Doable: AI Personal Productivity Agent

BMW Challenge Submission

DeepWiki & Advanced Documentation: https://deepwiki.com/RayyanNafees/doable

---

🚀 Project Overview

Doable is an intelligent, agentic assistant designed to redefine personal productivity. Built for the BMW Challenge, it goes beyond simple task tracking by integrating logical reasoning, psychological profiling, and external tools to act as a true partner in your workflow.

Tailored to specific personas (Software Developer, Product Manager, Team Leader), Doable bridges the gap between intent ("Schedule a coding session") and action (Calendar booking + Task creation) using a dual-backend architecture of Next.js 16 and a Python MCP Server.

💡 Core Functionality

• Natural Language Processing: Doable interprets complex goals like "Summarize my to-do list and estimate effort" or "Schedule a 2-hour coding session after lunch" using Google Gemini 2.5.
• Calendar Integration: Two-way sync with Google Calendar to manage schedules effectively.
• Task Management: Maintains "long-term memory" of tasks, handling priority ranking and deadlines automatically.
• Conversational Interface: A human-like assistant that provides status updates, motivation, and reasoning.
• MCP Integration: Exposes the agent's capabilities as a Model Context Protocol (MCP) Server (api/mcp_server.py).

🌟 Unique Features

Doable implements several innovative features to boost productivity and motivation:

• 🧠 Psychology-Driven (Ikigai):
  • Users take a psychology quiz during onboarding to identify their "Ikigai" (Reason for being).
  • The AI uses this profile to explain the "Why" behind every task, making even mundane work feel meaningful.
• ⏳ Reverse Pomodoro:
  • For tasks delayed by 1-2 days, Doable activates a unique timer: Work 5 mins, Rest 20 mins. This lowers the barrier to entry to get you started.
• 🧩 Smart Task Breakdown:
  • Automatically decomposes large, vague tasks into actionable 5-minute sub-steps.
• ⚖️ Automatic Eisenhower Matrix:
  • Uses psychological data and task metadata to auto-sort tasks into: Do First, Schedule, Delegate, Don't Do.
• 🏢 Automated Project Planning:
  • Companies can input a raw project plan; Doable generates all required tasks and recommends the best employees based on their psychology and completion history.
• 🔗 Ecosystem Integrations:
  • Email: Automatically creates tasks from flagged emails.
  • Tools: Syncs with ClickUp and Todoist.
• 🔔 Motivational Notifications:
  • Smart notifications for ignored tasks, featuring quotes and impact analysis: "Skipping this task delays your goal of becoming a Senior Dev by 2 weeks."

🎯 Bonus Objectives Achieved

• Self-Reflection: The agent verifies its own actions (e.g., "Did I successfully book the meeting?") using the reflect_on_task tool.
• AI Observability (Langfuse): Full integration with Langfuse for tracing agent thought processes, monitoring token usage, and debugging complex reasoning steps.

🛠️ Technical Architecture

Tech Stack

• Frontend: Next.js 16 (App Router), React 19, Tailwind CSS v4, Shadcn/UI.
• AI Engine: Google Gemini 2.5 Flash (via Vercel AI SDK v5).
• Agent Server: Python FastMCP (Model Context Protocol).
• Database: MongoDB (Mongoose) for flexible document storage.
• Observability: Langfuse (OpenTelemetry).

Project Structure

d:\web\doable
├── app/                  # Next.js App Router
│   ├── api/ai/           # AI Routes (Chat, Process)
│   └── dashboard/        # Main User Interface
├── api/                  # Python MCP Server (Agent Logic)
│   └── mcp_server.py     # FastMCP Implementation
├── components/           # React Components (Shadcn)
├── lib/                  # Utilities & Models
│   ├── models/           # Mongoose Schemas (User, Task, Project)
│   └── ai/               # AI Prompts & Logic
└── docs/                 # Documentation & Slides

🔧 Getting Started

Prerequisites

• Node.js 18+ & Python 3.10+
• MongoDB Instance
• Google AI Studio API Key
• Langfuse API Keys

Installation

1. Clone the repository

   git clone https://github.com/RayyanNafees/doable.git
   cd doable

2. Install Frontend Dependencies

   npm install

3. Setup Environment Variables Create a .env file:

   GOOGLE_API_KEY=your_key
   MONGO_URI=your_mongo_uri
   LANGFUSE_PUBLIC_KEY=pk-lf-...
   LANGFUSE_SECRET_KEY=sk-lf-...

4. Run Development Server

   npm run dev

   Open http://localhost:3000.

---

Submitted for the BMW Challenge Focused on: Context tracking, reasoning, tool orchestration, NLU, and time management logic.