Habita
Habita is an AI-powered health coach that helps users build better habits through dynamic, LLM-driven conversations. It provides personalized goal-setting, action plans, and real-time coaching, adapting responses based on past interactions for a fully engaging experience.
Demo
About
Habita is an LLM-powered AI health coach designed to help users build and sustain healthier habits. Unlike traditional habit trackers, Habita actively guides users through goal-setting, action planning, and progress tracking—all in a conversational, adaptive manner. Key features include: - AI-Powered Coaching: The bot dynamically generates responses tailored to the user’s health goals and progress. - Memory & Context Awareness: Stores conversation history to provide personalized, coherent coaching over time. - Motivational & Actionable Feedback: Encourages users with small, achievable steps to improve health. - Real-Time Adaptability: Moves beyond rigid scripted flows, responding intelligently to user input. By leveraging AI for real-time habit coaching, Habita provides an engaging, human-like experience to keep users accountable and motivated.
How it's made
Habita is built using a fully LLM-driven architecture, eliminating rigid state-based flows. 🔧 Tech Stack: - LLM: Groq API (for adaptive, real-time coaching responses) - Backend: Node.js (TypeScript) - Database: PostgreSQL (stores full conversation history for long-term memory) - Messaging: Telegram Bot API (integrates AI coaching into a chat interface) 🔑 Key Implementation Details: - Dynamic Conversation Engine: All responses are generated on-the-fly using Groq, making the coaching experience highly adaptive. - Memory-Driven Context: Instead of states, Habita retrieves past conversations, allowing AI to track user progress naturally. - Optimized Prompt Engineering: Carefully structured prompts guide the AI to act as a real coach, providing personalized, motivational feedback. This hacky but effective approach allowed for a quick, highly flexible health coaching system without needing predefined conversation flows.