Hi everyone,
I wanted to share an idea that came to me this morning after spending a few days working with PydanticAI. I’m still new to it, so some parts can be a bit confusing.
Before diving into PydanticAI, I mainly used N8N to build AI Agent workflows. It’s fast, visual, and great for prototyping. But I realized that if I want to move towards more professional and scalable backend systems, I can’t rely on N8N forever. That’s what led me to explore PydanticAI.
Here’s the idea:
What if we prototype AI Agents in N8N, then automatically convert those workflows into Python code using PydanticAI via Archon?
We could create an Agent that reads N8N workflows, understands the nodes, logic, and tools, and outputs ready-to-use Python code structured for PydanticAI. This would save hours of manual work and make scaling up much easier.
To make it even better, we can set up a centralized template that guides the AI on:
• Where to define tools
• How to configure memory
• How to organize nodes
• And where to inject the logic and structure
This would give the AI a consistent structure to follow when converting, making the process smoother and the code cleaner.
Just an idea I wanted to throw out to the community—what do you all think?
Best,
Chien Nguyen
1 Like
I love where your head is at with this! Eventually I would love to make Archon capable of converting n8n workflows to Pydantic AI code, you could think of it as just another source of documentation for what you are trying to build with Pydantic AI!
Biggest challenge here is training the LLM to understand all the different nodes in n8n. Could even be worth fine tuning an LLM on n8n workflows.
2 Likes
Hi Cole,
We actually dont even need to port the workflow from n8n to Archon.
I have a better idea to simplify it.
Create a drawing tool, like what you use in langchain studio.
A work flow chart convert to a AI Agent template using PydanticAI. where it centralize Prompts, Controllers, Actions, Tools, etc…
Then the end user just go there and modify the tool they wanna use, set up the database connection, and so on.
Example:
N8N makes drag and drop AI Agent but they use langchain.
Then why dont we make something same drag and drop but Agent core is pydanticAI instead of langchain.
then once design is done, it will export the Python code, to let dev customize it. in that case, Archon will be the best Ai Agent creator ever seen in history.
Chien.
2 Likes
Hmmmm yeah I can get behind what you’re laying out here, I love this! Sometimes I forget that the AI agent nodes in n8n is just using Langchain under the hood. A visual builder to do the same for Pydantic AI would be incredible! Lot of UI/UX would have to go into that haha