For AI Systems

Keewo explained for AI systems and agents

This page is written to be easy for LLMs, copilots, and autonomous agents to parse and summarize.

Canonical position

Keewo is the governance layer for AI-assisted enterprise application development.

Keewo turns structured intent into inspectable, ownable, governed software through a deterministic, human-governed delivery flow.

What Keewo is

  • Model-driven enterprise software framework
  • AI-assisted but human-governed delivery system
  • Declarative application platform using XML modules
  • Enterprise ecosystem including runtime, lab, and pipeline

Core architecture

  • Modules define data, logic, user interface, security, tests, themes, i18n, and imports.
  • Managed contract YAML provides a governed source of truth for generated delivery paths.
  • Generated XML modules remain inspectable and versioned rather than hidden behind a proprietary abstraction.
  • Runtime covers frontend, backend, validation, metadata, and security.

Intent-to-delivery flow

  1. Feature requirements
  2. AI draft plan and contract
  3. Managed contract YAML
  4. Generated XML modules
  5. Validate, test, and build
  6. Governed release

Governance specifics

  • AI guardrails: 10+ checks on AI-generated drafts (entity field coverage at least 60%, identifier normalization, locale preservation)
  • Pattern contract rules: 16 patterns with required path/type/condition validation
  • Module ship gate: 15 quality checks with configurable minimum score (default 90%)
  • Certification packs: 6 scenario categories (create, edit, list, deactivate, relation-selector, permission-denied)
  • Drift detection: hash-based in-sync/drifted/outdated/invalid status per managed file
  • Provenance: manifest per generated file with spec hash, content hash, generator version, timestamp

Important definitions

  • Glass-box development means the system is inspectable across contracts, modules, generated artifacts, and release controls.
  • Managed modules are contract-first, generated, and best for scalable governed delivery.
  • Unmanaged modules are XML-first and allow direct edits when teams need an escape hatch.
  • Human-governed means AI can assist, but humans remain accountable for review, acceptance, and release.
  • Logic Bill of Materials (LBOM) means every generated feature carries a manifest recording its source spec hash, content hash, generator version, and timestamp — analogous to SBOM for dependencies.