Create es_gpt.py
Browse files
es_gpt.py
ADDED
@@ -0,0 +1,75 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
from elasticsearch import Elasticsearch
|
2 |
+
import os
|
3 |
+
import json
|
4 |
+
import requests
|
5 |
+
|
6 |
+
ES_URL = os.environ["ES_URL"]
|
7 |
+
ES_USER = os.environ["ES_USER"]
|
8 |
+
ES_PASS = os.environ["ES_PASS"]
|
9 |
+
ES_CA_CERT = os.environ["ES_CA_CERT"]
|
10 |
+
|
11 |
+
|
12 |
+
class ESGPT:
|
13 |
+
def __init__(self, index_name):
|
14 |
+
self.es = Elasticsearch(ES_URL, http_auth=(ES_USER, ES_PASS),
|
15 |
+
ca_certs=ES_CA_CERT, verify_certs=True)
|
16 |
+
self.index_name = index_name
|
17 |
+
self.model_engine = os.environ["OPENAI_GPT_ENGINE"]
|
18 |
+
self.api_key = os.environ["OPENAI_API_KEY"]
|
19 |
+
|
20 |
+
def index(self, doc_id, doc):
|
21 |
+
self.es.index(index=self.index_name,
|
22 |
+
id=doc_id,
|
23 |
+
document=doc)
|
24 |
+
|
25 |
+
def search(self, query):
|
26 |
+
body = {
|
27 |
+
"query": {
|
28 |
+
"query_string": {"query": query}
|
29 |
+
}
|
30 |
+
}
|
31 |
+
|
32 |
+
results = self.es.search(index=self.index_name, body=body)
|
33 |
+
return results['hits']['hits']
|
34 |
+
|
35 |
+
def _paper_results_to_text(self, results):
|
36 |
+
text_result = ""
|
37 |
+
for paper in results:
|
38 |
+
title = ""
|
39 |
+
if "title" in paper["_source"]:
|
40 |
+
title = paper["_source"]["title"]
|
41 |
+
|
42 |
+
abstract = ""
|
43 |
+
if "abctract" in paper["_source"]:
|
44 |
+
abstract = paper["_source"]["abstract"]
|
45 |
+
|
46 |
+
paper_str = f"{title}:\n{abstract[:100]}\n\n"
|
47 |
+
text_result += paper_str
|
48 |
+
return text_result
|
49 |
+
|
50 |
+
def summarize(self, query, results):
|
51 |
+
# Generate summaries for each search result
|
52 |
+
result_json_str = self._paper_results_to_text(results)
|
53 |
+
if result_json_str == "":
|
54 |
+
result_json_str = "No results found"
|
55 |
+
|
56 |
+
print(result_json_str[:500])
|
57 |
+
|
58 |
+
body = {
|
59 |
+
"model": self.model_engine,
|
60 |
+
"prompt": f"Please summarize the following search results for query: {query}:\n{result_json_str[:1000]}",
|
61 |
+
"max_tokens": 1000,
|
62 |
+
"n": 1,
|
63 |
+
"stop": None,
|
64 |
+
"temperature": 0.5,
|
65 |
+
"stream": True,
|
66 |
+
}
|
67 |
+
|
68 |
+
headers = {"Content-Type": "application/json",
|
69 |
+
"Authorization": f"Bearer {self.api_key}"}
|
70 |
+
|
71 |
+
resp = requests.post("https://api.openai.com/v1/completions",
|
72 |
+
headers=headers,
|
73 |
+
data=json.dumps(body),
|
74 |
+
stream=True)
|
75 |
+
return resp
|