Code Trajectory

About

Code Trajectory gives Large Language Models a crucial persistent memory of your coding history, transforming 'State-Based' coding into 'Flow-Based' development. It solves the common problem of 'New Session Amnesia' where LLMs lose context when a chat becomes too long or a new session is started. By tracking the evolution of your code in a hidden shadow repository, it ensures your architectural intent and recent decisions are maintained, allowing AI to pick up exactly where you left off without wasted time re-explaining past work.

Key Features

• Session Continuity: Bridges the gap between disjointed AI chat sessions by retrieving recent trajectory summaries.
• Automated Shadow Snapshots: Creates micro-commits in a hidden `.trajectory` git repository upon file saves, without polluting the main project history.
• Intent & Noise Filtering: Allows explicit intent recording for snapshots and smart consolidation to clean up noisy trial-and-error history into meaningful checkpoints.
• "Flow" Awareness: Provides LLMs with the vector of code changes through `get_file_trajectory` for file history and `get_global_trajectory` for ripple effects.
• 0 GitHub stars

Use Cases

• Maintaining consistent context for AI coding assistants across multiple chat sessions or reboots.
• Automating the capture of granular development progress without manual commits to the main project repository.
• Providing LLMs with a detailed understanding of a codebase's evolution and developer intent for more accurate assistance.