Typically no. Upload the full archive.tar once at the start of a chat thread, then attach the smaller archive.diff.tar for subsequent turns. If you exhaust your context window, start a fresh chat, ask for a handoff in the old one, and paste that into the new chat along with the latest full archive.
STAN automatically excludes binary files from archives and prints a warning to the console. It also flags very large text files (by size or line count). You can add glob patterns to the excludes array in your stan.config.yml to ignore specific large files or directories you don't want included.
They are a portable, human-readable, and universally supported format for representing code changes. This makes them ideal for an AI-assisted workflow because they are auditable and tool-friendly. STAN’s FEEDBACK handshake provides a robust mechanism for automatically correcting patches that fail to apply.
Yes. The CLI is designed to be deterministic and scriptable. You can run stan run in a CI job to generate archives and text outputs, then upload them as build artifacts or use them in subsequent pipeline steps, such as automated documentation publishing or quality checks.
While STAN is packaged as an npm module with exports, its primary and supported interface is the CLI. For deep integration, you can consult the API reference on the docs site.