[Feature Request] Extend Prompt Caching Support to More Providers

[Wendong-Fan]

Required prerequisites

• I have searched the Issue Tracker and Discussions that this hasn't already been reported. (+1 or comment there if it has.)
• Consider asking first in a Discussion.

Motivation

Following #3587 which added prompt caching for Anthropic and OpenAI, we should extend support to:

• Google Gemini - Context caching with implicit/explicit modes
• AWS Bedrock - Prompt caching for
  Claude and Nova models
• Azure OpenAI - Prompt caching
  for GPT-4o+

Related: #3586, #3587

Solution

No response

Alternatives

No response

Additional context

No response

[Zephyroam]

Hi I did some inestigation:

Google Gemini

• Implicit caching works automatically, no changes needed.
• Explicit caching requires manual management. One option is to extend GeminiModel with new methods to allow users to explicitly operate on the cache.

Azure OpenAI

• Uses the same prompt caching mechanism as OpenAI.
• Already supported: AzureOpenAIModel relies on ChatGPTConfig, which includes prompt_cache_key.
• No changes are required.

AWS Bedrock

• Prompt caching is only supported via the native Bedrock API.
• The OpenAI-compatible endpoint currently in use does not support prompt caching.

[waleedalzarooni]

Hi I did some inestigation:

Google Gemini

• Implicit caching works automatically, no changes needed.
• Explicit caching requires manual management. One option is to extend GeminiModel with new methods to allow users to explicitly operate on the cache.

Azure OpenAI

• Uses the same prompt caching mechanism as OpenAI.
• Already supported: AzureOpenAIModel relies on ChatGPTConfig, which includes prompt_cache_key.
• No changes are required.

AWS Bedrock

• Prompt caching is only supported via the native Bedrock API.
• The OpenAI-compatible endpoint currently in use does not support prompt caching.

Great work @Zephyroam, let me know if you need any additional guidance before starting the PR!