codex-context-engine

You can reduce AI hallucination by generating a project context genome that maps macro-level project understanding and module structures. This provides the AI with a concise project briefing at session start to ground its responses in actual codebase architecture.

Generate project context documentation by running the codex-genome command, which creates a `.codex/context/genome.md` file and optional per-module maps. This genome is then automatically loaded when present to brief the AI on your workspace architecture.

You should load project context before starting any task on a project with 50+ files, or when issuing an explicit genome command. The context auto-loads at session start if a genome file is already present in the workspace.

Generating per-module maps helps manage large project context by targeting specific relevant context areas without overwhelming the model. This modular approach keeps the AI focused on the exact files needed for the current task.

Yes, you can auto-load project context at the start of a coding session if the `.codex/context/genome.md` file is already generated. The system detects the genome file and automatically loads it to provide immediate project understanding.

A project context genome is a multi-layer documentation file located at `.codex/context/genome.md` that provides macro project understanding. It works by being automatically loaded into the AI context to ground the model in your actual codebase architecture and reduce hallucinated outputs.