Hello I’ve been really stressed about this for a few days now ㅜㅜ
I have configured you to create an agent that rags using vector db in the supabase.
It was also confirmed that the data was well chunked and the vectorized value entered the table well.
The prompt also stated “reflect the results of the tool.” The problem is that you call the tool properly and get the value well, but you don’t write the value as the agent’s output.
If you look at the example of other YouTubers, it’s the form of openai → supabae → openai.
But for me, openai → openai (the output has been decided here, so there is no information) → superabase vector. This structure makes tool calls useless and lag is not done properly.
I did the same as everyone else does, but what’s wrong with mine? ㅜㅜ I’d appreciate it if you could help me with what the hell it is
Hey @dldpeka !
Welcome to n8n community!
Honestly i do nt understand what are you trying to say, neigher i i dont understand the text inside the image, but i can understand the connections between nodes…
I would follow this example in order for the agent to return the results form the tool (and to be up to date with informations, there s atrigger to check for new docs in Drive and upload then in the Vector store.).
I use other services, but the “concept” of store/retrieve should be the same.
Here are the nodes:
P.S connect the Vector Store Tool to yur Ai Agent.
I dont use Supabase, i use as you can see Pinecone.
Cheers and Happy New Year!