Search
Search steps enable information retrieval from indexed data using different search strategies. QType provides both vector-based semantic search and traditional document search capabilities, allowing applications to find relevant information based on user queries.
Search steps integrate with Indexes to perform efficient retrieval operations and can be combined with other steps to build sophisticated question-answering and information retrieval workflows.
Rules and Behaviors
- Index Dependency: Search steps require a valid index reference to perform search operations
- Query Processing: Input queries are processed according to the search type (semantic embedding for vector search, text matching for document search)
- Result Ranking: Search results are automatically ranked by relevance score
- Configurable Limits: Number of returned results can be controlled via configuration parameters
- Type Safety: Search steps validate that the referenced index supports the requested search operation
- Empty Results: Search steps handle cases where no matching documents are found gracefully
- Similarity Thresholds: Vector search can filter results based on minimum similarity scores
Vector Search
Vector search performs semantic similarity matching using embeddings to find conceptually related content.
Component Definition
VectorSearch
Performs vector similarity search against a vector index.
- default_top_k (
int | None
): Number of top results to retrieve if not provided in the inputs.
Configuration
Key Properties
- index: Reference to a vector index containing embedded documents
- top_k: Maximum number of results to return (default: 10)
- similarity_threshold: Minimum similarity score for results (0.0-1.0)
- embedding_model: Optional override for query embedding generation
Document Search
Document search performs traditional text-based search using keyword matching and full-text search capabilities.
Component Definition
DocumentSearch
Performs document search against a document index.
Configuration
Key Properties
- index: Reference to a document index with full-text search capabilities
- max_results: Maximum number of documents to return
- search_fields: Specific fields to search within documents
- boost_factors: Relevance boosting for specific fields
- filters: Additional filtering criteria for search results
Related Concepts
Search steps work with Indexes for data storage and retrieval, may use Models for embedding generation in vector search, and integrate with Flows for complex information retrieval pipelines. Results are typically consumed by other Steps for further processing.