Glossary: decode the terminology
Keep models, interfaces, integrations and workflows distinct.
Core terms
| Term | Meaning in this guide |
|---|---|
| TypeSafe | The provider of Jev and its developer interfaces |
| Jev | A model for bounded decisions |
| System One | Product terminology for this decision-model/interface family, not a claim to reproduce all human cognition |
| State | Evidence available to the current request |
| Choice | Selection among defined candidates |
| Score | Evaluation on a defined ordered rubric |
| Noul | A number from 0 to 1 representing probability of yes |
| Probability | A model signal for a particular answer |
| Confidence | A derived Choice / Score distribution signal, not independently validated accuracy |
Integration terms
| Term | Practical meaning |
|---|---|
| API | An interface for software calls, usually requiring provider credentials |
| API key | An authentication secret to protect like a password |
| SDK | A library wrapping API calls, not another model |
| Model ID | The requested model; fixed versions differ from moving aliases |
| Skill | Agent-readable instructions and resources, not automatic authorization |
| MCP | A protocol for connecting agents to tools; a server still needs client configuration |
| Provider / gateway | An access channel with its own authentication, pricing and interface |
Engineering terms
RAG combines retrieval with answer generation. Reranking reorders retrieved candidates. A guardrail checks or constrains behavior; model-based checks should not be the sole security boundary. A fallback is an explicit alternative path for failures or uncertainty. Calibration uses labeled data to examine or adjust the relationship between model signals, thresholds and observed performance.
These are introductory explanations. Use your provider and SDK version's documentation for implementation details.
Sources and verification boundary
Reviewed September 19, 2026. Documentation-based guidance, not a live API or business-performance test.
TypeSafe · Introduction →TypeSafe · Choice →TypeSafe · Score →TypeSafe · Noul →TypeSafe · Confidence →TypeSafe · Agent skill →Reading progress stays in this browser