From 8dec5624aa79393af9fe19d4f5137131040fdb9d Mon Sep 17 00:00:00 2001 From: administrator Date: Sun, 2 Apr 2023 09:31:48 +0800 Subject: [PATCH] =?UTF-8?q?qdrant=E6=95=B0=E6=8D=AE=E5=BA=93?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- gpt_2_create_question.py | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/gpt_2_create_question.py b/gpt_2_create_question.py index 85f56fd..bee67b1 100644 --- a/gpt_2_create_question.py +++ b/gpt_2_create_question.py @@ -1,4 +1,3 @@ -import numpy as np import openai from gpt_0_basic_info import api_key from gpt_0_create_qdrant import qdrant_url, collection_name @@ -28,6 +27,7 @@ def get_query_similarity(input_query): # 从Qdrant中搜索与query_vector最相似的两个向量 search_results = client.search(collection_name, query_vector, limit=2) + # 找到vector最接近的两个QandA two_largest = [] for result in search_results: @@ -37,6 +37,9 @@ def get_query_similarity(input_query): # [{'similarities': 0.87828124, 'QandA': '当有人问：亁颐堂是做什么的, 请回答：亁颐堂是一个网络培训公司'}, # {'similarities': 0.812168, 'QandA': '当有人问：公司名称, 请回答：亁颐堂科技有限责任公司'}] + # 如果最相似的two_largest[0]['similarities']都小于0.8，那么就返回空字符串 + # 如果第二相似的two_largest[1]['similarities']小于0.8，并且拼接后长度大于1500，那么就返回two_largest[0]['QandA'] + # 如果第二个相似的two_largest[1]['similarities']大于0.8，那么就返回两个拼接后的字符串 context = '' if two_largest[0]['similarities'] < 0.8 else two_largest[0]['QandA'] \ if (two_largest[1]['similarities'] < 0.8 or (len(two_largest[1]['QandA'] + '\n' + two_largest[0]['QandA']) >= max_context_len)) \ else (two_largest[1]['QandA'] + '\n' + two_largest[0]['QandA'])