Class Introduction

Function

Connects to the TEI service and calculates relevance scores for documents. It inherits the abstract class Reranker.

Prototype

from mx_rag.reranker.service import TEIReranker
TEIReranker(url,  k, client_param)

Parameters

Parameter	Data Type	Required/Optional	Description
url	String	Required	TEI re-ranking service address. The string contains [1, 128] characters. The /rerank and /v1/rerank APIs are supported. NOTE: Currently, the re-ranking service created based on the TEI framework does not support the HTTPS protocol. For security purposes, you can set up an Nginx service to ensure that the service and re-ranking service remain within the same trusted network. When the client accesses the Nginx in HTTPS mode, the Nginx forwards a request to the re-ranking service.
k	Integer	Optional	The most relevant k results after re-ranking. The value range is [1, 10000]. The default value is 1.
client_param	ClientParam	Optional	HTTPS client configuration parameter. The default value is ClientParam(). For details, see ClientParam.

Return Value

TEIReranker object.

Example

from paddle.base import libpaddle
from mx_rag.reranker.service import TEIReranker
from mx_rag.utils import ClientParam
# Same as LocalReranker(url="https://ip:port/rerank", client_param=ClientParam(xxx)).
rerank = TEIReranker.create(url="https://ip:port/rerank",
                            client_param=ClientParam(ca_file="/path/to/ca.crt"))
docs = ['I am Xiao Hong','I am Xiao Ming']
scores = rerank.rerank ('Hello', docs)
res = rerank.rerank_top_k(docs, scores)
print(res)

Parent topic: TEIReranker