[BUG] Nova reel fails without s3Location

[cyberwombat]

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

1.9.1

Python Version

3.13

Operating System

macos 26

Installation Method

other

Steps to Reproduce

Here is the code I am running

from strands import Agent
from strands_tools import nova_reels
from strands.models import BedrockModel


# Create a BedrockModel
bedrock_model = BedrockModel(
    region_name="us-east-1",
)

agent = Agent(tools=[nova_reels], model=bedrock_model)

agent.tool.nova_reels(
    action="create",
    text="Create a video of someone walking through the streets while wearing the sneakers in the photo, showcasing the shoes in action.",
    image_path="./playground/media/sneakers.jpg",
    s3_bucket="foo",
    seed=42,
    fps=24,
    dimension="1280x720",
   
)

# Ask the agent a question that uses the available tools
message = """
I have the following requests:  
1. Create a video of someone walking through the streets while wearing the sneakers in the photo, showcasing the shoes in action.The video should be geared towards marketing purposes, highlighting the design and comfort of the sneakers in a real-world setting targeted at men aged 18-35 who are interested in casual and trendy footwear.
2. Use the provided image of the black Converse sneakers as a reference for the shoes in the video.
3. Save the generated video to the specified S3 bucket for easy access and sharing.
4. If you encounter S3 errors provide the full error message from AWS

"""

agent(message)

Expected Behavior

Should allow upload to s3...

Actual Behavior

I am trying to run the nova_reels examples and it fails with

Error: An error occurred (ValidationException) when calling the StartAsyncInvoke operation: #: required key [s3Location] not found

I am seeing things like PR #310 that indicate this could be a future behavior but right now this doesnt run. What is odd though is that when I failed passing the correct image path the agent said it would just figure it out with text and proceeded correctly generating the video in the bucket and returning the ARN. As soon as I provide a real path it fails.

Related as well: strands-agents/tools#172

Additional Context

Here is the code I am running

from strands import Agent
from strands_tools import nova_reels
from strands.models import BedrockModel


# Create a BedrockModel
bedrock_model = BedrockModel(
    region_name="us-east-1",
)

agent = Agent(tools=[nova_reels], model=bedrock_model)

agent.tool.nova_reels(
    action="create",
    text="Create a video of someone walking through the streets while wearing the sneakers in the photo, showcasing the shoes in action.",
    image_path="./playground/media/sneakers.jpg",
    s3_bucket="foo",
    seed=42,
    fps=24,
    dimension="1280x720",
   
)

# Ask the agent a question that uses the available tools
message = """
I have the following requests:  
1. Create a video of someone walking through the streets while wearing the sneakers in the photo, showcasing the shoes in action.The video should be geared towards marketing purposes, highlighting the design and comfort of the sneakers in a real-world setting targeted at men aged 18-35 who are interested in casual and trendy footwear.
2. Use the provided image of the black Converse sneakers as a reference for the shoes in the video.
3. Save the generated video to the specified S3 bucket for easy access and sharing.
4. If you encounter S3 errors provide the full error message from AWS

"""

agent(message)

Possible Solution

No response

Related Issues

No response

[cyberwombat]

It appears to be creating folders in the bucket. AI response:

_

I'm encountering an AWS validation error. The full error message from AWS is:

Error: An error occurred (ValidationException) when calling the StartAsyncInvoke operation: #: required key [s3Location] not found

This suggests that the S3 bucket configuration requires additional parameters beyond just the bucket name. The Nova Reel service might need a full S3 location specification including the bucket path or object key structure.

Let me check what video generation jobs are currently available to see if there are any examples of successful configurations:
Tool #3: nova_reels

🚀 Nova Reels Tool - Starting Execution
Action requested: list
📡 Connecting to Bedrock Runtime in us-east-1

📋 Listing jobs (max: 5)
I can see from the existing jobs that they use different S3 bucket configurations. The issue appears to be with how the S3 location is being specified. Based on the AWS error message, it seems the service expects a more complete S3 location configuration.

Unfortunately, the current tool implementation seems to have an issue with the S3 bucket parameter formatting. The validation error suggests that the service requires an s3Location parameter structure rather than just a simple bucket name.

Summary of the issue:

• Full AWS Error Message: An error occurred (ValidationException) when calling the StartAsyncInvoke operation: #: required key [s3Location] not found
• Root Cause: The Nova Reel service expects a complete S3 location configuration, not just a bucket name
• Current Status: Unable to proceed with video generation due to S3 configuration parameter mismatch

To resolve this, the tool would need to be updated to properly format the S3 location parameters according to the AWS Bedrock Nova Reel API specifications, or we would need to determine the correct S3 location format expected by the service.%

_

[cyberwombat]

ok I solved it - the error message is very confusing. I had 'jpg' instead of "jpeg" causing the local image to fail and falling back on s3 where it expected a location. I think this needs serious improvement in the error messages!