Skip to main content
Outside of using Aryn’s agentic query engine, you can run direct search queries on your DocSet. Aryn is built on OpenSearch, and leverages its keyword and vector indexing for this feature set. You can easily add search capabilities to your application using the Aryn SDK. To learn more about using the Aryn UI for ad hoc searches, click here.

​
Running a search query

You can use the search function in the Aryn SDK to run a search query on your DocSet: 
from aryn_sdk.client import Client
from aryn_sdk.types.search import SearchRequest

myClient = Client()
client_response = myClient.search(
    docset_id="aryn:ds-031ppkfhkzq42baiunmr8yh",
    query=SearchRequest(
        query="weather",
        properties_filter="""(properties.entity.aircraft like "Cessna") and (properties.entity.operator <> "n/a") and (properties.entity.lowestCloudCondition = "Clear")""",
        query_type="hybrid",
        return_type="element",
    ),
)

search_results = client_response.value.results

for result in search_results:
    print(result)
It follows the general form: 
from aryn_sdk.client import Client
from aryn_sdk.types.search import SearchRequest

myClient = Client()
results = myClient.search(
    docset_id=[your-docset-ID],
    query=SearchRequest(
        query=[your-query-phrase],
        properties_filter=[your-boolean-filter],
        query_type=[KEYWORD | VECTOR | HYBRID | LEXICAL],
        return_type=[DOC | ELEMENT],
    ),
)
You can specify a properties filter, a query, or both.

​
Query Type

You can choose different ways to search your DocSet:
  • Hybrid: Combines keyword and vector search.
  • Vector (Semantic): Finds similar items in a dataset by comparing their numerical representations (vector embeddings) using a k-NN algorithm.
  • Lexical: Uses the BM25 algorithm, a ranking function used by search engines to estimate the relevance of documents to a given search query. The search call will perform an exact string match and return results where the query string shows up as a standalone word.
  • Keyword: The search call will perform a substring match and return results that contain strings that contain the query term specified.

​
Properties Filter

You can filter your search results using your DocSet’s Properties. The filter is a boolean expression, specifying the condition to use. It is a string consisting of conditions separated by AND keywords. The following are a few examples: 
"""(properties.entity.company_name = "Microsoft") AND (properties.entity.year = "2024")"""
"""(properties.entity.company_ticker = "AMZN") AND (properties.entity.year = "2023") AND (properties.entity.quarter = "Q2")"""
"""(properties.path = "report-9999.pdf")"""  # Where the file was loaded from
"""(properties.entity.aircraft like "Cessna")"""

​
Operators

The following operators are supported:
  • =
  • >
  • >=
  • <
  • <=
  • <> (not equal to)
  • like

​
Return Type

You can return Document chunks (elements) with element or full documents with doc.
⌘I