Your Personalized On-Device AI Companion Kouri is an intelligent, privacy-first, on-device AI assistant powered by the DeepSeek R1 7B model, handcrafted with purpose and personality by Xander. Designed to operate locally via Ollama, Kouri runs fully offline—preserving privacy, ensuring speed, and empowering personal ownership of AI.
✨ Core Philosophy Kouri isn’t just another chatbot. She’s a context-aware, emotionally tuned assistant that adapts to your personality, routines, and preferences—acting more like a digital companion than a tool.
On-device inference for low-latency, secure interactions (no cloud needed ☁️❌)
Dynamic prompt injection to personalize every response with tone, quirks, and user preferences
Modular intent classification to detect and respond to tasks like music playback, weather, alarms, and more
LLM-driven context understanding for natural, personality-rich conversations
🔧 Tech Stack 🧠 LLM: DeepSeek R1 7B running locally via Ollama
🐍 Backend: Python (subprocess for local model calls)
🧩 Intent Engine: Custom NLP classifier to route inputs
🧬 Personality Core: Dictionary-based traits injected into every prompt
🧠 Prompt Engineering: Structured prompts using YAML-style input for high LLM fidelity
💡 Kouri’s Capabilities 🤖 Fully offline, on-device AI interactions, Deep Reasoning
🗣️ Friendly and expressive communication with emoji-enriched tone
📱 Context-aware behavior (e.g., time-of-day mood shifting, user sentiment reflection)
🎯 Intent detection system for intelligent task execution
🧩 Modular design — easily extendable with new skills or moods
🔐 Privacy-First by Design Unlike cloud-based assistants, Kouri runs entirely on your machine. Your queries never leave your device, ensuring a fully private, secure AI experience—aligned with Apple’s values of personalized intelligence without compromise.
🌱 Future Vision 🔄 Memory modules to remember long-term user preferences, To have her own proprietary app and website, Option to be cloud-based (Requires Internet)
🧠 Lightweight fine-tuning using local conversation history
🪄 Contextual plugin system to interface with apps, automation tools, or APIs
🌐 Potential expansion into GUI-based systems using Tkinter, CustomTkinter, or SwiftUI
