[Feature Request] Check and support Prompt Caching for All Models

[Pakchoioioi]

Motivation

Support prompt caching for all models in the product. Prioritize cloud model providers first, and subsequently continue to add support for BYOK (Bring Your Own Key) models.

• Gemini 3.1pro/3flash/3pro
• GPT-4.1mini/4.1/5/5.1/5.2/5mini
• Claude Sonnet 4-5
• MiniMax M2.5

BYOK

• Qwen
• Deepseek
• Z.al
• Moonshot
• Grok
• Mistral
• ...

Please follow-up with CAMEL issue first： camel-ai/camel#3770

Solution

No response

Alternatives

No response

Additional context

No response

[Pakchoioioi]

For internal use only, not open to contributions.

[Zephyroam]

Blocked by pending reviews for camel-ai/camel#3784

[Zephyroam]

MiniMax M2.5 supports automatic caching, nothing need to do on our side.

Gemini supports Implicit caching and Explicit caching. Implicit caching needs nothing to do on our side. Explicit caching requires manually caching contents, which is not flexible.

[Zephyroam]

For BYOK models:

DeepSeek, Grok, Moonshot and Z.ai supports automatic caching.

Mistral does not support prompt caching.

Qwen supports Explicit cache and Implicit cache, but they both have pros and cons. Please check the doc before we can decide.