The problem I currently have is that when it generates a query it is not aware of what Q is so it won't know if it's a product, product_type, merchant, client.

Because of this confusion (that even a human being will have unless they are aware of what valued are in what column) it is not able to generate a relevant sql query

Hello guys

I am using arize-phoenix for my RAG observability along with the opentelemetry framework. I am having conflicts in my requirement.

My main project which I integrate the RAG with requires pydantic v1 while the opentelemetry requires pydantic v2.

How can I fix this issue.

Hi guys
How can I make my Text to SQL pipeline more faster

Currently I am using the llama-index Text-to-SQL query pipeline following the guide here https://docs.llamaindex.ai/en/stable/examples/pipeline/query_pipeline_sql/

I retrieve only the SQL query then execute it myself. Then the result is pass to an llm that synthesis a response.

This takes about 7 seconds to run and return the result to a front end.

The bulk of the time were consumed on the text-to-sql pipeline and the response synthesizer.

Is there a tip on how to make the system more faster