Function Calls and PostgreSQL Functions Training Data in Vanna.AI

[sinjup]

We are encountering issues when adding function calls and PostgreSQL functions as training data in Vanna.AI. The results are not accurate or consistent when functions and their parameters are included as training data. This raises concerns about whether Vanna.AI can effectively identify and handle functions with their parameters in queries.

Function Identification: Can Vanna.AI identify a function from a query and understand its parameters?
Function Execution: Is Vanna.AI capable of executing the identified function with the provided parameters when a user asks a related question?
Training Data: What best practices can be followed to include functions as training data in Vanna.AI to improve the quality of results?
Performance Impact: Will training with functions impact the model's performance, and are there any workarounds to ensure better results?
We would appreciate guidance on how to structure function-based training data effectively, and whether Vanna.AI can identify and execute functions as expected when interacting with users.

Expected Outcome: Better handling of function-based queries and improved accuracy when executing functions in response to user questions.