Skip to content

ebad66/FlowState

Repository files navigation

FlowState 🧠⚡

A real-time AI focus coach that helps you stay in flow, without burning out.


Inspiration

Staying focused is hard. Staying focused consistently is even harder—especially for students and remote workers who lack structure, accountability, or coaching.

Most productivity tools rely on:

  • rigid timers
  • manual checklists
  • or guilt-driven notifications

FlowState was built to be different.

Instead of forcing behavior, FlowState observes how you actually work, infers your cognitive state, and intervenes only when it matters.


What it does

FlowState is an AI-powered focus companion that uses voice, video, and system context to keep users in a productive flow state.

It continuously senses, predicts, and responds—like a real accountability coach.


Core capabilities

🎙️ Voice-first interaction

Talk to FlowState naturally:

  • “I’m working on my assignment.”
  • “Switch to crunch mode.”
  • “Start a 10 minute break.”
  • “What should I work on next?”

All feedback is delivered through real-time voice.


👀 Attention & presence detection

FlowState uses camera and system signals to estimate:

  • whether the user is present
  • whether they’re engaged or drifting
  • when they leave or switch contexts

This allows FlowState to detect:

  • unintentional distractions
  • prolonged inactivity
  • context switching (e.g. YouTube, social media)

⏱️ Smart breaks & task switching

  • Suggests breaks before burnout
  • Detects “fake breaks” (idle but not resting)
  • Encourages task switching when focus drops
  • Protects deep work by muting distractions

🔔 Adaptive nudges

FlowState intervenes only when necessary:

  • “You’ve been idle for a bit. Resume or switch?”
  • “You’re deep in flow. Muting distractions.”
  • “You switched tasks—was that intentional?”

No spam. No micromanagement. Just timely interventions.


Modes

😌 Casual Mode

  • Gentle reminders
  • Flexible timing
  • Ideal for everyday work

😈 Crunch SZN

  • Firmer nudges
  • Tighter break enforcement
  • Accountability-focused
  • Built for deadlines and exam weeks

System Architecture

FlowState runs as a closed-loop human-AI system:

Sensors → Presage → LLM → ElevenLabs → User

Inputs

  • Microphone → voice commands + activity cues
  • Webcam → presence and attention signals
  • System context → active apps, idle time
  • User intent → tasks and goals

Intelligence Layer (Presage)

Presage predicts the user’s cognitive state in real time:

  • focus score
  • distraction risk
  • fatigue risk

This determines when FlowState should intervene.

Decision Layer (LLM)

The LLM decides:

  • whether to speak
  • what to say
  • what tone to use

Output (ElevenLabs)

ElevenLabs generates:

  • realistic voice responses
  • adaptive emotional tone
  • real-time spoken feedback

This determines how the intervention feels.


Tech Stack

  • Presage – Real-time cognitive state prediction from sensor signals
  • ElevenLabs – Adaptive text-to-speech voice agent
  • Frontend – Web-based overlay UI
  • Backend – Python orchestration engine
  • AI – Policy-based reasoning + focus heuristics

Accessibility & Impact

FlowState is designed to reduce inequality around productivity:

  • No expensive coaching or subscriptions
  • Voice-first for accessibility
  • Flexible modes for neurodivergent users
  • Supports people who struggle with structure

Focus shouldn’t be a privilege.


Privacy

  • Camera and mic usage is transparent and optional
  • Only lightweight features are processed
  • No raw video or audio is stored
  • Users can disable any input at any time

What’s next

  • Personal focus pattern learning
  • Smarter fatigue detection
  • Cross-device sync
  • Calendar & task integrations

Built at

HackHive
By Ebad & Shay

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 2

  •  
  •