Getting started
DaRetriva turns your documents into a searchable, cited knowledge base. Here are the core building blocks you'll see across the app and the API.
Organization
Your account's top-level tenant — users, datasets, agents, billing and settings all live inside one organization. Every piece of data is scoped to it; nothing is ever shared across organizations.
Dataset
A named collection of documents (uploaded files or content synced from a connector). Datasets are the unit you attach to an AI agent to control exactly what it can answer from.
Data source connector
Pulls content into a dataset automatically — a web crawler, or an integration like Google Drive, Confluence, or SharePoint. Connectors sync on a schedule (or on demand) and keep the dataset up to date as source content changes.
Document & embeddings
Each document is chunked and turned into vector embeddings at upload/sync time. Search and chat both work by finding the most relevant chunks for a query, then grounding the AI's answer in that retrieved text — this is what "RAG" (Retrieval-Augmented Generation) means in practice.
AI agent
A configured assistant: a system prompt, model settings, and one or more datasets it can retrieve from. Your team chats with agents directly in the app, and an agent can also be exposed publicly as an embeddable chat widget.
Embeddable widget
A public, unauthenticated chat bubble you can drop onto any website with one script tag, backed by one of your agents. See the Widget Integration guide for the full setup.