We've just released GhostChat v2.0, a complete reimagining of AI chat applications with a local-first architecture. Unlike traditional cloud-dependent chat apps, v2.0 stores all conversations locally using IndexedDB with localStorage fallback, works completely offline (except for AI API calls), and gives users complete control over their data.
Key technical features:
- Zero external dependencies for core functionality
- Real-time UI with Zustand state management
- One-click export/import of all conversations
- Session management with custom AI prompts
- Privacy by default with optional cloud sync
- Complete TypeScript implementation with Next.js 15
The repository offers both the new local-first version (v2.0) and the traditional cloud-first version (v1.0) as production-ready templates.
This represents a shift toward "sovereign" AI applications where privacy and user control aren't afterthoughts but foundational principles. Would love to hear the community's thoughts on local-first AI architectures and whether this approach resonates with your privacy and performance needs.
Live demo available at the v2.0 branch. Full documentation in V2_README.md.
We've just released GhostChat v2.0, a complete reimagining of AI chat applications with a local-first architecture. Unlike traditional cloud-dependent chat apps, v2.0 stores all conversations locally using IndexedDB with localStorage fallback, works completely offline (except for AI API calls), and gives users complete control over their data.
Key technical features: - Zero external dependencies for core functionality - Real-time UI with Zustand state management - One-click export/import of all conversations - Session management with custom AI prompts - Privacy by default with optional cloud sync - Complete TypeScript implementation with Next.js 15
The repository offers both the new local-first version (v2.0) and the traditional cloud-first version (v1.0) as production-ready templates.
This represents a shift toward "sovereign" AI applications where privacy and user control aren't afterthoughts but foundational principles. Would love to hear the community's thoughts on local-first AI architectures and whether this approach resonates with your privacy and performance needs.
Live demo available at the v2.0 branch. Full documentation in V2_README.md.