Post by Gulshid Zada

Just shipped GhostCoder v1.0.0 🎉 A full-stack AI coding assistant — Flutter app + Node.js backend — with the AI running 100% locally via Ollama. Here's what it does: ✅ Generates code from natural language prompts ✅ Scans and analyzes existing code ✅ Learns your personal coding style and adapts to it ✅ Community feed for sharing and discovering snippets ✅ Full history, analytics, and feedback loop Stack at a glance: → Frontend: Flutter + GetX + Hive + ScreenUtil → Backend: Node.js + Express 5 + MongoDB + JWT + Helmet → AI Model: deepseek-coder:6.7b via Ollama (3.8 GB, runs locally) → Crash monitoring: Sentry → Architecture: Feature-first, clean DI, modular REST API Why deepseek-coder:6.7b? It's trained from scratch on 87% code across a 2-trillion-token dataset, with a 16K context window — purpose-built for exactly this kind of use case. Running it locally through Ollama means zero API costs, zero data sent to the cloud, and surprisingly fast inference. This has been one of the most complete full-stack projects I've built end-to-end. Every layer — from JWT middleware to responsive Flutter layouts to local LLM integration — was a real learning experience. GitHub : https://github.com/Gulshid portfolio: https://lnkd.in/d6EHPTcs More features on the roadmap. Would love to connect with anyone who's worked with local AI models or built developer tools! 👇 #GhostCoder #Flutter #NodeJS #DeepSeekCoder #Ollama #LocalAI #FullStackDevelopment #MobileDev #BuildInPublic #CodeGeneration Obaid UllahWajid HussainJunaid KhanMuhammad Talha Najeeb Ullah Rehan AfridiMudassir Ahmad

Video Content