Article Type: Conceptual Overview
Audience: Workspace Administrators, Builders, TeamAI Users
Last Updated: January 2026
Overview
Knowledge (Data Hubs) enables you to augment AI agents with contextual information by attaching collections, documents, and files. This feature provides greater flexibility than comparable platforms, allowing extensive knowledge integration for more accurate, context-aware agent responses.
Learning Objectives:
Understand the Knowledge (Data Hubs) feature and its architecture
Learn how knowledge processing works through chunking and embeddings
Distinguish between semantic search and document review methods
Identify optimal use cases for Knowledge implementation
Apply best practices for knowledge management and agent configuration
What is Knowledge?
Knowledge, or Data Hubs, is a feature in TeamAI that enables builders to attach multiple collections, documents, and files to an agent. Unlike ChatGPT, which only allows attaching up to 20 files to a GPT, TeamAI provides greater flexibility in the number of files you can include.
How Knowledge Works
The Knowledge Processing Pipeline
Selection: Use the agent configure editor tab to select desired collections or documents
Chunking: The agent breaks text into manageable chunks
Embedding: Creates mathematical representations (embeddings) of the text
Storage: Stores embeddings for later retrieval during user interactions
Result: When users interact with your agent, it accesses uploaded files to obtain additional context and augment queries.
Query Processing Methods
The agent automatically selects the optimal method based on prompt requirements:
Method | Use Case | Process |
Semantic Search | Q&A style prompts requiring specific document portions | Returns relevant text chunks from source material |
Document Review | Complex queries needing full context | Returns entire short documents or relevant excerpts from larger documents |
Result: The agent includes retrieved information with the prompt to generate contextually relevant responses.
The Knowledge Access Process
Query Submission: You ask your agent a question related to information in your data hubs
Knowledge Search: The agent searches through selected data hubs to find relevant information
Context Integration: The agent uses retrieved information to enhance its understanding
Response Generation: The agent provides a more accurate, contextually relevant response
Result: Knowledge-enabled agents unlock full potential by leveraging organizational information for valuable insights and assistance.
When to use Knowledge
Best suited for applications where context changes infrequently:
Employee handbooks and policy documents
School curricula and educational materials
Product documentation and technical specifications
Company standard operating procedures
Compliance and regulatory guidelines
Result: Knowledge provides stable, reliable context for agents handling evergreen content.
Tip: For frequently changing information, consider enabling web search tools to supplement knowledge bases.
Best Practices
Prioritize Knowledge: Configure agents to rely on Knowledge first before searching the internet or using custom plugins
Enable Source Citation: Set agents to cite their sources when providing information from data hubs to increase transparency and trust
Select Relevant Collections Only: Choose only collections and data hubs relevant to agent tasks to keep knowledge focused and improve response accuracy
Maintain Current Information: Regularly review and update data hubs to ensure agents have the most up-to-date information
Secure Sensitive Data: For sensitive information, create separate collections specifically for agent use rather than granting access to entire library
Structure Hierarchically: Organize knowledge into logical collections and datastores to improve retrieval efficiency
Test Knowledge Access: Validate that agents can properly access and cite knowledge before full deployment
Monitor Performance: Track agent responses to identify knowledge gaps or outdated information
Troubleshooting & FAQ
Q: How many files can I attach to a TeamAI agent?
A: TeamAI provides flexible knowledge integration without the restrictive file limits found in other platforms. You can attach multiple collections and datastores.
Q: What is the difference between semantic search and document review?
A: Semantic search returns specific text chunks for Q&A style queries. Document review returns entire documents or large excerpts for complex context requirements.
Q: When should I use Knowledge versus web search?
A: Use Knowledge for stable, internal documentation. Use web search for current events, market data, or information that changes frequently.
Q: Can I restrict which collections an agent can access?
A: Yes, in the agent configuration editor, select only the specific collections and datastores relevant to that agent's purpose.
Q: How often should I update my knowledge bases?
A: Review quarterly for policy documents, monthly for product information, and immediately when significant changes occur.
Q: What file types are supported for knowledge upload?
A: TeamAI supports CSV, TXT, MD, PDF, DOCX, and many others. See "Supported Document Types in TeamAI's Data Hubs" for a complete list.
Q: How do I know if my agent is using knowledge or web search?
A: Enable source citation in agent instructions. Knowledge citations reference specific documents; web search citations reference URLs.
Q: Can agents access knowledge in real-time?
A: Yes, agents query knowledge bases during each interaction to provide current, context-aware responses.