How to implement Claude Prompt Caching for intent classification in Neuron AI

[ololoDevs]

I'm building a chatbot using Neuron AI framework and Claude API, and I have a use case that would greatly benefit from prompt caching.
My scenario:
At one of my checkpoints, I'm doing intent classification where I constantly send the same static content in every request:

A comprehensive list of user intents (~20-30 intents)
Descriptions for each intent
Multiple example queries for each intent

This static intent list is sent with every user message to classify their intent, which is inefficient in terms of both cost and latency.
What I need:
I'd like to leverage Claude's Prompt Caching feature to cache this static intent list. According to Anthropic's documentation, this requires:

Adding the anthropic-beta: prompt-caching-2024-07-31 header
Structuring the prompt with cache_control parameter on the content blocks that should be cached

My questions:

Does Neuron AI framework support Claude's Prompt Caching feature?
If yes, how do I:

Add the required beta header to requests?
Structure my prompts with cache_control blocks in the system/messages array?

Are there any examples or documentation for implementing prompt caching with Neuron AI?
If it's not currently supported, are there plans to add this feature?

Any guidance or code examples would be greatly appreciated!
Thanks in advance!

[ilvalerione]

It's not supported out of box. Provider specific features could be implemented hooking the current framework design. If it can't be done it's an opportunity to propose a design change to make the framework more flexible overtime.