feat: add session token support for AWS Nova Sonic

[TamaryOmer]

Summary

Adds support for AWS session tokens in the Nova Sonic speech-to-speech service integration, enabling authentication with temporary AWS credentials.

Changes

• Removed explicit credential parameters: No longer requires secret_access_key, access_key_id, or session_token in constructor
• Added environment credential resolver: Uses EnvironmentCredentialsResolver to automatically detect credentials from: Environment variables (AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_SESSION_TOKEN), IAM roles , AWS credential files or other standard AWS credential sources • Updated dependencies: Added smithy-aws-core for credential resolution • Simplified configuration: Removed manual credential management code

Benefits

• Automatic session token support: Works seamlessly with temporary credentials • Follows AWS best practices: Uses standard AWS credential chain • Reduced complexity: No need to manually pass credentials • Better security: Supports IAM roles and other secure credential methods

Breaking Changes

• Constructor no longer accepts secret_access_key, access_key_id, or session_token parameters • Credentials must now be provided via environment variables or other AWS standard methods

Migration

Before: python

llm = AWSNovaSonicLLMService(
    secret_access_key="...",
    access_key_id="...",
    session_token="...",  # optional
    region="us-east-1"
)

After:

python
# Set environment variables or use IAM roles
llm = AWSNovaSonicLLMService(
    region="us-east-1"  # region now defaults to us-east-1
)

[markbackman]

Thanks for the submission. This unfortunately makes a breaking change that will impact developers using AWSNovaSonicLLMService. We're happy to support this pattern, but it needs to work with the current authentication scheme as well and can't be a breaking change. If that's something you're interested in submitting, we can review it.