HackStack

Project Info

bluberri

Devpost

Inspiration

We wanted to create a platform that unifies opinions and reviews from across the internet — from Best Buy and Amazon to Reddit and beyond. It’s hard to get a full picture when feedback is scattered across multiple sites, so our goal with bluberri was to make it simple to explore, analyze, and understand what people really think about any product or topic — all in one place.

What it does

bluberri collects and analyzes social media conversations and product reviews from multiple platforms. It uses AI-powered sentiment analysis to detect emotions and opinions, providing clear visual insights and downloadable CSV reports for any product, feature, or campaign. The frontend allows users to run analyses and view live results, while the backend manages scraping, data aggregation, and model inference.

How we built it

We built bluberri with a Python backend using FastAPI, SQLAlchemy, and transformer models like BERT and RoBERTa, with Google Gemini handling summaries. The backend scrapes data from APIs and web sources, runs sentiment and emotion analysis, and streams progress and results to the frontend — built with Next.js and React. We designed a modular pipeline to integrate multiple data sources, handle rate limits, and provide a real-time analysis experience.

Challenges we ran into

Getting transformer models to run efficiently and reliably Handling rate limits and API restrictions across multiple platforms Building a unified pipeline that merges data from different APIs and formats Keeping real-time updates smooth while streaming results to the frontend

Accomplishments we're proud of

Integrated transformer models (BERT + fast sentiment models) for scalable analysis Created a modular, multi-source data pipeline Built a live, streaming frontend for progress and results Enabled AI-powered summaries and CSV exports for actionable insights

What we learned

How to manage API rate limits and platform-specific restrictions Techniques for optimizing transformer model inference How to design robust error handling and streaming for a better user experience The importance of combining multiple data sources for richer insights

What's next

Integrate more public APIs (e.g., Threads, Bluesky) Add web scrapers for platforms without public APIs Optimize model performance to reduce full-analysis time Expand support for more content types and review formats

Analysis

Compare with all teams

View

Metric

Figures cover GitHub contributors during the hackathon window. A co-authored commit counts in full for each author, so per-member totals add up to more than the whole-team figures.

Technology

Found in codeClaimed only
  • CSSIn code
  • FastAPIIn code
  • Hugging FaceIn code
  • Next.jsIn code
  • PythonIn code
  • PyTorchIn code
  • ReactIn code
  • TypeScriptIn code
  • Google GeminiClaimed

8 of 9 appear in the indexed code. 1 claimed on Devpost could not be matched to code, which may simply mean the tool leaves no trace in the repository.

AI coding agents

  • Claude CodeCommits

Detected from committed agent config files and commit authorship. Absence of a signal is not proof an agent was unused.

Codebase size

Source size

163 KB

Source files

32

Counts recognized source files only; vendored directories, binaries and lockfiles are excluded, so this is smaller than the repository on disk.

0 stars