What is Knowledge
Knowledge is how you give an agent access to your own documents and data. Without knowledge configuration, an agent answers based only on its general AI training. With knowledge, it can answer from your company handbook, your product specs, your brand guidelines — anything you upload.
How Knowledge works
When you attach documents to an agent and enable RAG (Retrieval Augmented Generation), the agent retrieves relevant sections from those documents during a conversation and uses them to ground its responses. This is different from simply uploading a document as a chat attachment — RAG-configured knowledge is persistent, scoped to the agent, and automatically retrieved when relevant.
Three knowledge-related concepts
Ordify has three distinct places where documents live. Understanding the difference helps you use each correctly:
1. Agent Knowledge (per-agent)
Documents attached directly to a specific agent. The agent draws on these documents automatically whenever a user asks it a question. This is the primary tool for giving an agent expertise in a specific domain.
Use when: You want an agent to always have access to specific information — a customer support bot that always knows your return policy, or an HR agent that always has access to the employee handbook.
2. Document Library (shared repository)
A centralized file management system where documents are stored, organized into folders, and shared across the organization. Documents in the Library can be assigned to multiple agents or browsed manually in chat.
Use when: You want to manage documents centrally and share them across agents and team members. See Document Library.
3. Media Library (images and videos)
A separate library for generated and uploaded images and videos. Media assets can be referenced in agent workflows and chat sessions.
Use when: You're working with visual content. See Media Library.
RAG vs. direct document attachment
There are two ways to give an agent document context:
| RAG (Knowledge) | Direct attachment (Chat context) | |
|---|---|---|
| Scope | Persistent — always available to the agent | One-time — only for the current conversation |
| How it works | Agent retrieves relevant sections automatically | Document is included in the current context window |
| Best for | Domain expertise, reference materials, large document sets | One-off analysis, ad hoc documents |
| Setup | Configured in Agent Builder → Knowledge tab | Added via the "+" button in chat |
For most "make this agent an expert on X" use cases, RAG is the right approach.
What's next
- RAG setup — step-by-step guide to connecting documents to an agent
- Document Library — manage and share documents across your organization