RAG Search API

The RAG Search was introduced to look for passages that are relevant to the requested query. It is only available on enabling the “Enable for RAG” switch during the Collection creation.

URL

https://localhost:8443/rest/v2/api/rag-search

Method

GET

params

Below given are the basic parameters to use in RAG Search API

param	description
query	search query
cname	name of the Collection
col	collection id
searchtype	can be keyword/vector/hybrid
passages	Number of results to display on response. Integer value/Numbers
v.weight	vector weight in double value.
v.threshold	vector threshold in double value.

Header

Header	Description
X-Correlation-ID	A unique identifier that can be sent through the search API request. The X-Correlation-IDs will be logged in ../webapps/ROOT/logs/query.log and can also be viewed in exported query logs file from SearchBlox Analytics.

Response

The JSON response from SearchBlox by default

Example For RAG Search API

URL with Parameters : https://localhost:8443/rest/v2/api/rag-search?passages=10&public=true&query=tools&searchtype=hybrid&v.threshold=0.7&v.weight=0.2

METHOD : GET

Response :

{
    "query": "tools",
    "end": 10,
    "time": 215.0,
    "result": [
        {
            "no": 1,
            "score": 6.1731944,
            "uid": "b8d31bf6df153a16bb9483a11e180fa4",
            "col": "7",
            "colName": "amazon",
            "passage": "Motivation 25 Industry standard and newly developed design tools, decision-making tools, and rating tools helped to evaluate and benchmark existing and future performance. II. PROJECT DEVELOPMENT PROCESS 3) A variety of tools were used and developed to triangulate to the best answer. Design Tools Decision-Making Tools Rating Tools III.",
            "indexDate": "2024-12-02T14:52:40.214Z",
            "isStructured": false,
            "contentType": "PDF",
            "url": "https://searchblox-downloads.s3.amazonaws.com/testdata/allfileformats/ESBOverviewDeck.pdf"
        },
        {...}
    ],
    "pageSize": 10,
    "vectorWeight": 0.2,
    "searchType": "hybrid",
    "maxScore": 6.1731944,
    "passages": 10,
    "vectorThreshold": 0.7,
    "lastPage": 1,
    "currentPage": 1,
    "hits": 10,
    "start": 1
}

Terms	Description
score	Double value out of 10, for the generated result.
uid	Unique ID of the url
col	Collection ID
colName	Name of the collection
indexDate	Date on when the document/url was indexed.
isStructured	Collections like CSV, Database, MongoDB are of structured data and other Collections like WEB, Dynamic AUTO, SharepointOnline and others are considered as unstructured. For example if it is stored in any table or can be stored in databases, they are considered as structured else not.
contentType	Type/Format of the document like PDF, HTML and so on.
url	indexed url