The architecture surrounding a model: retrieval, identity, permissions, tools, monitoring and execution.
Each line below is retained only when a VOCE publication or formal research record supports the relationship.
Human systems technical architecture conditions action AI systems.
One component, control, identity layer, interface or system capability depends on or is constrained by another.
A documented dependency does not prove that the component is sufficient for security, compliance, reliability or system effectiveness.
AI systems contains Model.
One component, control, identity layer, interface or system capability depends on or is constrained by another.
A documented dependency does not prove that the component is sufficient for security, compliance, reliability or system effectiveness.
AI systems can include RAG.
One component, control, identity layer, interface or system capability depends on or is constrained by another.
A documented dependency does not prove that the component is sufficient for security, compliance, reliability or system effectiveness.
AI systems can orchestrate AI agents.
One component, control, identity layer, interface or system capability depends on or is constrained by another.
A documented dependency does not prove that the component is sufficient for security, compliance, reliability or system effectiveness.
AI systems must be governed beyond model choice AI governance.
A descriptive, taxonomic, historical, clinical or institutional relation directly supported by the cited record without claiming a stronger mechanism.
Do not generalize this relation beyond the population, system, period or documentary scope supported by the cited record.
VOCE Atlas is a curated knowledge graph. Co-occurrence alone does not create a published relationship. Every connection now carries a semantic class and an inference boundary that remain part of the record alongside its source.