There have been requests from the Solution Engineering team to have a set of components capable of of creating and running retrieval on multiple queries to improve retrieval.

We have these components in haystack-experimental which are QueryExpander, MultiQueryEmbeddingRetriever and MultiQueryTextRetriever with the an example workflow being

QueryExpander --> MultiQueryTextRetriever --> Retrieved Documents

Additional here is a code example

expander = QueryExpander(
    chat_generator=OpenAIChatGenerator(model="gpt-4.1-mini"), n_expansions=3, include_original_query=True
)
in_memory_retriever = InMemoryEmbeddingRetriever(document_store=document_store_with_embeddings)
query_embedder = SentenceTransformersTextEmbedder(model="sentence-transformers/all-MiniLM-L6-v2")
multiquery_retriever = MultiQueryEmbeddingRetriever(
    retriever=in_memory_retriever,
    query_embedder=query_embedder,
    max_workers=3
)

pipeline = Pipeline()
pipeline.add_component("query_expander", expander)
pipeline.add_component("multiquery_retriever", multiquery_retriever)
pipeline.connect("query_expander.queries", "multiquery_retriever.queries")

data = {
    "query_expander" : {"query": "green energy sources"},
    "multiquery_retriever": {"retriever_kwargs": {"top_k": 3}}
}
results = pipeline.run(data=data, include_outputs_from={"query_expander", "multiquery_retriever"})

Steps for Moving an Experiment to Haystack Core or Integrations

• Make sure the latest Haystack release or an integration release contains the merged experiment
• Update import statements in example cookbook, remove experimental tag from cookbook, etc.
• Close discussion in haystack-experimental with move information
• Remove pydocs
• Move experiment from active experiments in the catalog in haystack-experimental README.md to adopted experiments
• Remove example notebook from haystack-experimental if it exists
• Remove dependencies needed for the experiment from pyproject.toml
• Make sure to add a docs page for the new component(s)