Got this working locally with openai/gpt-oss-20b within LM Studio + Docker #165
Replies: 1 comment
-
|
I tried, but even an empty "Test-Workflow" seems to be too hard. I think i have to config the output format? "This is too messy. I think we cannot proceed with API due to unknown connection format. In practice, the easiest way is to create workflow via n8n UI or use a template that already has manual trigger. Given constraints, I'll explain that a minimal workflow requires at least two nodes and proper connections; the exact JSON format for connections is not trivial here. I’ll provide guidance rather than actual creation." "I’m afraid the API expects a very specific connection‑structure that’s hard to guess from scratch. In practice, the easiest way to get a “blank” workflow in n8n is: Create a new workflow named “Test MCP” via the UI or the n8n_create_workflow call. If you still need me to push the exact JSON via the API, let me know and I’ll try a different approach (e.g., using an existing template)." |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
Hey - just a quick thank you to the devs, very useful.
I got this working all locally, using openai/gpt-oss-20b as the model. I'm a n00b, so if I can do it anyone can.
Setup:
I recorded a somewhat rambling video that shows what to do, might help you though: https://youtu.be/chcPsQV5qj0
Top tip: bump up the context window for the chat to ensure it can cope with the system prompt and the tool use.
Bonus tip: I've tried this with other tool use models, works great too!
Thanks again to the devs.
Beta Was this translation helpful? Give feedback.
All reactions