Concepts

This section explains the key ideas, design decisions, and internal architecture behind KITT. These pages are intended to help you understand why KITT works the way it does, not just how to use it.

If you are looking for step-by-step instructions, see the Guides section. If you need command-line reference, see the CLI Reference.

Topics

Architecture

How KITT is structured: the engine plugin system, Docker-based container management, the sibling container pattern, and the overall project layout.

Hardware Fingerprinting

How KITT uniquely identifies the hardware it runs on, the fingerprint format, detection methods for GPUs, CPUs, RAM, and storage, and the supported environment types.

KARR — Results Storage

KARR (Kitt's AI Results Repository) is KITT's results storage system. Covers the database backend (SQLite / PostgreSQL), the hybrid data model, schema migrations, and the evolution from flat files through Git-backed storage to the current database architecture.

Engine Lifecycle

The full lifecycle of an inference engine container: image pull, container creation, health checking with exponential backoff, benchmark execution, GPU memory tracking, and cleanup.

Benchmark System

The benchmark plugin architecture, built-in performance and quality benchmarks, YAML-defined custom benchmarks, checkpoint recovery, and suite orchestration.

Security

Mutual TLS for agent-server communication, automatic certificate generation, bearer token authentication, and development-mode options.