zaclys/searxng
1
0
Fork 1
mirror of https://github.com/searxng/searxng synced 2024-01-01 19:24:07 +01:00
Code Issues Projects Releases Packages Wiki Activity

o

Browse source
This commit is contained in:
Joseph Cheung 2023-02-23 17:02:57 +08:00
parent b8557e9f99
commit 8ae80a3b73
4 changed files with 10 additions and 13 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

2
Dockerfile
View file

@ -45,8 +45,6 @@ RUN apk add --no-cache -t build-dependencies \
uwsgi \
uwsgi-python3 \
brotli \
rust \
cargo \
&& pip3 install --no-cache -r requirements.txt \
&& apk del build-dependencies \
&& rm -rf /root/.cache

1
requirements-dev.txt
View file

@ -22,4 +22,3 @@ wlc==1.13
coloredlogs==15.0.1
requests
markdown
transformers

1
requirements.txt
View file

@ -18,4 +18,3 @@ typing_extensions==4.5.0
fasttext-predict==0.9.2.1
requests
markdown
transformers

15
searx/webapp.py
View file

@ -17,7 +17,6 @@ import requests
import markdown
import re
import datetime
from transformers import GPT2TokenizerFast
from timeit import default_timer
from html import escape
@ -769,8 +768,10 @@ def search():
res['content'] = res['content'].replace("This Tweet was deleted by the Tweet author.","Deleted Tweet.")
tmp_prompt = res['title'] +'\n'+ res['content'] + '\n' + new_url +'\n'
tokenizer = GPT2TokenizerFast.from_pretrained("gpt2")
if len( tokenizer(prompt + tmp_prompt +'\n' + "\n以上是问题 " + original_search_query + " 的搜索结果，删除与问题相关度低的内容，用简体中文分条总结简报，在文中用(链接)标注对应内容来源链接，不要把链接都放在最后。结果：")['input_ids'] )<2990:
if original_search_query == search_query.query and len( prompt + tmp_prompt +'\n' + "\n以上是关键词 " + original_search_query + " 的搜索结果，删除无关内容，用简体中文分条总结简报，在文中用(链接)标注对应内容来源链接，链接不要放在最后。结果：" ) <2500:
prompt += tmp_prompt +'\n'
if len( prompt + tmp_prompt +'\n' + "\n以上是任务 " + original_search_query + " 的网络知识。用简体中文完成任务，如果使用了网络知识，删除无关内容，在文中用(链接)标注对应内容来源链接，链接不要放在最后。结果：") <2500:
prompt += tmp_prompt +'\n'
if prompt != "":
gpt = ""
@ -782,7 +783,7 @@ def search():
}
if original_search_query != search_query.query:
gpt_data = {
"prompt": prompt+"\n以上是问题 " + original_search_query + " 的搜索结果，删除与问题相关度低的内容，用简体中文分条总结简报，在文中用(链接)标注对应内容来源链接，不要把链接都放在最后。结果：",
"prompt": prompt+"\n以上是任务 " + original_search_query + " 的网络知识。用简体中文完成任务，如果使用了网络知识，删除无关内容，在文中用(链接)标注对应内容来源链接，链接不要放在最后。结果：",
"max_tokens": 1000,
"temperature": 0.2,
"top_p": 1,
@ -795,7 +796,7 @@ def search():
}
else:
gpt_data = {
"prompt": prompt+"\n以上是关键词 " + search_query.query + " 的搜索结果，删除与关键词相关度低的内容，用简体中文分条总结简报，在文中用(链接)标注对应内容来源链接，不要把链接都放在最后。结果：",
"prompt": prompt+"\n以上是关键词 " + search_query.query + " 的搜索结果，删除无关内容，用简体中文分条总结简报，在文中用(链接)标注对应内容来源链接，链接不要放在最后。结果：",
"max_tokens": 1000,
"temperature": 0.2,
"top_p": 1,
@ -844,7 +845,7 @@ def search():
background: rgba(211, 173, 209, 0.4);
}
</style>
''' + '<div id="prompt" style="display:none">' + gpt + '</div>'
''' + '<div id="prompt" style="display:none">' + str(base64.b64encode(gpt.encode("utf-8"))) + '</div>'
# gpt_response = requests.post(gpt_url, headers=gpt_headers, data=json.dumps(gpt_data))
# gpt_json = gpt_response.json()
# if 'choices' in gpt_json:
@ -1013,7 +1014,7 @@ const headers = {
"Content-Type": "application/json",
Authorization: "Bearer '''+ os.environ['GPTKEY'] + r'''",
};
let prompt = JSON.parse(document.querySelector("#prompt").textContent);
let prompt = JSON.parse(atob(document.querySelector("#prompt").textContent));
chatTextRawIntro = "";
text_offset = -1;