bmad-document-project

To generate architecture documentation for a brownfield project, you can scan the codebase to automatically discover, classify, and output structured architecture, API contracts, and data models. The process uses configurable scan depths and template-driven outputs written directly to a project knowledge folder.

Yes, you can resume an interrupted documentation scan on a large repository because the process implements resumable state tracking. It detects existing state, archives older runs, and stores incremental progress to enable long-running analyses without losing data.

The best way to document API contracts and data models for a monolith codebase is using an automated deep-dive scan. This method performs exhaustive full-file reviews to produce file inventories, dependency graphs, data-flow analyses, and structured markdown artifacts.

Yes, brownfield project documentation supports quick scans for pattern-only analysis. You can choose between quick, deep, and exhaustive scan modes to control the depth of file analysis and documentation generation based on your immediate needs.

To handle exhaustive scans for multi-part repositories without timing out, the documentation process uses per-subfolder batching and write-as-you-go outputs. This approach breaks down large scans into manageable chunks and saves progress incrementally.