Runchat Updates: May 16, 2025

🧨 EXPLODE RUNCHAT NODES

• It was always possible to open a runchat node in a separate canvas, now you can "explode" a runchat to view it directly in the context of your current workflow
• Exploded runchats are grouped so they are easy to move around, and parent groups resize to fit content (most of the time)
• You can still run the parent runchat node, or any children within it.

🛠️ CUSTOMIZE RUNCHATS

• you can make changes to runchats that override the version published in libraries. This makes it super fast to make tweaks to runchats for a specific use case
• you can change node values as well as add or remove nodes

🔍 INSPECT RUNCHATS

• When you run a runchat node all child nodes on the canvas are updated with the new values. This makes it super easy to inspect exactly that is happening when you work with runchat nodes that might themselves contain many nested runchat nodes.

We plan to update how tool calling works in runchat to use this pattern so that it will be possible to inspect and edit llm tool calls as runchats. This will open up use cases like using the prompt node to create a runchat for a particular task, using nested agents to create complex workflows that you can then fully customize and configure, reviewing raw inputs to responses to complex queries and more.