FindItTogether

Find It Together is a Fire TV memory-assistance prototype that connects Bee conversation transcripts and Ring Playground footage using Amazon Bedrock. It helps users review their day and ask follow-up questions about everyday events, such as an expected package.

Inspiration

An expected delivery can be mentioned in a conversation and later appear on a camera, but those clues live in different places. We wanted a simpler way to connect what was said with what was seen, without making someone search separate histories.

Find It Together grew from the idea of helping families with everyday recall. We brought that idea into a TV-based experience: review your day, ask a question, and follow up for the details that matter.

What it does

Find It Together offers two paths: a daily summary and voice questions. For the summary, Amazon Bedrock analyzes conversational context from Bee alongside visual material from Ring Playground.

In the demo, asking “Did I forget anything today?” brings up an expected package. The user can then ask where the package is and what it looks like. The app answers that it is at the door, describes its appearance, and includes timestamps in its answers.

The recorded development setup uses Ring Playground footage and the Mac microphone for voice input.

How we built it

Bee CLI supplies the conversational context, Ring Playground supplies the visual stream, and Amazon Bedrock analyzes the transcripts and visual material for summaries and question answering. The Fire TV application brings those results into one interface.

The package scenario shows why this combination matters: spoken context identifies an expected event, while visual context helps answer questions about the item. The demonstration does not establish a deployed real-home camera integration or Fire TV remote-microphone support.

Challenges we ran into

The central integration challenge is connecting conversational context with a visual observation while keeping timing meaningful. Knowing that a package was expected is different from knowing where it was observed, and an earlier observation does not prove where it is now.

The latest iteration adds timestamps and package descriptions to make the answers more concrete. Distinguishing historical observations from current state remains an important next step.

Accomplishments that we’re proud of

The prototype combines conversational and visual context in one experience instead of presenting two disconnected histories. It supports both a broad daily recap and a specific sequence of follow-up questions about a package.

Adding timestamps and package descriptions makes the demonstration more specific than a generic summary: the user can ask about the day, then follow up about a particular item.

What we learned

Connecting sources is only part of making a useful assistant. An answer also needs enough detail and timing for the person to understand what it refers to.

The package scenario gives us a concrete way to evaluate that idea: can the user move from a general question about the day to a useful description of a particular item? Clearer source evidence and uncertainty handling are the next steps, not simply more fluent answers.

What’s next for Find It Together

Our next priorities are to make supporting source moments easier to inspect, distinguish historical observations from current state, and handle missing or conflicting evidence explicitly.

We also want to expand testing on supported Fire TV environments and consented household scenarios, and develop the broader family-assistance experience with clear access, consent, and sharing controls. These are future directions. The current prototype is not a clinically validated memory-care product.

Built With

• amazon
• amazon-fire-tv
• amazon-web-services
• bedrock
• bee-cli
• ring-playground