Skip to content
BranchesTags

Repository files navigation

KarmaOS

The Workforce Operating System — Intelligence · Operations · Automation

React TypeScript Vite Gemini React Flow

🧠 What is KarmaOS?

Most productivity tools measure activity—keystrokes, hours logged, or messages sent. KarmaOS is different. It measures impact and friction.

KarmaOS is a passive, zero-friction intelligence layer that reads metadata across your tools (Slack, GitHub, Calendar, Jira). It doesn't surveil message content—it analyzes patterns (e.g., context switching, thread density, PR review times) and correctly classifies work patterns into distinct states: Deep Work, Shallow Work, and Coordination.

By applying intelligent classification and LLM reasoning via Google Gemini, KarmaOS turns these raw metadata signals into clear, actionable human narratives. Instead of forcing you to interpret a chaotic dashboard of charts, it acts as a thinking partner, proactively telling you where the friction is: "Good morning. Your team is highly focused today, but engineering is currently blocked on design approvals for the new checkout flow."

KarmaOS v2 extends this intelligence layer with a complete HR Operations backbone and an Agentic Automation Engine, transforming it from an analytics tool into a complete, modern workforce operating system.

🚀 Key Capabilities & Differentiation

  • Activity vs. Impact: KarmaOS understands that 100 Slack messages could signify high team alignment, or a completely broken process. It measures the quality of time, not just the quantity.
  • Quantifying "Invisible Work": KarmaOS detects workflow bottlenecks that happen between tools (e.g., a ticket that sits in review while 50 Slack messages are exchanged about it), pinpointing systemic friction rather than blaming individuals.
  • Zero-Friction Passive Intelligence: No timers, no forms, no status updates. KarmaOS operates completely passively in the background, requiring absolutely zero behavioral changes from your team.
  • Proactive Burnout Prevention: By tracking context switching and after-hours coordination, KarmaOS can objectively identify overload risks before burnout occurs, shifting management from reactive to proactive.
  • Synthesized AI Narratives, Not Fake Charts: We strictly avoid "Fake AI." Models aren't used to draw charts or tally numbers. We use deterministic algorithms for classification, and rely on Gemini exclusively for synthesis—taking disparate data points and telling the objective human story behind them.

✨ Core Features

🔍 Intelligence Layer

Feature Description
System Overview Daily narrative-driven dashboard with AI-synthesized workforce insights
Recruiter Agent Automated candidate sourcing, screening, and Work DNA matching
Team Insights Performance analysis across departments with deep work metrics
Work DNA Profiles Per-employee behavioral classification (Maker, Synchronizer, Operator)
AI Assistant Conversational interface powered by Gemini for deep-dive queries

🏢 HR Operations

Feature Description
Leave Management Apply, approve/reject, balance tracking with sick/casual/unpaid types
Attendance Check-in/out, working hours calculation, daily status grid
Payroll Summary Monthly salary breakdown with unpaid leave deductions

⚡ Automation Engine

Feature Description
Natural Language Rules Describe rules in plain English (e.g., "If an employee takes more than 3 leaves in a month, flag them")
Gemini-Powered Parsing AI converts prompts into structured {condition, trigger, action} rules
Visual Workflow React Flow pipeline visualization (Data Source → Trigger → Condition → Action → Output)
Live Evaluation One-click rule execution against live workforce data

🔎 Hiring Automation

Feature Description
Prompt-Based Sourcing Describe your ideal candidate — Gemini parses it into structured hiring filters
LinkedIn + Naukri Scraping Simulated web scraping from both platforms with source badges
Manual Filter Editing Fine-tune AI-generated rules via modal (experience, skills, location, exclusions)
Candidate Pipeline Shortlist/reject candidates, view match scores, Work DNA compatibility

🔌 Integrations

Provider Data Points
Slack Channel activity, message frequency, response time, focus hours
Microsoft Teams Meeting frequency, chat activity, org structure, calendar data
Microsoft Outlook Email volume, calendar density, response patterns, meeting conflicts
Discord Server activity, voice participation, thread engagement, sentiment
Google Workspace Gmail volume, calendar events, Drive activity, Meet usage

Integrations connect in-app with simulated OAuth flow (localStorage-persisted). For production Outlook, register an Azure AD app with Mail.Read + Calendars.Read Graph API permissions.

🏗️ System Architecture

graph TD
    subgraph Data Sources
        S[Slack/Teams]
        G[GitHub/GitLab]
        C[Calendar/Outlook]
        J[LinkedIn/Naukri]
        HR[HR Data<br>Leave, Attendance]
    end

    subgraph KarmaOS Engine
        A[Aggregator<br>15-min Blocks]
        ML[Classifier<br>Deep Work vs Coordination]
        RE[Rule Engine<br>Automation Rules]
        HA[Hiring Engine<br>Candidate Matching]
    end

    subgraph Intelligence Layer
        LLM((Google Gemini<br>Narrative + Parsing))
    end

    subgraph User Experience
        UI[Dashboard<br>Insights & Operations]
        AI[AI Assistant<br>Conversational]
        RF[React Flow<br>Visual Workflows]
    end

    S --> A
    G --> A
    C --> A
    J --> HA
    HR --> RE

    A --> ML
    ML --> LLM
    RE --> LLM
    HA --> LLM
    LLM --> UI
    LLM --> AI
    LLM --> RF

    style S fill:#36C5F0,stroke:#none,color:#fff
    style G fill:#24292e,stroke:#none,color:#fff
    style C fill:#0078D4,stroke:#none,color:#fff
    style J fill:#0A66C2,stroke:#none,color:#fff
    style HR fill:#10b981,stroke:#none,color:#fff
    style A fill:#2A2A35,stroke:#4f46e5,color:#fff,stroke-width:2px
    style ML fill:#2A2A35,stroke:#4f46e5,color:#fff,stroke-width:2px
    style RE fill:#2A2A35,stroke:#f59e0b,color:#fff,stroke-width:2px
    style HA fill:#2A2A35,stroke:#3b82f6,color:#fff,stroke-width:2px
    style LLM fill:#5E1B89,stroke:#d8b4fe,color:#fff,stroke-width:2px
    style UI fill:#18181B,stroke:#3f3f46,color:#fff
    style AI fill:#18181B,stroke:#3f3f46,color:#fff
    style RF fill:#18181B,stroke:#FF0072,color:#fff,stroke-width:2px
Loading

Data Flow

  1. Aggregates metadata events into 15-minute chronological blocks
  2. Classifies blocks into distinct states (Deep Work, Shallow Coordination, Friction)
  3. Prompts Gemini with aggregated summaries — narratives for insights, structured JSON for automation/hiring rules
  4. Evaluates automation and hiring rules against live workforce + HR data
  5. Renders everything in a premium, narrative-first UI

⚡ Getting Started

Important

A GEMINI_API_KEY is required for all AI features (dashboard insights, automation rule parsing, hiring rule creation, AI assistant).

Prerequisites

Installation & Setup

  1. Install dependencies:

    npm install

  2. Configure Environment Variables: Create or edit .env.local and add your API key:

    GEMINI_API_KEY=your_gemini_api_key_here

  3. Start the Development Server:

    npm run dev

    The app will be available at http://localhost:3000.

Tip

If port 3000 is busy, run npx kill-port 3000 before starting the server.

📁 Project Structure

├── server.ts                    # Express server — all APIs, Gemini calls, in-memory stores
├── src/
│   ├── App.tsx                  # Routing (12 routes)
│   ├── main.tsx                 # Entry point with ThemeProvider
│   ├── index.css                # Design system + dark mode variables
│   ├── contexts/
│   │   ├── AuthContext.tsx       # Firebase auth wrapper
│   │   └── ThemeContext.tsx      # Dark mode toggle (localStorage-backed)
│   ├── layouts/
│   │   └── DashboardLayout.tsx  # Sidebar nav, search, theme toggle
│   ├── pages/
│   │   ├── Dashboard.tsx        # System overview with AI insights
│   │   ├── TalentIntelligence.tsx  # Recruiter agent
│   │   ├── HiringAutomation.tsx # LinkedIn/Naukri sourcing rules
│   │   ├── TeamInsights.tsx     # Performance analysis
│   │   ├── EmployeeView.tsx     # Individual Work DNA profile
│   │   ├── LeaveManagement.tsx  # Leave apply/approve flow
│   │   ├── Attendance.tsx       # Check-in/out grid
│   │   ├── Automation.tsx       # React Flow workflow builder
│   │   ├── Settings.tsx         # Profile + integrations
│   │   └── LandingPage.tsx      # Public landing
│   ├── components/
│   │   └── AiAssistant.tsx      # Floating chat widget
│   └── lib/
│       ├── dataService.ts       # All API fetch wrappers (30+ functions)
│       ├── gemini.ts            # Client-side Gemini helpers
│       ├── firebase.ts          # Firebase config
│       └── utils.ts             # cn() class merge utility
├── data/
│   └── Employee_Data.xlsx       # Source employee dataset
└── .env.local                   # GEMINI_API_KEY

🛠️ Tech Stack

Layer Technology
Frontend React 18, Vite, Framer Motion
Language TypeScript (strict mode)
Styling Tailwind CSS, Lucide React icons
Intelligence Google Gemini 2.5 Flash API
Workflow Viz React Flow (@xyflow/react)
Server Express.js (custom dev server)
Auth Firebase Authentication (optional)
Data In-memory stores (MVP), Excel source data

🔐 Security Model

  • Gemini API key is server-side only — never exposed to the browser
  • All AI calls routed through Express endpoints in server.ts
  • Firebase auth gates dashboard access when configured
  • Integration tokens stored in localStorage (simulated OAuth for MVP)

📝 MVP Notes

Note

All HR data (leave, attendance, payroll) and automation rules are stored in-memory and reset on server restart. This is intentional for the MVP — the architecture cleanly separates data stores for easy database migration.

Note

LinkedIn/Naukri scraping is simulated with 20 seeded Indian tech market profiles. The API structure is production-ready for plugging in real scraping services (Proxycurl, RapidAPI, etc.).

🗺️ Roadmap

  • Persistent database (PostgreSQL/MongoDB)
  • Real OAuth flows for Slack, Teams, Outlook
  • Live LinkedIn/Naukri API integration
  • Scheduled automation rule execution (cron)
  • Email/Slack notification delivery for triggered rules
  • Role-based access control (Admin, Manager, Employee)
  • Employee self-service portal (own leave/attendance)

About

HR workforce

karma-pi-seven.vercel.app

Resources

Readme
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

Packages

Contributors

Languages