TaskScribe 📝⚡

AI-Powered Meeting → Task Management SaaS

TaskScribe is a full-stack web application that automatically converts meeting transcripts into structured, actionable tasks using a Large Language Model (LLM). It helps teams capture decisions, track ownership, and maintain momentum after meetings.

This project was built as a production-style SaaS, not a demo — with authentication, persistence, AI extraction, and deployment-ready architecture.

---

✨ Features

🔐 Authentication

• Secure user registration & login
• JWT-based authentication
• Per-user task isolation

🤖 AI Task Extraction

• Converts raw meeting transcripts into tasks
• Extracts:
  • Task description
  • Owner
  • Deadline
  • Priority
• Uses Groq-hosted LLMs with structured JSON prompting for high accuracy
• Ignores filler discussion, questions, and non-commitments

📋 Task Management

• View all extracted tasks
• Edit tasks (description, owner, deadline, priority)
• Mark tasks complete / reopen
• Delete individual tasks
• Delete all tasks (useful for testing)
• Timestamps (created_at, updated_at)

🎨 Modern UI / UX

• Dark-mode SaaS interface
• Glassmorphism & gradient styling
• Responsive layout
• Toast notifications & confirmation dialogs
• Empty states and loading states

---

🧠 How Task Extraction Works

Instead of brittle regex or rule-based NLP, TaskScribe uses:

• Groq API
• LLaMA-3.3-70B (versatile) model
• JSON-structured prompting to enforce deterministic outputs

Why JSON prompting?

• Precise control over model behavior
• Consistent, machine-parsable output
• Easy validation and post-processing
• Production-grade reliability

---

🏗️ Tech Stack

Frontend

• React
• Bootstrap (dark theme)
• Axios
• Vercel (deployment)

Backend

• FastAPI
• SQLAlchemy
• JWT Authentication (python-jose)
• Groq Python SDK
• Render (deployment)

AI / LLM

• Groq API
• llama-3.3-70b-versatile
• Structured JSON prompting

---