Standardize & Abstract Prompt/Context Caching #848

[rvndbalaji]

Checked

• I searched existing ideas and did not find a similar one
• I added a very descriptive title
• I've clearly described the feature request and motivation for it

Feature request

Hi,

I noticed that Context Caching is supported by both Gemini and Claude but both of the syntax is different
Is it possible to abstract away this and allow a general purpose cache_options that caters to different LLM providers?

LangChain is supposed to be the abstraction that performs all of this, and I was hoping this can be implemented

Motivation

I need to implement caching mechanism for both Gemini and Anthropic, but I want this to be vendor neutral

FYI - @lkuligin @baskaryan @efriis

Proposal (If applicable)

No response