| Project | The top-level container for one practice, course or product. Holds knowledge bases, agents and their conversations. Your first project is created when you sign up. |
| Knowledge base | The materials an agent answers from: documents and transcribed recordings, optionally in folders. A project can have several. |
| Assistant (agent) | The assistant itself: a model, instructions and memory, answering from one knowledge base. Listed under Assistants; some screens call it an agent. |
| Instructions | Text that tells an agent who it speaks for, its tone and what it must never do. Shapes how it answers, not what it knows. |
| Conversation memory | How long an agent remembers a conversation with one person, from none to 24 hours. |
| Model | The underlying AI model an agent uses to write answers. Smarter models cost more credits per answer. |
| Channel | A way people reach an agent: Telegram, WhatsApp, a website widget, or the API. |
| Contact | A person who has written to one of your agents. |
| Blocked contact | A contact you blocked: their messages are saved but never reach the agent, and cost no credits. See Block a contact. |
| Conversation | All messages between one contact and an agent, shown in Inbox. |
| Transcription | The text AOVA makes from an uploaded audio or video recording. The agent answers from the transcript; the original recording isn't kept. See Add audio and video. |
| Buttons under the answer | Up to 3 links set on a file. Clients see them whenever an answer uses that file. |
| Credits | The unit usage is billed in. Spent on answers (depending on the model), speech to text (8.8 credits per minute) and voice replies (0.8–6.1 credits per minute), from one balance per account. |
| Top-up | Credits bought separately from a plan. Used after plan credits. |