AI agents can work autonomously by delegating tasks to different agents, models and tools. Various protocols are evolving to standardize the communication between agents. This post describes some p...
Why Agents and Agentic Systems can fail
At the ChatBot Summit in Berlin I gave a talk ‘How to understand why Agents fail’. Below is a transcript and the slides. As you all know there is a lot of innovation and buzz around agents and age...
Introduction to Docling
Docling is an open-source framework to extract information from PDFs, DOCX, tables, images and other document types into structured data for scenarios like Generative AI and agentic applications. T...
Leveraging Docling in Langflow
Docling and Langflow are two popular open-source frameworks to extract information from documents and to build agentic applications. This post describes how to leverage Docling within Langflow. La...
Benefits of IBM watsonx Orchestrate
My colleagues Volker Kuhm and Leonard Brucksch and I have written an article outlining the benefits of IBM watsonx Orchestrate titled ‘AI Agents for Business’. Below is our short version. For more...
Getting User Input in watsonx Orchestrate Agents
Watsonx Orchestrate is IBM’s platform to build and run agentic systems. Autonomous agents can automate tasks but often humans need to be in the loop. This post explains how Orchestrate agents can a...
Top five Reasons why to use watsonx Orchestrate Agents
Watsonx Orchestrate is IBM’s platform to build and run agentic systems, from structured flows to autonomous systems. You can start simple via the Orchestrate user interfaces, but as a developer you...
Creating Ground Truth Datasets for Agent Evaluations
Agentic applications can fail for multiple reasons. For developers it’s important to understand why they fail to improve the quality. To run evaluations, ground truth information is required which ...
Evaluating Agents in watsonx Orchestrate
There is a lot of buzz around autonomous agents. However, agents can fail. In these cases, it’s important for developers and AI engineers to understand why they fail to improve prompts, models and ...
Converting SSE to STDIO via the MCP Gateway
The Model Context Protocol (MCP) has quickly become the standard how tools can be invoked from agents. There are different transport mechanisms, but some clients only support STDIO at this point. T...