OpenAI MCP (SQL) or LightRAG / GraphRAG for Product Availability Locator Chatbot

API
, , ,
1

Hi everyone!

I am building a chatbot that helps users find available products across stores of a retail chain.

The data I have is structured:

One XML file with a list of stores, their IDs, and location.
One Excel file containing available product listings (store id, description, features, specs, price, etc) (updated frequently).

I have to choose between three options:

  1. GraphRAG (basically create a graph to model relationships between stores, products, products descriptions specs and features.)

  2. LightRAG (a lighter-weight graph-based RAG approach. lower cost and lower complexity)

  3. OpenAI MCP with SQL (PostgresSQL)
    Use OpenAI’s MCP to allow the model to interact directly with the database (that i will create). This solution is easier because: the data is structured, it will cost me peanuts compared to the other solution, fairly simpler than the other solutions, the model can directly infer the user requirement and generate SQL queries. Easy Geolocation filter with PostGIS (“search within 30km”).

I understand that GraphRAG maybe overkill and expensive, so really the choice comes down to either LightRAG or OpenAI MCP with SQL Server.

I would love to hear your thoughts and suggestions.

Thanks.