Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,73 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import requests
|
3 |
+
import os
|
4 |
+
import re
|
5 |
+
|
6 |
+
API_TOKEN = os.getenv('API_TOKEN')
|
7 |
+
API_URL = "https://api-inference.huggingface.co/models/nasa-impact/nasa-smd-ibm-st-v2"
|
8 |
+
headers = {"Authorization": f"Bearer {API_TOKEN}"}
|
9 |
+
|
10 |
+
def query_similarity(source_sentence, sentences):
|
11 |
+
payload = {
|
12 |
+
"inputs": {
|
13 |
+
"source_sentence": source_sentence,
|
14 |
+
"sentences": sentences
|
15 |
+
}
|
16 |
+
}
|
17 |
+
response = requests.post(API_URL, headers=headers, json=payload)
|
18 |
+
return response.json()
|
19 |
+
|
20 |
+
def format_output(response):
|
21 |
+
results = sorted(response, key=lambda x: x['score'], reverse=True)
|
22 |
+
formatted_results = []
|
23 |
+
for item in results:
|
24 |
+
formatted_results.append(f"Sentence: {item['sentence']}, Score: {item['score']:.4f}")
|
25 |
+
return "\n".join(formatted_results)
|
26 |
+
|
27 |
+
def split_into_chunks(text, chunk_size=100):
|
28 |
+
sentences = re.split(r'(?<=[.!?]) +', text) # Split text into sentences
|
29 |
+
chunks = []
|
30 |
+
current_chunk = []
|
31 |
+
current_length = 0
|
32 |
+
|
33 |
+
for sentence in sentences:
|
34 |
+
sentence_length = len(sentence.split())
|
35 |
+
if current_length + sentence_length > chunk_size:
|
36 |
+
chunks.append(" ".join(current_chunk))
|
37 |
+
current_chunk = [sentence]
|
38 |
+
current_length = sentence_length
|
39 |
+
else:
|
40 |
+
current_chunk.append(sentence)
|
41 |
+
current_length += sentence_length
|
42 |
+
|
43 |
+
if current_chunk:
|
44 |
+
chunks.append(" ".join(current_chunk))
|
45 |
+
|
46 |
+
return chunks
|
47 |
+
|
48 |
+
def semantic_search(query, document):
|
49 |
+
chunks = split_into_chunks(document)
|
50 |
+
response = query_similarity(query, chunks)
|
51 |
+
return format_output(response)
|
52 |
+
|
53 |
+
def read_file(file):
|
54 |
+
text = file.read().decode('utf-8')
|
55 |
+
return text
|
56 |
+
|
57 |
+
# Define Gradio interface
|
58 |
+
iface = gr.Interface(
|
59 |
+
fn=semantic_search,
|
60 |
+
inputs=[
|
61 |
+
gr.Textbox(lines=2, placeholder="Enter your query here..."),
|
62 |
+
gr.File(label="Upload a .txt file")
|
63 |
+
],
|
64 |
+
outputs="text",
|
65 |
+
title="Document Semantic Search",
|
66 |
+
description="Input a query and upload a document (.txt) to find the most semantically similar paragraphs or sentences.",
|
67 |
+
examples=[
|
68 |
+
["Enter a sample query here...", None]
|
69 |
+
]
|
70 |
+
)
|
71 |
+
|
72 |
+
iface.launch()
|
73 |
+
|