feat(models): add configurable TTL for Bedrock cache points

[mattdai01]

Support 1-hour and 5-minute TTL options for Bedrock prompt caching. This allows users to extend cache lifetime beyond the default 5 minutes for supported models (Claude Opus 4.5, Haiku 4.5, Sonnet 4.5).

• Add ttl field to CacheConfig for messages cache point
• Add cache_tools_ttl field to BedrockConfig for tools cache point
• Add optional ttl field to CachePoint TypedDict
• Pass through TTL on user-supplied cachePoint blocks

Description

https://aws.amazon.com/about-aws/whats-new/2026/01/amazon-bedrock-one-hour-duration-prompt-caching/
Support 1-hour and 5-minute TTL options for Bedrock prompt caching. This allows users to extend cache lifetime beyond the default 5 minutes for supported models (Claude Opus 4.5, Haiku 4.5, Sonnet 4.5).

Previously, all cache points used Bedrock's default 5-minute TTL. With this change, users can configure extended TTL through:

• CacheConfig(strategy="auto", ttl="1h") for messages cache points
• cache_tools_ttl="1h" for tools cache points
• Manual {"cachePoint": {"type": "default", "ttl": "1h"}} in SystemContentBlock (previously ttl was silently stripped)

Changes

• Add ttl field to CacheConfig for auto-injected messages cache point
• Add cache_tools_ttl field to BedrockConfig for tools cache point
• Add optional ttl field to CachePoint TypedDict for type safety
• Fix _format_request_message_content to pass through ttl on user-supplied cachePoint blocks (was previously dropped)
• Refactor _inject_cache_point to accept ttl parameter from caller instead of re-reading config

• Related Issues

Documentation PR

Type of Change

New feature

Testing

Unit tests passed, did a quick test checking cacheReadInputTokens in bedrock response was > 0 even after 5 min ttl when 1 hr was enabled.

• [ X ] I ran hatch run prepare

Checklist

• [ X ] I have read the CONTRIBUTING document
• [ X ] I have added any necessary tests that prove my fix is effective or my feature works
• I have updated the documentation accordingly
• I have added an appropriate example to the documentation to outline the feature, or no new docs are needed
• [ X ] My changes generate no new warnings
• [ X ] Any dependent changes have been merged and published

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[JackYPCOnline]

Is this PR related to #1219

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[mattdai01]

I think we no longer have pass through on the cachePoint right? If you look at the _format_request_message_content function in the PR, it seems that it only passes through type currently, and drops ttl. now the behavior is that it drops everything except type and ttl.