AI Capabilities
Inspect recommendations and embedding workflows while keeping models outside the system-of-record and verification boundaries.
Workspai can expose AI recommendations and embedding operations, but a model is never the authority for workspace identity, Graph truth, repair closure, or a verification verdict.
npx workspai ai info --json
npx workspai ai recommend --jsonInspect configuration and availability before invoking model-backed work. Recommendation output remains advisory and should carry its model, provider, scope, and evidence provenance.
npx workspai ai generate-embeddings --json
npx workspai ai update-embeddings --jsonEmbeddings are retrieval indexes, not the canonical Workspace Model or Knowledge Graph. They may be regenerated when their inputs or provider configuration changes without rewriting source facts.
Evaluation records model usage and verified outcomes without storing prompt or
response bodies. See ai,
ai recommend, and
workspace eval.
Observation, Feedback, and Evaluation
Observe workspace changes, record structured outcomes, and measure agent value without storing prompt or response bodies.
Infrastructure Operations
Plan, inspect, start, and stop governed infrastructure without confusing runtime provisioning with Workspace Intelligence evidence.