TaskMaster AI - Email Triage Commander

UI interface

Inspiration Every day, professionals waste 30-60 minutes manually reading through dozens of emails, extracting action items, and drafting replies. This repetitive chore inspired me to build an autonomous agent that handles email triage automatically. The goal was simple: let AI do the heavy lifting so I can focus on actual work.

What it does TaskMaster AI is an autonomous email triage agent that: Summarizes messy email inbox content instantly Extract action items (tasks to do, people to reply to, deadlines). Automatically saves tasks to a persistent database Drafts replies and saves them for review Operates completely autonomously—just paste your emails and let the agent work.

How I built it Tech Stack: Gemini 2.5 Flash (via Google AI Studio)—The brain that understands emails and decides actions Firestore—Persistent NoSQL database for tasks and drafts Flask—Lightweight Python web framework for the interface Google GenAI SDK – Function calling for autonomous tool execution

Key Technical Decisions: Used Gemini's Function Calling capability to allow the agent to actually execute actions (saving to database, drafting emails) Decoupled the agent logic from the web interface for clean separation of concerns Used Firestore for persistent state—tasks survive between sessions

Challenges I faced Authentication complexity: Setting up Firestore credentials required careful environment variable management. Function Calling accuracy: Training Gemini to correctly identify which tool to call and extract proper parameters from messy email text Time constraint: Building a production-ready agent within limited time required ruthless prioritization.

Accomplishments that I'm proud of Built a fully functional autonomous agent Successfully implemented Gemini's function calling for real-world actions Created a clean web interface for easy demonstration The agent consistently extracts and saves tasks from real emails. All code is publicly available on GitHub with clear setup instructions.

What I learned Gemini's function calling is incredibly powerful for building autonomous agents. Firestore is perfect for agent state persistence. Less code is better—focus on the core agent logic.

What's next for TaskMaster AI? Gmail API integration: Directly read emails instead of copy-pasting. Email sending: Actually send drafted replies automatically. Slack integration: Send task notifications to Slack User authentication: Multiple users with their own task lists Advanced prioritization: ML-based urgency detection