NoticeAI - AI Agent for Email Extraction

Overview

NoticeAI is an AI agent built using LangChain and LangGraph to process incoming emails from regulatory bodies, such as safety violations or compliance notices, and extract critical information. The extracted data can then be used to notify internal teams in a Django-based e-commerce platform, based on microservices and orchestrated with Terraform in the AWS cloud.

Key Features

• Email Parsing: Extracts essential fields from regulatory notice emails, including project ID, violation type, entity name, contact information, required changes, compliance deadlines, and potential fines.
• Date Parsing: Extracted dates are converted into proper date objects using Pydantic computed fields.
• Modular Design: Designed to be easily extendable for additional fields and email formats.
• Integration: Intended for integration with a growing e-commerce platform.

Technology Stack

• LangChain: Used to create a language model pipeline for email parsing.
• LangGraph: For orchestrating the chain and handling the flow of extracted information.
• Pydantic: Data validation and settings management.
• Python: The primary programming language used for the implementation.
• AWS: For deployment and cloud infrastructure.
• Django: Backend for the e-commerce platform.
• Terraform: Infrastructure as code to manage AWS resources.
• Poetry: Dependency management and packaging for Python projects.
• Pytest: Testing framework for unit tests and integration testing.
• pytest-cov: Test coverage reporting for ensuring code quality.

Project Structure

chains/notice_email_extraction.py

This file contains the NoticeEmailExtract class, which defines the fields to be extracted from the email, such as:

• entity_name
• entity_phone
• entity_email
• project_id
• violation_type
• site_location
• compliance_deadline
• max_potential_fine

The class includes methods to convert string dates to datetime.date objects.

Installation

To install the dependencies for this project, use Poetry:

poetry install

Make sure you have a valid OpenAI API key and the necessary AWS credentials set up for deployment.

Tests

The project includes unit tests for the following:

Model Tests: Ensuring that the data fields are properly extracted and converted.

Validation Tests: Ensuring correct handling of edge cases like missing fields or invalid dates.

Numeric and Text Validations: Ensuring the agent handles invalid numeric inputs and extra spaces in text fields.

To run the tests, use:

pytest --cov=chains --cov-report=term-missing

Test Coverage

To generate a test coverage report, you can use pytest-cov with the following command:

pytest --cov=chains --cov-report=htmlcov

This will generate an HTML test coverage report that can be shared with your team.

Future Work

As the project grows, I plan to integrate additional features, such as:

Expanding the types of emails parsed (e.g., notifications for other types of regulatory issues).

Integrating with the larger Django e-commerce system for automated notifications and actions based on the parsed data.

Scaling the solution using AWS Lambda and AWS SQS for handling large volumes of incoming emails.

License

This project is licensed under the terms of the MIT License.

Author

Alina Bazavan

Email: sempedia@gmail.com