Internals

These pages explain how OntoCast works inside: the mechanisms behind the settings, the counters in the run telemetry, and the boundaries between components. Read them when you contribute to OntoCast, or when you tune a deployment beyond what the guides cover. They name modules and classes, which may change between releases.

Page	What it explains
How ontology retrieval works	How each content unit's ontology context is retrieved, scored and sized
Validation layers and the critic loop	The checks extracted ontology and facts pass through, and how critic and repair budgets are spent
How prompts are built	What goes into each render and critic prompt, and in which order
Telemetry counters	Every counter in budget and retrieval_metrics
How LLM responses are parsed	What happens to a model response before it becomes a result
Ontology catalog	Which component owns stored ontologies, how tenancy scopes them, and what a custom triple-store backend implements