process_html.py

import os
import pinecone

from pydantic import Field
from vector_db import Document
from html_parser import HTMLParser
from langchain.vectorstores import Pinecone
from config import PINECONE_API_KEY, PINECONE_ENVIRONMENT, INDEX_NAME
from config import EMBEDDING_API_BASE, EMBEDDING_API_KEY, OPENAI_API_TYPE, OPENAI_API_VERSION, EMBEDDING_DEPLOYMENT_ID
from langchain.embeddings import OpenAIEmbeddings

WEBSITE_FOLDER = 'website'
parser = HTMLParser()

# initialize pinecone
pinecone.init(
    api_key=PINECONE_API_KEY,  # find at app.pinecone.io
    environment=PINECONE_ENVIRONMENT,  # next to api key in console
)

# Azure embedding model definition
embeddings = OpenAIEmbeddings(
    deployment=EMBEDDING_DEPLOYMENT_ID, 
    openai_api_key=EMBEDDING_API_KEY,
    openai_api_base=EMBEDDING_API_BASE,
    openai_api_type=OPENAI_API_TYPE,
    openai_api_version=OPENAI_API_VERSION,
    chunk_size=16
)

if INDEX_NAME and INDEX_NAME not in pinecone.list_indexes():
    pinecone.create_index(
        INDEX_NAME,
        metric="cosine",
        dimension=1536
    )
    print(f"Index {INDEX_NAME} created successfully")

index = pinecone.Index(INDEX_NAME) 
index.delete(delete_all=True)

files_src = os.listdir(WEBSITE_FOLDER)
documents = []
for file in files_src:
    filepath = os.path.join(WEBSITE_FOLDER, file)
    filename = os.path.basename(filepath)
    data = parser.parse_file(filepath)
    texts= []
    for d in data:  
        texts.append(Document(page_content=d, metadata={"source": filepath, "document_id": filename}))
    documents.extend(texts)
# print(len(documents))
# if len(documents)>0:
#     document_id = [d.metadata['document_id'] + f"_{idx}" for (idx, d) in enumerate(documents)]
#     Pinecone.from_documents(documents, embeddings, ids=document_id, index_name=INDEX_NAME)
#     message = f"Add files to {INDEX_NAME} sucessfully"