can try the Hybrid Search skill right away.

  2. Describe your software development task

    Ask in plain language, or type /hybrid-search-implementation to invoke the skill directly. Zeplik recognizes the Hybrid Search skill and applies its method.

  3. Review and refine the result

    Zeplik returns a clear, structured answer. Ask follow-ups in the same chat to refine it or take the next step.

Source and credit

Author
wshobson
License
MIT

Adapted from the open-source wshobson/agents project and tuned to run natively on Zeplik. View source on GitHub.

Frequently asked questions

What is the Hybrid Search skill?
Hybrid Search is a ready-to-run software development skill on Zeplik. Not for full RAG pipelines (use rag-implementation) or model/chunking choice (use embedding-strategies). Ask in plain language and Zeplik applies the skill's method for you inside the conversation, on whichever AI model you prefer.
How do I use Hybrid Search on Zeplik?
Sign in to Zeplik and ask in plain language, or type /hybrid-search-implementation in any chat to invoke it directly. The skill applies its method and returns a result you can refine in the same conversation.
Which AI model does the Hybrid Search skill use?
Any model you choose. Zeplik works across every model in one chat, so the Hybrid Search skill runs on your preferred model for the task.
Where does the Hybrid Search skill come from?
The Hybrid Search skill is adapted from the open-source wshobson/agents project (MIT) and tuned to run natively on Zeplik. The original source is linked on this page.
How much does the Hybrid Search skill cost?
Using the skill is free to start. You only spend Zeplik credits when the assistant runs, and new accounts begin with free credits.

Related software development skills

More on Zeplik

Try Hybrid Search on Zeplik

Every model, one chat. Bring the Hybrid Search skill into your next conversation and let the assistant do the work.

Browse all skills
Hybrid Search - Software development skill for Zeplik AI | Zeplik Chat