RAG System for Finding Local Service Providers: Agent Selection Question

Questions
1

Hello, quick question about a RAG system (chatbot + Supabase vector database):

Use case: The chat should display the best service providers based on a given problem and location.
Example: plumbing issue in Paris, the RAG should find plumbers in that city.

Agent choice : Is it better to use the AI AGENT node or the Question and Answer Chain node? In my case, there might be several back-and-forth exchanges for the user to clarify their request. If using AI AGENT, which type of agent would be best: tools, conversational, or OpenAI function (I’m using gpt4-turbo)?

I’m trying to improve the relevance of results and performance (response speed).

Thanks friends.

Information on your n8n setup

  • n8n version: 1.81.4
  • Database (default: SQLite): PSQL
  • n8n EXECUTIONS_PROCESS setting (default: own, main): default
  • Running n8n via (Docker, npm, n8n cloud, desktop app): self-hosted docker
  • Operating system: win10
2

I’d use the tools Agent for sure.

Specially because for that case the best solution would be a database containing the service providers, instead of a vector store.

Structured data is more reliable.
Vector Store is more of a “unstructured knowledge” where pieces of information can end up disconnected from each other and only connected by context.

World you be comfortable using a database? Do you know how to proceed?

3

Thank you for your feedback, Solomon! So you suggest a classic PostgreSQL database on Supabase would indeed be more appropriate than a vector database?

My source data is currently in a Google Sheet, it musbe a way to load in a classic Psql db or I can set the Google sheet as a source/tool?

I guess this approach will allow me to perform precise queries on attributes like location, service type, ratings, etc., without the complexity and uncertainties associated with semantic searches in a vector database.

For the agent, I’ll follow your advice to use the tools agent.

4

You can use Supabase for both: normal database structure and also vector store.

But to store the information in your example, I’d say it’s better to use a normal database structure.

You can also use Google Sheets for that, by using a Google Sheets tool connected to the Ai Tools agent. But Google Sheets won’t be the same as using a database because it’s harder to run complex requests, with many filters.

But it’s worth the shot. Maybe it solves your problem in a simpler way.

And yes, using a table is better for executing precise queries! That’s why I think the table/database structure is better for this case.

:point_right: If my reply answers your question, please remember to mark it as a solution.

1 Like
5

Thanks, Solomon, for taking the time to answer.

1 Like