Creating a basic AI agent for polling inventory. #478
danielrosehill
started this conversation in
Show and tell
Replies: 1 comment 1 reply
-
Isn't this WAY more complicated and slower than a standard search? |
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Cleaning Up & Exporting A Static Inventory
So my ultimate objective is to get a proper setup that uses the Homebox API in order to be able to look up items dynamically.
But for a simple proof of concept I decided to do it the dumb way and export my inventory statically.
Step one, export your inventory from the Homebox.
This is totally optional, but I thought it would make sense to do a little bit of cleaning of the data so that irrelevant data for retrieving inventory information could be excluded:
Basically, I got rid of the columns that I thought didn't need to be there.
One thing you can also try is rewriting the headers so that they're a bit more obvious for an AI agent to interpret.
Something like this:
For the purpose of scalability testing, here's where I'm at currently in my inventory:
Once I thought that I'd done enough cleaning, I downloaded a CSV.
Creating Agent In LibreChat
The next step is to create an agent configuration in LibreChat or your preferred large-language model frontend.
The agent is basically a model with a system prompt attached and related to the custom knowledge.
These instructions could almost certainly be improved upon but they should get the job done:
Once you upload your knowledge, the CSV is being vectorized, embedded into a RAG pipeline using the built-in vector database that comes with the stack:
When you can see the file name display, the process has been successful and your agent can now access the inventory:
Make sure to click on save to make sure that the knowledge has been added:
Now our chatbot is ready for use and we can ask you questions about our inventory:
You can see that it's checking the files when asked a question:
It was able to identify all the infrared controllers in my tech cabinet and isolate the one that I was looking for, number 3!
You can ask follow-up questions to retrieve details about the thing you're looking for:
And the chatbot will even respond in the format you requested with the data you wanted and provide a link to the catalog entry:
Beta Was this translation helpful? Give feedback.
All reactions