Allow s3Location as Document, Image, and Video

[Unshure]

Description

Adds native S3 location support for media content types in Bedrock, allowing users to reference images, documents, and videos stored in S3 directly rather than requiring base64 encoding.

One decision I made in this pr is the choice between a generic Location source, vs the more specific S3Location source. Some model providers support url for a document location (ref), so it could make sense to make this generic for s3 as well. But since S3 also has an optional bucketOwner, I thought it would make more sense to have a separate S3Location, and then later add something like a URLLocation to cover the openai case.

What Changed

Type System Updates

• Extended ImageSource, DocumentSource, and VideoSource in types/media.py to accept s3Location as an alternative to bytes
• Added S3Location TypedDict with required uri field and optional bucketOwner for cross-account access

Provider Behavior

• Bedrock: Passes S3 locations through unchanged to the API
• Non-Bedrock providers: These will be addressed in a follow-up pr because this one would be huge otherwise

Updated Integ Tests

I added integ tests which create an s3 bucket (post-fixed with the accountId) where the local file resources are uploaded for testing. Then they are referenced in the tests to ensure the model can recognize and identify the content in them.

Ive also added a short 5 second video resource to the integ tests to cover video modalities

Usage Example

from strands import Agent
from strands.models.bedrock import BedrockModel

agent = Agent(model=BedrockModel())

# Reference an S3-hosted document directly
response = agent(
    [
        {
            "role": "user",
            "content": [
                {"text": "Summarize this document:"},
                {
                    "document": {
                        "format": "pdf",
                        "name": "report.pdf",
                        "source": {
                            "s3Location": {
                                "uri": "s3://my-bucket/documents/report.pdf",
                                "bucketOwner": "123456789012"  # Optional
                            }
                        }
                    }
                }
            ]
        }
    ]
)

Notes on backwards compatibility:
This change updates ImageSource, DocumentSource, and VideoSource to use total=False in their typed dict definition, ultimately making the bytes parameter optional now when it was required before. This is considered safe since all existing code should not be impacted by this. However, going forward, since these fields are essentially union types now, they need to be treated as such by customers going forward.

Related Issues

#1482

Documentation PR

TODO

Type of Change

New feature

Testing

How have you tested the change? Verify that the changes do not break functionality or introduce warnings in consuming repositories: agents-docs, agents-tools, agents-cli

• I ran hatch run prepare

Checklist

• I have read the CONTRIBUTING document
• 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
• My changes generate no new warnings
• 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.

[codecov]

Codecov Report

❌ Patch coverage is 90.74074% with 5 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/strands/models/bedrock.py	87.17%	1 Missing and 4 partials ⚠️

📢 Thoughts on this report? Let us know!