You can build a complete RAG workflow visually. Document chunking, retrieval, generation, response formatting. All accessible through the UI. The visual builder handles the flow and data mapping automatically between steps.
Custom code is optional, not required. You might use a custom code node to handle document preprocessing if your files have weird formatting. But for 80% of use cases, the standard nodes do the job.
Prompt engineering lives in the generation node itself. You write and test prompts right there. Error handling is built into the nodes—retries, fallbacks, all configurable from the UI.
Non-technical users ship complete RAG systems this way. I’ve seen it happen. The visual builder and the built-in nodes are designed for exactly this.
Where custom code helps is optimization. If you want to implement a specific chunking strategy or custom validation logic, you add a code node. But that’s enhancement, not necessity.