2 ms·

Node-RED is cool but I didn't realise it could do the AI workflows that n8n is aiming for?

rubenfiszel 1y ago
For the AI workflows, we (windmill.dev) added AI Agent steps as first class primitives very recently.
nemanja030 1y ago
Node-Red is probably a more hands on approach, than n8n, as someone who have 6-7 years of experience in node-red i find n8n really complicated for some reason. A lot of windows and configurations. Started with node-red and quickly built great skills in backend development. Also utilized Cursor to quickly spinup a OpenRouter nodes for node-red, install node-red and check them out :D https://www.npmjs.com/package/@lizzard-solutions/node-red-contrib-openrouter https://www.npmjs.com/package/@lizzard-solutions/node-red-co...
- kristopherleads 1y ago
  Love to hear you love using Node-RED! That package looks pretty cool! I'll check it out - I'm super in the weeds today with a flow so I'm in deep-dive mode.
rcarmo 1y ago
It can, absolutely. I have been prototyping Agentic workflows on it for a long while now with my own custom nodes. Also, it runs great on bun now.
kristopherleads 1y ago
It definitely can! Full disclosure, I'm a DevRel at FlowFuse, so I'm definitely biased here - but there's quite a lot of support for AI workflows, from lightweight automation to fairly complex pipelines. For example, I'm currently building a flow that connects device monitoring data to a central reporting structure, compares the incoming values against internal spec docs, and uses OpenAI to summarise overall operational status and any drift/out-of-spec issues. The summaries (along with the raw JSON objects) are then sent over MQTT for multi-site compliance and operations. Once you map it out, it's surprisingly straightforward to build. What makes Node-RED and FlowFuse stand out on this particular use case IMO is that AI flows get treated just like any other message payload. That means your AI output becomes a first-class data object in the system - you can remix, transform, mutate, or extract from it the same way you would any other JSON payload, making it pretty portable and easy to integrate.

For the AI workflows, we (windmill.dev) added AI Agent steps as first class primitives very recently.

Node-Red is probably a more hands on approach, than n8n, as someone who have 6-7 years of experience in node-red i find n8n really complicated for some reason. A lot of windows and configurations. Started with node-red and quickly built great skills in backend development. Also utilized Cursor to quickly spinup a OpenRouter nodes for node-red, install node-red and check them out :D https://www.npmjs.com/package/@lizzard-solutions/node-red-contrib-openrouter https://www.npmjs.com/package/@lizzard-solutions/node-red-co...

kristopherleads 1y ago
Love to hear you love using Node-RED! That package looks pretty cool! I'll check it out - I'm super in the weeds today with a flow so I'm in deep-dive mode.

rcarmo 1y ago

It can, absolutely. I have been prototyping Agentic workflows on it for a long while now with my own custom nodes. Also, it runs great on bun now.

kristopherleads 1y ago

It definitely can! Full disclosure, I'm a DevRel at FlowFuse, so I'm definitely biased here - but there's quite a lot of support for AI workflows, from lightweight automation to fairly complex pipelines. For example, I'm currently building a flow that connects device monitoring data to a central reporting structure, compares the incoming values against internal spec docs, and uses OpenAI to summarise overall operational status and any drift/out-of-spec issues. The summaries (along with the raw JSON objects) are then sent over MQTT for multi-site compliance and operations. Once you map it out, it's surprisingly straightforward to build. What makes Node-RED and FlowFuse stand out on this particular use case IMO is that AI flows get treated just like any other message payload. That means your AI output becomes a first-class data object in the system - you can remix, transform, mutate, or extract from it the same way you would any other JSON payload, making it pretty portable and easy to integrate.