haive.core.engine.retriever.providers.MultiQueryRetrieverConfig

Multi-Query Retriever implementation for the Haive framework.

This module provides a configuration class for the Multi-Query retriever, which generates multiple query variations to improve retrieval coverage and find more relevant documents for complex or ambiguous queries.

The MultiQueryRetriever works by: 1. Using an LLM to generate multiple query variations from the original query 2. Running each generated query against the base retriever 3. Collecting and deduplicating all retrieved documents 4. Returning the combined set of unique documents

This retriever is particularly useful when: - Dealing with complex or ambiguous user queries - Need to improve recall by finding documents with different phrasings - User queries might miss relevant documents due to vocabulary mismatch - Building systems that need comprehensive document coverage

The implementation integrates with LangChain’s MultiQueryRetriever while providing a consistent Haive configuration interface with LLM integration.

Classes

MultiQueryRetrieverConfig

Configuration for Multi-Query retriever in the Haive framework.

Module Contents

class haive.core.engine.retriever.providers.MultiQueryRetrieverConfig.MultiQueryRetrieverConfig

Bases: haive.core.engine.retriever.retriever.BaseRetrieverConfig

Configuration for Multi-Query retriever in the Haive framework.

This retriever generates multiple query variations using an LLM to improve retrieval coverage and find more relevant documents for complex queries.

retriever_type

The type of retriever (always MULTI_QUERY).

Type:

RetrieverType

base_retriever

The underlying retriever to query with variations.

Type:

BaseRetrieverConfig

llm_config

LLM configuration for generating query variations.

Type:

AugLLMConfig

num_queries

Number of query variations to generate.

Type:

int

include_original

Whether to include the original query in the set.

Type:

bool

Examples

>>> from haive.core.engine.retriever import MultiQueryRetrieverConfig
>>> from haive.core.engine.retriever.providers.VectorStoreRetrieverConfig import VectorStoreRetrieverConfig
>>> from haive.core.engine.aug_llm import AugLLMConfig
>>>
>>> # Create base retriever and LLM config
>>> base_config = VectorStoreRetrieverConfig(name="base", vectorstore_config=vs_config)
>>> llm_config = AugLLMConfig(model_name="gpt-3.5-turbo", provider="openai")
>>>
>>> # Create multi-query retriever
>>> config = MultiQueryRetrieverConfig(
...     name="multi_query_retriever",
...     base_retriever=base_config,
...     llm_config=llm_config,
...     num_queries=3,
...     include_original=True
... )
>>>
>>> # Instantiate and use the retriever
>>> retriever = config.instantiate()
>>> docs = retriever.get_relevant_documents("machine learning algorithms")

get_input_fields()

Return input field definitions for Multi-Query retriever.

Return type:

dict[str, tuple[type, Any]]

get_output_fields()

Return output field definitions for Multi-Query retriever.

Return type:

dict[str, tuple[type, Any]]

instantiate()

Create a Multi-Query retriever from this configuration.

Returns:

Instantiated retriever ready for multi-query retrieval.

Return type:

MultiQueryRetriever

Raises:

• ImportError – If required packages are not available.

• ValueError – If configuration is invalid.

classmethod validate_num_queries(v)

Ensure reasonable number of queries.