[BUG] CachePointBlock ttl field dropped by SDK + cache_tools ordering violation with 1h TTL

[namartaknev]

Checks

• I have updated to the lastest minor and patch version of Strands
• I have checked the documentation and this is not expected behavior
• I have searched ./issues and there are no duplicates of my issue

Strands Version

1.35

Python Version

3.12

Operating System

linux

Installation Method

pip

Steps to Reproduce

1. Create a BedrockModel with cache_tools="default" and a system prompt containing a cachePoint with ttl="1h"
2. Create an Agent with ConversationCacheHook that sets CACHE_POINT_BLOCK = {"cachePoint": {"type": "default", "ttl": "1h"}}
3. Invoke the agent with a multi-turn conversation
4. Observe: (a) ValidationException due to TTL ordering, or (b) cache falls back to 5m TTL silently

Expected Behavior

cachePoint blocks with "ttl": "1h" should be forwarded to Bedrock as-is
cache_tools="default" should allow configuring its TTL to match other cache points
Bedrock should honor 1-hour TTL on all configured cache checkpoints

Actual Behavior

Bug 1: _format_request_message_content in bedrock.py silently drops the ttl field from cachePoint blocks — all cache silently falls back to 5m TTL regardless of what is set
Bug 2: cache_tools="default" hardcodes a 5m TTL on the toolConfig cache point. Since toolConfig is processed before system and messages, a 1h TTL on a subsequent cache point violates Bedrock's decreasing TTL ordering rule → ValidationException

Additional Context

The Bedrock CachePointBlock API supports an optional ttl field with values "5m" or "1h": https://docs.aws.amazon.com/bedrock/latest/APIReference/API_runtime_CachePointBlock.html
Bedrock processes cache checkpoints in order: toolConfig → system → messages, and requires TTLs to be in non-increasing order
Users leveraging ConversationCacheHook (manually injected cache points via hooks) cannot access 1h TTL because the SDK strips the field before the API call

Possible Solution

# Bug 1 fix — bedrock.py _format_request_message_content
if "cachePoint" in content:
    cp = {"type": content["cachePoint"]["type"]}
    if "ttl" in content["cachePoint"]:
        cp["ttl"] = content["cachePoint"]["ttl"]
    return {"cachePoint": cp}

# Bug 2 fix — add cache_tools_ttl param to BedrockModel
# e.g. BedrockModel(cache_tools="default", cache_tools_ttl="1h")
# so toolConfig cache point TTL can be aligned with system/conversation cache points

Related Issues

No response