Ingoring Workflow - Vector Store Tool

Questions
1

Firstly, I’m completely new to n8n so still coming to terms with the logic. I’m trying to get the workflow to call the vector store first before it goes to the LLM. At the moment its ignoring the vector store entirely. Is it possible to force the Agent to work with the Vector Store first before engaging the LLM. If so how?

I have accomplished this in other solutions but I really like n8n’s setup and would like to keep it if possible. I’m using the latest version of n8n on the cloud.

What is the error message (if any)?

None.

Please share your workflow

Share the output returned by the last node

Information on your n8n setup

  • n8n version:
  • Latest Online Version
  • Database (default: SQLite):
  • Pinecone or Supabase
  • n8n EXECUTIONS_PROCESS setting (default: own, main):
  • Running n8n via (Docker, npm, n8n cloud, desktop app):
    n8n Cloud
  • Operating system:
    Online
1 Like
2

Hi, I dont think you can force it. But maybe by adding the instruction to prompt / system prompt.

1 Like
3

Yes, so you should instruct it by a system message, i’ve managed to get it to always use, but you have to lieterally say like ‘always use tools and vector store etc’ generally not needed in user message, but prompting there can help too.

Also defining what the ai agent does, help to further itterate to use the tools, like you are an ai agent which does this, before doing this you must use the tool answer questions with a vector store etc

It’s quite fun actually ahaa Hope this helps bit more too

2 Likes
4

Thanks, do you have an example I could try to follow or know a video I could watch to learn from?

5

this seems to work everytime @gofly.online

6

I mean you don’t have to complicate it, try different variations. Like “In the beginning of every session you must invoke the tool…” etc.

Another thing is the model you are using. The most reliable models are usally the bigger ones.

1 Like
7

Thank you.

It utterly odd. I’ll try different models as well. Its just randomly getting items and messing them up.

But the main thing is that I set it up correctly. I really haven’t changed anything different from the example. Hence why I was perplexed.

Thank you again!

Cheers!

1 Like
8

I wish you luck on the workflow. And just as a sidenote from personal experience putting less on the LLMs is usallly for the best. Sometimes they are unpredictable. But with a bigger model and lower temperature setting should work in your case.

Regards

2 Likes
9

Agreed, I’m going to have to go back to the drawing board to which model I use.

Thank you again.

1 Like
10

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.