Parsea

A secure desktop tool that reads your inbox and flags job updates, so you don't waste time checking every email.

Parsea (Quality of Life Update - 27th June 2025)

Parsea (UI Update - 20th June 2025)

Parsea (CyberSecurity Update - 19th June 2025)

Recent Updates (June 27, 2025)

• Refined UI
• Removed Electron Menu Bar
• Removed "Interview", "Offer", "Rejection" Option
• Setup logic for planned MongoDB and React Integration
• Setup LLM logic for other services options

Recent Updates (June 20, 2025)

• Rehauled the entire UI
• Added a DEMO settings Page
• Modified the system prompt to flag each interesting email with a category
• Did some magic

Recent Updates (June 19, 2025)

• Refactored email classification logic to use a detailed system prompt for Cohere AI
• Improved CSV parsing and email formatting for AI input
• Ensured only email titles are returned for classified emails
• Added error handling for Cohere API failures
• Added advanced cybersecurity filter: The app now removes emails that match common authentication code, OTP, or security code patterns before any AI processing. This helps protect sensitive information and keeps your dataset clean.

Cybersecurity Filter Details

CSV FILE AFTER FILERING:

The filter removes emails whose subject or body matches any of these patterns:

• "verification code"
• "authentication code" or "auth code"
• "use code"
• "your code"
• "login code"
• "security code"
• "2fa code"
• "otp"
• Any 4–8 digit number only if it appears near the words "code" or "pin" (e.g., "Your code is 123456")

This is implemented with regular expressions for robust detection. Only emails that do not match these patterns are kept for further processing and AI classification.

Features

• Fetches recent emails from your inbox and saves them to a CSV file using IMAP
• Parses the CSV into structured email objects (subject, sender, date, etc.)
• Cybersecurity filter: Automatically removes emails that appear to be authentication code, OTP, or security code messages using advanced pattern matching (see below)
• Uses Cohere AI (Command-a-03-2025) to classify job application–related emails into categories: Updates, Confirmation, and Other
• Returns only the titles of relevant emails for easy identification (no email contents are exposed)
• Provides a summary count for each category in the AI response
• Robust error handling for API and parsing issues
• Modular code for easy maintenance and extension

Problems

• Requires a better way to parse all the emails locally before sending it to the model
• Design could have a better UI that is more intuitive compared to just a text
• Could experiment with a RAG type workflow for this application
• Context Length might become an issue, because the emails parsed are a lot
• Need another filter to not parse google/outlook/etc authuntication codes
• Need a local model that looks for private information before prompting Cohere LLM with all emails

Use the Product

Prerequisites

• Node.js (version 14 or higher)
• npm or yarn package manager

Installation

1. Clone the repository:

   git clone <repository-url>
   cd parseML

2. Install dependencies:

   npm install

Dependencies

This project uses the following dependencies:

Production Dependencies:

• @azure/msal-node@^3.6.0 - Microsoft Authentication Library for Azure AD
• axios@^1.10.0 - HTTP client for API requests
• cohere-ai@^7.17.1 - Cohere AI API client
• dotenv@^16.5.0 - Environment variable loader
• imap@^0.8.17 - IMAP client library
• imapflow@^1.0.188 - Modern IMAP client
• mailparser@^3.7.3 - Email parsing utilities

Development Dependencies:

• electron@^36.4.0 - Desktop application framework
• electron-reload@^2.0.0-alpha.1 - Hot reload for Electron development

Running the Application

npm start

Contributing

Contributions are welcome! Please open an issue or submit a pull request for improvements or bug fixes.

License

This project is licensed under the MIT License. See the LICENSE file for details.