Show HN: Engrim – A universal, local-first SQLite memory engine for AI CLIs (github.com)

Engrim has unveiled a groundbreaking SQLite memory engine designed specifically for AI command-line interfaces (CLIs). This local-first, project-scoped engine empowers developers to seamlessly switch between various AI models and environments—such as Google Antigravity, Claude Code, and Cursor—without losing important project context or architectural decisions. By addressing the challenges of attention dilution as context windows expand, Engrim allows for the maintenance of a curated episodic memory with a capacity to store around 4,000 characters, drastically reducing the token costs associated with memory retrieval. The significance of Engrim lies in its ability to preserve project intelligence across multiple AI environments, effectively decoupling user constraints and architectural decisions from specific vendor platforms. Tested extensively in a live algorithmic trading environment, Engrim achieved a remarkable reduction in context costs—over 99%—while ensuring zero context amnesia during model transitions. Its hybrid memory retrieval system combines SQLite's full-text search capabilities and vector embeddings, optimizing performance while keeping all memory data local, secure, and offline. This innovation offers a new level of flexibility and efficiency for AI developers looking to leverage multiple models in complex projects.