Developer Reference
Explore command specs, system design models, integration workflows, and traversal filters for our agent infrastructure tooling.
1. Core Design & Architecture
Unlike traditional codebase packagers that buffer full directories or file arrays in RAM, Kord operates as a streaming ingestion pipeline. The filesystem traversal loop is coupled directly to the output serializing encoder.
Streaming Ingestion Pipeline Flow
Constant Memory Footprint
Memory consumption remains flat regardless of repository size. XML and JSON elements are serialized token-by-token directly to standard output or the targeted file, completely avoiding buffering full file maps in RAM.
Zero-Dependency Engine
Kord is constructed entirely using the Go standard library (encoding/xml, path/filepath, io), keeping compilation trivial and compile-time artifacts clean.
Stream Formatting
Output contents (specifically source code and plain text files) are automatically wrapped inside CDATA blocks for XML or string-encoded for JSON to prevent escaping issues and ensure parsing fidelity by downstream LLM endpoints.