AI Engineering
Status: Canonical section
Last updated: 2026-07-05
Purpose
Complete reference for ZAIXOS AI-native engineering — lifecycles, closed loop, evaluation, and governance.
Section map
| Document | Topic |
|---|---|
| agent-lifecycle.md | Agent create → run → retire |
| conversation-lifecycle.md | Turn-by-turn flow |
| intent-understanding.md | Intent taxonomy and resolution |
| signals.md | Behavior and capability signals |
| capability-matching.md | Intent → capability routing |
| planning.md | Goal and task planners |
| execution.md | Tool execution |
| reasoning.md | LLM vs deterministic paths |
| memory-and-knowledge.md | PL-003 vs PL-005 |
| learning.md | PL-004 learning loops |
| benchmark.md | Dataset evaluation |
| regression.md | Continuous regression |
| acceptance-lab.md | Human-gated promotion |
| release.md | AI release process |
| version-registry.md | Prompt/model registry |
| prompt-versioning.md | Prompt semver |
| closed-loop.md | End-to-end quality loop |
| human-approval.md | Mandatory approval gates |
Stack diagram
mermaid
flowchart TB
subgraph Product["Product Layer"]
UI[Chat / Filament UI]
ADP[AI Adapter Module]
TOOLS[Domain Tools]
end
subgraph PL003["PL-003 Runtime"]
AG[AgentRuntime]
INT[Intent]
PL[Planner]
EX[Tool Execution]
end
subgraph PL004["PL-004 Intelligence"]
VR[Version Registry]
LR[Learning]
end
subgraph PL006["PL-006 Observability"]
BM[Benchmark]
ACC[Acceptance]
end
UI --> ADP --> AG
AG --> INT --> PL --> EX --> TOOLS
AG --> VR
BM --> ACC --> VR
LR --> VRRelated
Breadcrumbs: Home → AI