feat(bedrock): add docs for SystemContentBlock caching approach

[dbschmigelski]

Description

This is the followup to feat(models): add SystemContentBlock support for provider-agnostic caching strands-agents/sdk-python#1112. It includes documentation to describe the new approach.

After strands-agents/sdk-python#1141 is merged, we will update the LiteLLM section. Then after that as we begin progressing on [FEATURE] Prompt caching support for all models strands-agents/sdk-python#1140 we will update docs to provide a general caching section

Type of Change

• Content update/revision

Motivation and Context

strands-agents/sdk-python#1112

Areas Affected

http://127.0.0.1:8000/user-guide/concepts/model-providers/amazon-bedrock/#caching

Screenshots

Checklist

• I have read the CONTRIBUTING document
• My changes follow the project's documentation style
• I have tested the documentation locally using mkdocs serve
• Links in the documentation are valid and working
• Images/diagrams are properly sized and formatted
• All new and existing tests pass

Additional Notes

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

[rickkk856]

@dbschmigelski Is there any specific order? Or support to few dynamic SystemContentBlock?

Eg:

# Define system content with cache points
system_content = [
    SystemContentBlock(
        text="You are a helpful assistant that provides concise answers. "
             "This is a long system prompt with detailed instructions..."
             "..." * 1600  # needs to be at least 1,024 tokens
    ),
    SystemContentBlock(cachePoint={"type": "default"}),
    SystemContentBlock(text=f"Some other aditional text, eg; with dynamic {date}")
]

# Create an agent with SystemContentBlock array
agent = Agent(system_prompt=system_content)

or eg:

# Define system content with cache points
system_content = [
    SystemContentBlock(text=f"Some other aditional text, eg; with dynamic {date}"),
    SystemContentBlock(
        text="You are a helpful assistant that provides concise answers. "
             "This is a long system prompt with detailed instructions..."
             "..." * 1600  # needs to be at least 1,024 tokens
    ),
    SystemContentBlock(cachePoint={"type": "default"})
]

# Create an agent with SystemContentBlock array
agent = Agent(system_prompt=system_content)

Or could it even be built like this?

# Define system content with cache points
dynami_system_content = SystemContentBlock(text=f"Some other aditional text, eg; with dynamic {date}")
    
system_content = [
        SystemContentBlock(
        text="You are a helpful assistant that provides concise answers. "
             "This is a long system prompt with detailed instructions..."
             "..." * 1600  # needs to be at least 1,024 tokens
    ),
    SystemContentBlock(cachePoint={"type": "default"})
]

# Create an agent with SystemContentBlock array
agent = Agent(system_prompt=[dynami_system_content , *system_content])

I'm getting strange errors like this:
botocore.exceptions.ParamValidationError: Parameter validation failed:
Invalid type for parameter system[0].text, value:

Any best recommendation? Tested on versions 1.17.0 and 1.15.0

[dbschmigelski]

Hi,

from strands import Agent
from strands.types.content import SystemContentBlock
import datetime

# Define system content with cache points
system_content = [
    SystemContentBlock(
        text="You are a helpful assistant that provides concise answers. "
             "This is a long system prompt with detailed instructions..."
             "..." * 1600  # needs to be at least 1,024 tokens
    ),
    SystemContentBlock(cachePoint={"type": "default"}),
    SystemContentBlock(text=f"The current time is {datetime.datetime.now()}")
]

# Create an agent with SystemContentBlock array
agent = Agent(system_prompt=system_content)
agent("What time is it")

and

from strands import Agent
from strands.types.content import SystemContentBlock
import datetime

dynami_system_content = SystemContentBlock(text=f"the time is {datetime.datetime.now()}")

system_content = [
        SystemContentBlock(
        text="You are a helpful assistant that provides concise answers. "
             "This is a long system prompt with detailed instructions..."
             "..." * 1600  # needs to be at least 1,024 tokens
    ),
    SystemContentBlock(cachePoint={"type": "default"})
]

# Create an agent with SystemContentBlock array
agent = Agent(system_prompt=[dynami_system_content , *system_content])


agent("What time is it")

appear valid to me.

Can you take a look at the message array? Maybe there is an issue like an invalid character relating to date

[rickkk856]

Nevermind, problem was in a corrupted environment.

By the way tested again and order didn't influence, although it might be worth documenting that dynamic content-blocks are supported. Also the default TTL of 5 minutes of the system prompt cache.