Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,128 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
+
import time
|
3 |
+
import pdfplumber
|
4 |
+
import docx
|
5 |
+
import nltk
|
6 |
+
import gradio as gr
|
7 |
+
from langchain.embeddings import HuggingFaceEmbeddings
|
8 |
+
from langchain.vectorstores import FAISS
|
9 |
+
from langchain.text_splitter import RecursiveCharacterTextSplitter, SentenceTextSplitter
|
10 |
+
from sentence_transformers import SentenceTransformer
|
11 |
+
from transformers import AutoTokenizer
|
12 |
+
from nltk import sent_tokenize
|
13 |
+
from typing import List, Tuple
|
14 |
+
from transformers import AutoModel, AutoTokenizer
|
15 |
+
|
16 |
+
# Ensure nltk sentence tokenizer is downloaded
|
17 |
+
nltk.download('punkt')
|
18 |
+
|
19 |
+
FILES_DIR = './files'
|
20 |
+
|
21 |
+
# Supported embedding models
|
22 |
+
MODELS = {
|
23 |
+
'e5-base': "danielheinz/e5-base-sts-en-de",
|
24 |
+
'multilingual-e5-base': "multilingual-e5-base",
|
25 |
+
'paraphrase-miniLM': "paraphrase-multilingual-MiniLM-L12-v2",
|
26 |
+
'paraphrase-mpnet': "paraphrase-multilingual-mpnet-base-v2",
|
27 |
+
'gte-large': "gte-large",
|
28 |
+
'gbert-base': "gbert-base"
|
29 |
+
}
|
30 |
+
|
31 |
+
class FileHandler:
|
32 |
+
@staticmethod
|
33 |
+
def extract_text(file_path):
|
34 |
+
ext = os.path.splitext(file_path)[-1].lower()
|
35 |
+
if ext == '.pdf':
|
36 |
+
return FileHandler._extract_from_pdf(file_path)
|
37 |
+
elif ext == '.docx':
|
38 |
+
return FileHandler._extract_from_docx(file_path)
|
39 |
+
elif ext == '.txt':
|
40 |
+
return FileHandler._extract_from_txt(file_path)
|
41 |
+
else:
|
42 |
+
raise ValueError(f"Unsupported file type: {ext}")
|
43 |
+
|
44 |
+
@staticmethod
|
45 |
+
def _extract_from_pdf(file_path):
|
46 |
+
with pdfplumber.open(file_path) as pdf:
|
47 |
+
return ' '.join([page.extract_text() for page in pdf.pages])
|
48 |
+
|
49 |
+
@staticmethod
|
50 |
+
def _extract_from_docx(file_path):
|
51 |
+
doc = docx.Document(file_path)
|
52 |
+
return ' '.join([para.text for para in doc.paragraphs])
|
53 |
+
|
54 |
+
@staticmethod
|
55 |
+
def _extract_from_txt(file_path):
|
56 |
+
with open(file_path, 'r', encoding='utf-8') as f:
|
57 |
+
return f.read()
|
58 |
+
|
59 |
+
class EmbeddingModel:
|
60 |
+
def __init__(self, model_name, max_tokens=None):
|
61 |
+
self.model = HuggingFaceEmbeddings(model_name=model_name)
|
62 |
+
self.max_tokens = max_tokens
|
63 |
+
|
64 |
+
def embed(self, text):
|
65 |
+
return self.model.embed_documents([text])
|
66 |
+
|
67 |
+
def process_files(model_name, split_strategy, chunk_size=500, overlap_size=50, max_tokens=None):
|
68 |
+
# File processing
|
69 |
+
text = ""
|
70 |
+
for file in os.listdir(FILES_DIR):
|
71 |
+
file_path = os.path.join(FILES_DIR, file)
|
72 |
+
text += FileHandler.extract_text(file_path)
|
73 |
+
|
74 |
+
# Split text
|
75 |
+
if split_strategy == 'sentence':
|
76 |
+
splitter = SentenceTextSplitter(chunk_size=chunk_size, chunk_overlap=overlap_size)
|
77 |
+
else:
|
78 |
+
splitter = RecursiveCharacterTextSplitter(chunk_size=chunk_size, chunk_overlap=overlap_size)
|
79 |
+
|
80 |
+
chunks = splitter.split_text(text)
|
81 |
+
model = EmbeddingModel(MODELS[model_name], max_tokens=max_tokens)
|
82 |
+
embeddings = model.embed(text)
|
83 |
+
|
84 |
+
return embeddings, chunks
|
85 |
+
|
86 |
+
def search_embeddings(query, model_name, top_k):
|
87 |
+
model = HuggingFaceEmbeddings(model_name=MODELS[model_name])
|
88 |
+
embeddings = model.embed_query(query)
|
89 |
+
return embeddings
|
90 |
+
|
91 |
+
def calculate_statistics(embeddings):
|
92 |
+
# Return time taken, token count, etc.
|
93 |
+
return {"tokens": len(embeddings), "time_taken": time.time()}
|
94 |
+
|
95 |
+
# Gradio frontend
|
96 |
+
def upload_file(file, model_name, split_strategy, chunk_size, overlap_size, max_tokens, query, top_k):
|
97 |
+
with open(os.path.join(FILES_DIR, file.name), "wb") as f:
|
98 |
+
f.write(file.read())
|
99 |
+
|
100 |
+
# Process files and get embeddings
|
101 |
+
embeddings, chunks = process_files(model_name, split_strategy, chunk_size, overlap_size, max_tokens)
|
102 |
+
|
103 |
+
# Perform search
|
104 |
+
results = search_embeddings(query, model_name, top_k)
|
105 |
+
|
106 |
+
# Calculate statistics
|
107 |
+
stats = calculate_statistics(embeddings)
|
108 |
+
|
109 |
+
return {"results": results, "stats": stats}
|
110 |
+
|
111 |
+
# Gradio interface
|
112 |
+
iface = gr.Interface(
|
113 |
+
fn=upload_file,
|
114 |
+
inputs=[
|
115 |
+
gr.File(label="Upload File"),
|
116 |
+
gr.Dropdown(choices=list(MODELS.keys()), label="Embedding Model"),
|
117 |
+
gr.Radio(choices=["sentence", "recursive"], label="Split Strategy"),
|
118 |
+
gr.Slider(100, 1000, step=100, value=500, label="Chunk Size"),
|
119 |
+
gr.Slider(0, 100, step=10, value=50, label="Overlap Size"),
|
120 |
+
gr.Slider(50, 500, step=50, value=200, label="Max Tokens"),
|
121 |
+
gr.Textbox(label="Search Query"),
|
122 |
+
gr.Slider(1, 10, step=1, value=5, label="Top K")
|
123 |
+
],
|
124 |
+
outputs="json"
|
125 |
+
)
|
126 |
+
|
127 |
+
iface.launch()
|
128 |
+
|