Feature request
Type
Describe the solution you'd like
Since the LLM might call the API and then do post-processing in SQL based on the user needs, that process of generating the SQL takes time and might require a few retries for complex API schemas. However those SQL transformations can then be saved as a cached tool with some LRU such that simialr requests in the future can be faster and more dependable.
in similar way also learnings about how to use the API can be cached in some limited memory per api so the agent can "learn and improve" how to work with the API beyond what the schema provides
Feature request
Type
Describe the solution you'd like
Since the LLM might call the API and then do post-processing in SQL based on the user needs, that process of generating the SQL takes time and might require a few retries for complex API schemas. However those SQL transformations can then be saved as a cached tool with some LRU such that simialr requests in the future can be faster and more dependable.
in similar way also learnings about how to use the API can be cached in some limited memory per api so the agent can "learn and improve" how to work with the API beyond what the schema provides