Building Smarter Agents in KNIME: New Tools, Live Demo, and Workflow Deep Dive

New Tools: RAG, DB Query and an improved web search for my AI Personal Assistant!
Sub: / @lowcodeaiforthewin-fx2bo
In this video, we continue the series on exploring agentic systems in KNIME. I share updates on the progress made since the first video, highlighting improvements, new tools, and a live demonstration of a personal assistant agent in action. Key advancements include enhancing the web search tool with Python libraries like DuckDuckGo Search and Goose3, as well as developing two new tools: a retrieval-augmented generation (RAG) tool for interacting with documents and a database query tool for answering natural language questions using structured data.
The centerpiece of this video is a fully integrated personal assistant agent that dynamically delegates tasks to specialized agents. You'll see how the agent orchestrator decides which agent to activate-be it for web search, database queries, or document interaction-and handles user questions seamlessly. I walk you through a live demonstration of querying financial data, searching for information on recent events, and generating Python extension development tips using the new tools.
I also dive into the technical side, showing how agent workflows are defined, emphasizing the use of structured prompts in XML-like formats, and explaining the logic behind recursive loops, state variables, and message histories. Finally, I provide a high-level breakdown of the generic agent workflow, which forms the backbone of this system.
If you're interested in AI-driven tools, prompt engineering, or building agent-based systems, this video offers practical insights and inspiration for your own projects.