What is before and after converse API?

 Before vs. After the Converse API

​To truly understand their relationship, it helps to see what boto3 looked like before AWS introduced the Converse API.

​The Old Way: invoke_model()

​Originally, boto3 only gave you a raw method called invoke_model(). Every AI provider (Anthropic, Meta, Mistral) required a completely different data structure. If you used boto3 to talk to Claude, your code looked entirely different than if you used boto3 to talk to Llama. You had to manually format strings into JSON and decode bytes.

​The Modern Way: converse()

​AWS realized developers were writing too much "glue code" to switch between models. They built the Converse API directly into the boto3 library to act as a universal translator.

​Now, boto3 translates a single, unified Python dictionary format into whatever specific format the underlying model requires behind the scenes.