gavinzli's picture
Refactor email metadata handling in list_emails function and update vectorstore initialization with Pinecone integration
8dae98c
"""Module to search and list emails from Gmail."""
import os
import re
import base64
import hashlib
from datetime import datetime, timedelta
from venv import logger
from ics import Calendar
from langchain_core.documents import Document
from langchain_community.document_loaders import (
PyPDFLoader,
UnstructuredExcelLoader,
CSVLoader,
UnstructuredImageLoader,
)
from models.db import vectorstore
SCOPES = ["https://www.googleapis.com/auth/gmail.readonly"]
EMAIL_PATTERN = r"[a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,}"
ATTACHMENTS_DIR = "cache"
os.makedirs(ATTACHMENTS_DIR, exist_ok=True)
# service = build_gmail_service()
def search_emails(service, query):
"""Search emails based on a query."""
result = service.users().messages().list(userId="me", q=query).execute()
messages = []
if "messages" in result:
messages.extend(result["messages"])
while "nextPageToken" in result:
page_token = result["nextPageToken"]
result = (
service.users().messages().list(userId="me", q=query, pageToken=page_token).execute()
)
if "messages" in result:
messages.extend(result["messages"])
return messages
def list_emails(service, messages):
"""
Processes a list of email messages, extracts metadata, decodes content, and handles attachments.
Args:
messages (list): A list of email message dictionaries, where each dictionary contains
at least an 'id' key representing the email's unique identifier.
Returns:
None: The function processes the emails and adds the extracted documents to a vector store.
Functionality:
- Retrieves email details using the Gmail API.
- Extracts metadata such as sender, recipient, subject, CC, and date.
- Decodes email content in plain text or HTML format.
- Handles multipart emails, including attachments.
- Processes attachments based on their MIME type:
- PDF files are loaded using PyPDFLoader.
- Images (PNG, JPEG) are loaded using UnstructuredImageLoader.
- CSV files are loaded using CSVLoader.
- Excel files are loaded using UnstructuredExcelLoader.
- Calendar files (ICS) are parsed to extract event details.
- Removes HTML tags from email content.
- Stores processed documents and metadata in a vector store.
- Deletes temporary files created during attachment processing.
Notes:
- The function assumes the existence of a global `service` object for Gmail API.
- The `vectorstore.add_documents` method is used to store the processed documents.
- Attachments are temporarily saved in `ATTACHMENTS_DIR` and deleted after processing.
- The function logs information about attachments being downloaded.
"""
ids = []
documents = []
for message in messages:
msg = service.users().messages().get(userId="me", id=message["id"], format="full").execute()
metadata = {}
metadata["threadId"] = msg["threadId"]
metadata["msgId"] = msg["id"]
msgId = f"{msg['threadId']}-{msg['id']}"
# for docstore_id in list(vectorstore.index_to_docstore_id.values()):
# if docstore_id.startswith(msgId):
# logger.info("Already indexed: %s", msgId)
# continue
for header in msg["payload"]["headers"]:
if header["name"] == "From":
metadata["from"] = header["value"]
elif header["name"] == "To":
metadata["to"] = header["value"]
elif header["name"] == "Subject":
metadata["subject"] = header["value"]
logger.info("subject: %s", metadata["subject"])
elif header["name"] == "Cc":
metadata["cc"] = header["value"]
metadata["date"] = datetime.fromtimestamp(int(msg["internalDate"]) / 1000).strftime(
"%d/%m/%Y %H:%M:%S"
)
metadata["userId"] = service.users().getProfile(userId="me").execute().get("emailAddress")
# print(metadata, msg["payload"]["mimeType"])
ids = []
documents = []
mime_types = []
if msg["payload"]["mimeType"] in [
"multipart/alternative",
"multipart/related",
"multipart/mixed",
]:
mime_types = []
attach_docs = []
for part in msg["payload"]["parts"]:
print("mimeType: ", part["mimeType"])
mime_types.append(part["mimeType"])
if part["mimeType"] == "text/plain" and "text/html" not in mime_types:
body = base64.urlsafe_b64decode(part["body"]["data"]).decode("utf-8")
body = re.sub(r"<[^>]+>", "", body) # Remove HTML tags
metadata["mimeType"] = part["mimeType"]
documents.append(Document(page_content=body, metadata=metadata))
ids.append(msg["id"])
elif part["mimeType"] == "text/html" and "text/plain" not in mime_types:
body = base64.urlsafe_b64decode(part["body"]["data"]).decode("utf-8")
body = re.sub(r"<[^>]+>", "", body)
metadata["mimeType"] = part["mimeType"]
documents.append(Document(page_content=body, metadata=metadata))
ids.append(msg["id"])
if part["filename"]:
attachment_id = part["body"]["attachmentId"]
logger.info("Downloading attachment: %s", part["filename"])
attachment = (
service.users()
.messages()
.attachments()
.get(userId="me", messageId=message["id"], id=attachment_id)
.execute()
)
file_data = base64.urlsafe_b64decode(attachment["data"].encode("UTF-8"))
path = os.path.join(".", ATTACHMENTS_DIR, part["filename"])
with open(path, "wb") as f:
f.write(file_data)
if part["mimeType"] == "application/pdf":
attach_docs = PyPDFLoader(path).load()
elif part["mimeType"] == "image/png" or part["mimeType"] == "image/jpeg":
try:
attach_docs = UnstructuredImageLoader(path).load()
except Exception as e:
logger.error("Error loading image: %s", e)
elif part["filename"].endswith(".csv"):
attach_docs = CSVLoader(path).load()
elif (
part["mimeType"]
== "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"
):
attach_docs = UnstructuredExcelLoader(path).load()
elif part["mimeType"] == "application/ics":
with open(path, "r", encoding="utf-8") as f:
calendar = Calendar(f.read())
for event in calendar.events:
documents.append(
Document(
page_content=f"Event: {event.name}\n\Description: {event.description}\nStart: {event.begin}\nEnd: {event.end}",
metadata={
"attachment": part["filename"],
"mimeType": part["mimeType"],
"location": event.location,
"created": event.created.strftime("%d/%m/%Y %H:%M:%S"),
"last_modified": event.last_modified.strftime(
"%d/%m/%Y %H:%M:%S"
),
"start": event.begin.strftime("%d/%m/%Y %H:%M:%S"),
"end": event.end.strftime("%d/%m/%Y %H:%M:%S"),
},
)
)
ids.append(f"{msgId}-{part['filename']}-{hashlib.sha256(file_data).hexdigest()}")
if os.path.exists(path):
os.remove(path)
for index, document in enumerate(attach_docs or []):
document.metadata["mimeType"] = part["mimeType"]
if "page_label" in document.metadata:
document.metadata["page"] = document.metadata["page_label"]
document.metadata["attachment"] = part["filename"]
document.metadata = {
key: value
for key, value in document.metadata.items()
if key in ["attachment", "page"]
}
document.metadata.update(metadata)
documents.append(document)
ids.append(f"{msgId}-{hashlib.sha256(file_data).hexdigest()}-{index}")
elif msg["payload"]["mimeType"] == "text/plain" and "data" in msg["payload"]["body"]:
body = base64.urlsafe_b64decode(msg["payload"]["body"]["data"]).decode("utf-8")
body = re.sub(r"<[^>]+>", "", body)
metadata["mimeType"] = msg["payload"]["mimeType"]
documents.append(Document(page_content=body, metadata=metadata))
ids.append(msgId)
elif msg["payload"]["mimeType"] == "text/html" and "data" in msg["payload"]["body"]:
body = base64.urlsafe_b64decode(msg["payload"]["body"]["data"]).decode("utf-8")
body = re.sub(r"<[^>]+>", "", body)
metadata["mimeType"] = msg["payload"]["mimeType"]
documents.append(Document(page_content=body, metadata=metadata))
ids.append(msgId)
if "multipart/alternative" in mime_types and len(mime_types) == 1:
print("Only multipart/alternative found in the email.")
else:
try:
vectorstore.add_documents(documents=documents, ids=ids)
except Exception as e:
logger.error("Error adding documents to vectorstore: %s", e)
def collect(service, query=(datetime.today() - timedelta(days=10)).strftime("after:%Y/%m/%d")):
"""
Main function to search and list emails from Gmail.
This function builds a Gmail service, constructs a query to search for emails
received in the last 14 days, and lists the found emails. If no emails are found,
it prints a message indicating so.
Returns:
None
"""
# query = "subject:Re: Smartcareers algorithm debug and improvement'"
emails = search_emails(service, query)
if emails:
logger.info("Found %d emails:\n", len(emails))
logger.info("Found %d emails after two_weeks_ago:\n", len(emails))
list_emails(service, emails)
logger.info("Listing emails...")
return f"{len(emails)} emails added to the collection."
else:
logger.info("No emails found after two weeks ago.")