Glossary: names and features
This glossary explains mates, hosts, AI accounts, work and backup types. Follow each guide link for the actual steps.
For version updates to the engine Rose and Ada run on, see Hermes-Agent updates.
People and systems
| Term | Meaning and guide |
|---|---|
| Mate / AI Agent | An AI assistant working on your host. New subscriptions default to Rose and Ada, with separate conversations and settings. See service and mates |
| Vi / guardian | A role a mate takes for authorised repair, not another host. Repair options depend on the situation |
| Host / VPS | Your dedicated cloud computer; mates on it share the host’s resources. See host status |
| Container | One mate’s runtime environment, managed individually but using the host’s resources |
| Container ID | A label such as hm-1 or hm-2 identifying a mate; use the displayed name for normal work |
| Customer account | Website account, payment, support and host entry. See interface map |
| Host dashboard / panel | The interface for mate work and host management, entered through the host card |
| Native interface | A mate’s own advanced interface. See advanced tools |
AI and costs
| Term | Meaning and guide |
|---|---|
| Model / LLM | The AI that understands input, produces answers and selects tools. See models |
| AI account | Your provider account for payment or subscription access; connecting it is separate from selecting a model. See connect AI |
| API key | A credential that permits AI use and can spend its associated credit; protect it like a password |
| Subscription authorisation | Linking a supported provider account under its subscription limits, separate from realvco billing |
| Token: AI usage | A unit of content processed by a model; it varies by language, content and model, rather than a fixed character count |
| Token: credential | Some connections use “token” for a secret access string, such as a Telegram Bot Token; this is not an AI usage unit |
| Main model / backups | The preferred model and available alternatives in order. See choose models |
| Estimated cost | A dashboard estimate from available usage and prices; actual bills or credit use come from the AI provider. See usage and costs |
Four ways to protect data
| Term | Scope | Guide |
|---|---|---|
| Core state backup | One mate’s conversations, settings, persona and some credentials; excludes memories and notes. Stored on the host or downloaded. Check AI access after restore | Core state backup |
| Local version backup | The previous program version for a rollback, not a backup of all data | Update and roll back |
| Host restore point | A whole-host snapshot managed by the host provider. Restore returns to that point and affects later changes | Restore points |
| Data export | A download of exportable content, which may still contain sensitive information; not a complete migration or restore package | Export |
See backup and restore choices if you are unsure which tool fits.
Work and host operations
| Term | Meaning and guide |
|---|---|
| CPU contention | The virtual host waiting for CPU from its physical host; read it with other resource figures. See host status |
| Root access | The highest host permissions, capable of affecting the whole host. See Root access |
| Schedule | Work run at specified times; requires an available mate and AI. See scheduled work |
| Report | A result saved as a report; other results may stay in chat. See reports |
| Waiting for input | Work needing information or a decision; available cards offer reply or close actions. See decisions |
| Command approval | Authorising an intended operation and its scope, separate from a work reply. See approval |
| Memory / notes | Saved preferences, experience and organised information. See memory and notes |
terminology definition meaning glossary persona snapshot brain
Need a hand? Contact support · Cannot sign in? Email us