Roxabi · Documentation
Documentation
Public, plain-spoken guides and references across the Roxabi portfolio — architecture, cognition, and the operating principles behind the open-source primitives. Organized by tag.
The Exocortex
Overview of the exocortex: the problem of siloed agent data, the living-memory architecture, and the ouroboros feedback loop.
Read →
Cognitive Framework
The three-layer model (D1 Searle / D2 Husserl / D3 Tulving) that gives Roxabi Cortex principled, predictive cognitive structure.
Read →
The Observation Contract
The typed contract that separates fast, source-aware encoding from slow, graph-aware memory consolidation.
Read →
Entity Taxonomy
The biomimetic data model behind the Cortex graph: referent axis, memory type axis, decay, and compiled truth.
Read →
Temporal Decay
The forgetting-curve model behind Cortex memory: decay formulas, reinforcement on access, and the thresholds that silence stale knowledge.
Read →
The Multi-Agent Audit
Sixty-seven parallel agents across eight domains, two-step drift verification, and one prioritized report — the read half of the factory, with two real recaps in the open.
Read →
Autonomous Remediation
A field report: an overnight Kimi audit, an Opus-derived epic, and a workflow that drove 22 findings to merged PRs autonomously — harness, scheduling, blockers, and cost.
Read →
Derived Reflexes
How well-formed beliefs at Dilts level 4 automatically generate correct behavior at levels 2–3, with nine engineering maxims derived from first principles.
Read →
The Roxabi Factory
An agent engine extended with a jobs layer — the architecture choices (worker vs tool, five tool layers, domains by nature) and the order we build them in.
Read →