ReviewSense

Daily workflow
Brightdata scrappers
LlamaIndex responses
Workflow output

Demo Video

https://www.loom.com/share/612c562db8264f9fac796d1a3c71322f?sid=22820034-13da-49e8-ad1d-df1164b66152

Inspiration

App reviews are a goldmine of user feedback, but most of the time they get buried under thousands of repetitive, unstructured comments. Developers struggle to keep up, users feel unheard, and opportunities to improve apps slip through the cracks. We wanted to build something that gives reviews the attention they deserve — automatically, intelligently, and continuously.

What it does

ReviewSense is an AI-powered workflow that runs daily to:

Collect fresh app reviews every day
Analyze sentiment and detect common themes
Prioritize issues and feature requests
Generate actionable responses developers can send back
Suggest improvements that feed directly into product workflows

In short, it helps apps listen, learn, and respond to their users at scale.

How we built it

We set up an agentic AI workflow that:

Extracts reviews from app stores using Data
Runs daily workflow using Airia that invokes a LLamaIndex agent to analyze the reviews for sentiment, categorization, and clustering.
Maps findings to recommended actions.
Generates natural, context-aware replies developers can use right away.
Honeyhive is used to get a rating on clarity, correctness and relevancy.

Challenges we ran into

Setting up a reliable daily pipeline that works without manual intervention.

Accomplishments that we're proud of

Building a working end-to-end daily agentic workflow within hackathon time.
Going beyond simple sentiment analysis to deliver actionable insights.
Turning scattered reviews into a clear feedback loop for developers.

What we learned

How powerful agentic AI workflows can be when run continuously instead of one-off.
That developers want not just insights, but also ready-to-use actions.
How much impact small changes in prompt design have on output quality.