-
Notifications
You must be signed in to change notification settings - Fork 0
/
build_database.py
77 lines (64 loc) · 2.52 KB
/
build_database.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
import os
import re
from langchain.document_loaders import TextLoader
from tqdm import tqdm
from langchain.text_splitter import RecursiveCharacterTextSplitter
import chromadb
from chromadb.config import Settings
# Adding \n at end of each sentence and \n\n at end of each paragraph
# It will help in recursive splitter
def add_line_breaks(documentText):
sentence_splits = documentText.split(".")
sentence_with_delimiter = ".\n".join(sentence_splits)
# If number of whitespaces is more than 2, then it is another paragraph
sentence_with_delimiter = re.sub(r' {2,}', '\n\n', sentence_with_delimiter)
return sentence_with_delimiter
documents = []
metadata = []
# Getting all the documents and metadata, and storing it in a list
for comp in tqdm(os.listdir("datasets")):
for prod in os.listdir(f"datasets/{comp}"):
# if prod == "Apple Fitness+.docx":continue
loader = TextLoader(f"datasets/{comp}/{prod}", autodetect_encoding=True)
doc = loader.load()
text = add_line_breaks(doc[0].page_content)
documents.append(text)
metadata.append({"company": comp, "product": prod.split(".")[0]})
# Recursive splitter
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=250)
# Split each element in the list
split_list = [text_splitter.split_text(element) for element in documents]
splitted_docs = []
splitted_metadata = []
# Aggregating the split texts
for idx, docs in enumerate(split_list):
curr_metadata = metadata[idx]
if isinstance(docs, list):
for doc in docs:
splitted_docs.append(doc)
splitted_metadata.append(curr_metadata)
else:
splitted_docs.append(docs)
splitted_metadata.append(curr_metadata)
chroma_client = chromadb.Client(
Settings(
chroma_db_impl="duckdb+parquet",
persist_directory="tandc-db"
))
collection_name = "TandC-project"
# Vector database
if len(chroma_client.list_collections()) > 0 and collection_name in [
chroma_client.list_collections()[0].name
]:
chroma_client.delete_collection(name=collection_name)
else:
print(f"Creating collection: '{collection_name}'")
collection = chroma_client.create_collection(name=collection_name)
print("Building the vector database")
collection.add(
documents=splitted_docs,
metadatas=splitted_metadata,
ids=[f"id{i}" for i in range(1, len(splitted_docs) + 1)]
)
print(f"Completed building vector database")
chroma_client.persist()