I’ve spent a lot of time researching AI and building projects with it over the last years. I started out coding projects by hand (partly with help from AI), just to prototype and get a feel for the ins & outs of working with AI. As an architecture purist I fell in love with LangChain early on, even though it was barely usable and manageable. I ran into many problems and tried many workarounds, and now feel confident enough to say: I know what to build with AI and how. But something changed me along the way that I never dreamt would happen.
You see, I have always coded my solutions with the help of code editors and all the powerful plugins that supported me in this, and this served me well until AI came onto the scene. Because, as an individual developer I now coded too brittle solutions too slow, even when supported by AI as a dev tool. Surely this is the time to iterate fast and so I researched a promising no-code platform that would suit me for a long time to come. As an old school dev with 30+ years under my belt I did not foresee the value n8n would bring me. Being able to very rapidly assemble automation workflows with AI agents in the middle that have been given a very clear and limited scope of operation, I feel like I am on a different path now.
What are those advantages you ask?
Besides the speed of operation I realized the following:
Devs can build and share their solutions so easily, that rich solutions are just one copy paste away, waiting to be adapted and integrated into your own projects.
Sure, we have open source git repos to copy code from, but that code is distributed over multiple files, making understanding and integration a much more complex (and brittle) endeavor.
Workflows are visual: one can easily see how a workflow should work and what to operate on during the entire lifecycle of the solution, being a hardcore dev or any intelligent newcomer.
Now this is a joyful evolution in my day to day work, as I happily and effortlessly add new tools to my AI agents in no time! I still code stuff like frontends, but having n8n as a backend makes me iterate so much faster that I recommend it to anyone 😉
Of course there are some limitations to using n8n, like still not being able to get streaming responses out of it, or hitting their oversimplified RAG solution (it does not even output metadata ffs), but I hope that will be fixed in the near future. Let's just help the n8n team to grow so feature requests are implemented quicker!
Oh, and while we're at it, lets mention my list of outstanding requests here that I think have a lot of value:
- AOP style hooks to be able to call http endpoints before and after a tool gets called, or something similar like an n8n event node so that we can build one listener that translates n8n events into whatever action we come up with ourselves. Use case: notifications.
- Failover LLM: Retry with another LLM provider on failure
- Conditional AI Agent tools: ability to bypass tools based on conditions, so that we can disable tools with logic, minimizing the noise for the LLM to act on.
The post My AI dev journey towards no code with n8n appeared first on iD!OTZ.
Top comments (0)