Introducing Agent Search: The Next Evolution in Enterprise Knowledge Discovery

At Onyx, we're on a mission to make enterprise knowledge more accessible and actionable. While our existing search capabilities have helped teams find relevant documents across their knowledge bases, we've noticed that many workplace questions require more than just document retrieval–they need a more comprehensive analysis and require a synthesis of disparate pieces of information..

Today, we're excited to announce Agent Search, a powerful new feature that transforms how teams interact with their institutional knowledge.

What is Agent Search?

Agent Search is an AI-powered advanced search capability that approaches questions the way your most knowledgeable colleague would. Instead of simply matching keywords, it:

• Breaks down complex questions into more manageable and narrow sub-questions
• Explores multiple sub-questions in parallel
• Synthesizes findings into comprehensive, contextual answers
• Refines responses to ensure comprehensiveness

For example, when asked "How did we close [customer] so quickly?", Agent Search doesn't just search for [customer]-related documents. It may explore the [customer]'s sales process, compare it with typical sales cycles, investigate any unique aspects like PoC processes, and synthesize this information into a coherent explanation.

How Agent Search Works

Compared to existing Onyx search, Agent search can:

• Enable answers to more complex, ambiguous questions
• Improve answer quality, particularly when multiple entities are involved
• Provide richer context around key aspects of the question
• Scale effectively across large document collections

All of the above could happen because Agent Search operates through a sophisticated multi-step process:

1. Initial Context Gathering: When you ask a question, Agent Search first performs a broad search to understand the context and relevant entities.
2. Question Decomposition: The agent breaks down complex queries into focused sub-questions that can be answered more precisely.
3. Parallel Investigation: Multiple searches run simultaneously to gather comprehensive information.
4. Answer Synthesis: The agent combines findings from all sources into an initial response.
5. Refinement: The answer is refined by performing additional searches to fill gaps or resolve

Get Started with Agent Search

Agent Search is available now for all Onyx deployments. Here's how you can start using it:

1. Onyx Cloud Users: Agent Search is automatically enabled in any assistants with search capability. Simply start asking more complex questions to see it in action.
2. Self-hosted Users: Update to the latest version of Onyx through our GitHub repository. Check our documentation for configuration options.
3. Enterprise Deployment: If you are an existing customer, simply update to the newest version to get access to Agent search. If you are new, Book a demo call today to discuss your needs.

Agent Search represents a significant step forward in making enterprise knowledge more accessible and actionable. We're excited to see how teams will use this capability to work more efficiently and make better-informed decisions.

Visit our documentation to learn more about Agent Search and join our community slack, discord #agent-search to discuss your experience with this new feature.