File size: 4,313 Bytes
d99956a eb7475d 4c6ab9d eb7475d 4c6ab9d d99956a 4c6ab9d eb7475d d99956a 4c6ab9d eb7475d 1e54781 eb7475d 1e54781 4c6ab9d eb7475d d99956a eb7475d 4c6ab9d d99956a b25b520 d99956a b25b520 d99956a b25b520 d99956a 4c6ab9d b25b520 eb7475d b25b520 eb7475d 1e54781 4c6ab9d d99956a 4c6ab9d b25b520 4c6ab9d eb7475d d99956a b25b520 d99956a 4c6ab9d d99956a b25b520 d99956a eb7475d d99956a eb7475d d99956a eb7475d 4c6ab9d eb7475d b25b520 | 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 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 | from tqdm import tqdm
import os
import chardet
import numpy as np
from pinecone import Pinecone
from langchain.docstore.document import Document as LangchainDocument
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_community.embeddings import OpenAIEmbeddings
import spacy
# Constants
DATA_DIR = "data"
JOURNAL_DIR = "journals"
# Load environment variables
OPENAI_API_KEY = os.environ["OPENAI_API_KEY"]
PINECONE_API_KEY = os.environ["PINECONE_API_KEY"]
PINECONE_INDEX = os.environ["PINECONE_INDEX"]
# Initialize Pinecone
pc = Pinecone(api_key=PINECONE_API_KEY)
index = pc.Index(PINECONE_INDEX)
# Initialize embedding model
embedding_model = OpenAIEmbeddings(
model="text-embedding-3-small",
api_key=OPENAI_API_KEY
)
nlp = spacy.load("xx_sent_ud_sm") # For multilingual support including Arabic
def sentence_overlap_chunks(text, chunk_size=2000):
doc = nlp(text)
sentences = [sent.text.strip() for sent in doc.sents]
chunks = []
i = 0
while i < len(sentences):
chunk = []
length = 0
start_i = i
# Fill chunk up to chunk_size characters
while i < len(sentences) and length + len(sentences[i]) <= chunk_size:
chunk.append(sentences[i])
length += len(sentences[i]) + 1
i += 1
# Join and store chunk
if chunk:
try:
chunk = " ".join(chunk)
chunks.append(chunk)
i+=1
except:
print("can't process line.")
i+=4
# Overlap: start next chunk with last sentence of current chunk
i = i - 3
return chunks
# for data
for filename in os.listdir(DATA_DIR):
if filename.endswith(".pdf"):
filepath = os.path.join(DATA_DIR, filename)
namespace = "ns4"
print(f"Processing {filename} → namespace: {namespace}")
reader = fitz.open(filepath)
content = ""
for page in reader:
text = page.get_text()
if text:
content += text + "\n"
# Wrap in Langchain doc
docs_processed = sentence_overlap_chunks(content)
# Embed and prepare for upsert
upsert_data = []
for i, chunk in tqdm(enumerate(docs_processed), total=len(docs_processed), desc="Embedding chunks"):
vector = embedding_model.embed_query(chunk)
upsert_data.append({
"id": f"{filename[:-4]}_chunk_{i}",
"values": vector,
"metadata": {
"text": chunk,
"source": filename
}
})
# Upsert to Pinecone under this file's namespace
print(f"⬆️ Upserting {len(upsert_data)} vectors to namespace '{namespace}'...")
index.upsert(vectors=upsert_data, namespace=namespace)
print(f"✅ Done with {filename}\n")
# for journals
for filename in os.listdir(JOURNAL_DIR):
if filename.endswith(".txt"):
filepath = os.path.join(JOURNAL_DIR, filename)
namespace = filename[:-4]
print(f"Processing {filename} → namespace: {namespace}")
# Detect encoding
with open(filepath, "rb") as f:
raw_data = f.read()
encoding = chardet.detect(raw_data)['encoding']
# Read file
with open(filepath, "r", encoding=encoding) as f:
content = f.read()
# Wrap in Langchain doc
docs_processed = sentence_overlap_chunks(content, chunk_size=400)
# Embed and prepare for upsert
upsert_data = []
for i, chunk in tqdm(enumerate(docs_processed), total=len(docs_processed), desc="Embedding chunks"):
vector = embedding_model.embed_query(chunk)
upsert_data.append({
"id": f"{namespace}_chunk_{i}",
"values": vector,
"metadata": {
"text": chunk,
"source": filename
}
})
# Upsert to Pinecone under this file's namespace
print(f"⬆️ Upserting {len(upsert_data)} vectors to namespace '{namespace}'...")
index.upsert(vectors=upsert_data, namespace=namespace)
print(f"✅ Done with {filename}\n") |