Skip to content

mangalgithub/jarvis

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

41 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Jarvis Personal OS 🤖

The Ultimate Production-Grade Personal AI Operating System.

Jarvis is a sophisticated, multimodal AI assistant designed to manage your life with high-fidelity intelligence, deterministic accuracy, and stateful memory. Jarvis moves beyond simple chat and implements complex multi-turn agency, robust logic-NLP decoupling, and real-time situational awareness.


🏥 High-Fidelity Health OS (V4) 🏆

Jarvis implements an elite health tracking pipeline that prioritizes accuracy and interactive reliability.

  • Deterministic Math Engine: Moves away from LLM-based "guessing." NLP is used only for entity extraction, while a pure Python backend handles all mathematical computations using a curated Food Database.
  • Priority Entity Matching: Implements "Longest-Match-First" logic to ensure composite dishes (e.g., "Chicken Curry") are matched before raw ingredients ("Chicken"), preventing protein overestimation.
  • Interactive Clarification Loop: Jarvis identifies vague inputs (e.g., "I had some chicken") and pauses the log to ask for clarification, ensuring 100% data integrity.
  • Context-Aware Multipliers: Automatically detects preparation context (Restaurant vs. Home) and applies caloric multipliers for hidden fats/oils.
  • Protein Density Guardrails: Built-in sanity checks that flag physically impossible macro claims (e.g., >35% protein density in cooked meals).

🧠 Contextual Memory & Stateful Session Management

  • Short-Term Session State: Powered by a MongoDB-backed ConversationStateService, Jarvis can "remember" pending actions across multiple turns.
  • Message Re-Hydration: Jarvis "stitches" user answers back into original commands (e.g., "I had some chicken" + "1 bowl""I had 1 bowl of chicken").
  • Temporal Awareness: Real-time clock injection ensures Jarvis knows the time and date, providing situational greetings and accurate log timestamps.
  • Deep RAG Integration: Long-term memory storage of user preferences, habits, and constraints for personalized advice.

👁️ Vision Intelligence (Multimodal)

  • AI Nutritionist: Snap a photo of your food. Jarvis identifies the meal and automatically logs deterministic calories and protein.
  • Smart Receipts: Show Jarvis a shopping bill or restaurant receipt. He extracts the total and items to log your expenses instantly.
  • Visual Context: Jarvis "sees" what you see, allowing for natural conversations about images.

💰 Advanced Finance Agent

  • Natural Language Logging: "Spent 500 on dinner at Pizza Hut" — Jarvis handles the rest.
  • Categorization: Automatically sorts spending into Food, Travel, Bills, Shopping, etc.
  • Budgets & Goals: Set monthly limits and track progress toward major savings goals.

⚡ Technical Superpowers

  • Multi-Agent Orchestrator: A central brain that handles intent classification and routes requests to specialized agents (Health, Finance, News, etc.).
  • Logic-NLP Decoupling: Core architectural pattern that uses LLMs for understanding and pure code for execution, ensuring zero-hallucination math.
  • Real-time Sync: WebSockets ensure your dashboard updates instantly across all devices.

🛠️ Tech Stack

  • Frontend: Next.js 15, Tailwind CSS, Framer Motion, HeroUI, Recharts.
  • Backend: Node.js, Express, Socket.io, MongoDB Atlas.
  • AI Core: Python, FastAPI, Groq (Llama 3), Gemini 2.0 Flash, MongoDB.
  • Infrastructure: Vercel (Web), Hugging Face (Dockerized Agents).

📦 Deployment Guide

  1. GitHub: Connect your repo to Vercel for the Frontend/Backend.
  2. Hugging Face: Create a Docker Space for the agents/ folder.
  3. Environment: Ensure GOOGLE_API_KEY, GROQ_API_KEY, and MONGODB_URI are set.

Created with ❤️ by Mangal Gupta.

About

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors