API ReferenceCollections

Search Advanced

Advanced search with comprehensive filtering and options.

This endpoint supports:

  • Metadata filtering using Qdrant’s native filter syntax
  • Pagination with offset and limit
  • Score threshold filtering
  • Query expansion strategies (default: AUTO, generates up to 4 variations)
  • Automatic filter extraction from natural language (default: ON)
  • LLM-based result reranking (default: ON)

Default behavior:

  • Query expansion: ON (AUTO strategy)
  • Query interpretation: ON (extracts filters from natural language)
  • Reranking: ON (improves relevance using LLM)
  • Score threshold: None (no filtering)

To disable features, explicitly set:

  • enable_reranking: false
  • enable_query_interpretation: false
  • expansion_strategy: “no_expansion”

Path parameters

readable_idstringRequired
The unique readable identifier of the collection to search

Headers

x-api-keystringRequired

Request

This endpoint expects an object.
querystringRequired>=1 character<=1000 characters
The search query text
filterobject or nullOptional

Qdrant native filter for metadata-based filtering

offsetinteger or nullOptional>=0Defaults to 0

Number of results to skip (DEFAULT: 0)

limitinteger or nullOptional>=1<=1000Defaults to 20

Maximum number of results to return (DEFAULT: 20)

score_thresholddouble or nullOptional>=0<=1

Minimum similarity score threshold (DEFAULT: None - no filtering)

response_typeenumOptional

Type of response - ‘raw’ or ‘completion’ (DEFAULT: ‘raw’)

Allowed values:
search_methodenum or nullOptional

Search method to use (DEFAULT: ‘hybrid’ - combines neural + BM25)

Allowed values:
recency_biasdouble or nullOptional>=0<=1

How much document age penalizes the similarity score (0..1). 0 = no age penalty (pure similarity); 0.5 = old docs lose up to 50% of their score; 1 = old docs get zero score (pure recency). Applied as: score × (1 - bias + bias × age_factor). Works within top ~10,000 semantic matches. DEFAULT: 0.3

expansion_strategyenum or nullOptional

Query expansion strategy (DEFAULT: ‘auto’ - generates up to 4 query variations). Options: ‘auto’, ‘llm’, ‘no_expansion’

Allowed values:
enable_rerankingboolean or nullOptional

Enable LLM-based reranking to improve result relevance (DEFAULT: True - enabled, set to False to disable)

enable_query_interpretationboolean or nullOptional

Enable automatic filter extraction from natural language query (DEFAULT: True - enabled, set to False to disable)

Response

Successful Response
resultslist of maps from strings to any
Array of search result objects containing the found documents, records, or data entities.
response_typeenum

Indicates whether results are raw search matches or AI-generated completions based on the found content.

Allowed values:
statusenum

Status of the search operation indicating the quality and availability of results:
• success: Search found relevant results matching your query
• no_relevant_results: Search completed but found no sufficiently relevant matches
• no_results: Search found no results at all, possibly indicating empty collections or very specific queries

Allowed values:
completionstring or null

AI-generated natural language answer when response_type is ‘completion’. This provides natural language answers to your query based on the content found across your connected data sources.

Errors