TrendCast

• 

  TrendCast front page

Inspiration

I barely keep up with the news myself. Wouldn't it be nice to have a way to stay up to date?

What it does

TrendCast gives a quick, spoken snapshot of what tech-related things people are talking about across Reddit, YouTube, and Instagram. The TTS uses a friendly tone to simulate talking to a friend (so it's less awkward).

How we built it

• Data Pipeline: Cartesia STT and TTS for natural conversation, text filtering, parsing, and summarization powered by openai/gpt-oss-20b via Groq, and Apify web scraping
• Frontend: React + TypeScript for a simple and clean UI/UX experience
• Backend: Python with FastAPI for API endpoints, Pydantic for request validation

Challenges we ran into

• Empty responses: Apify actors often timed out or gave no response
• Messy data: Apify results were sometimes irrelevant to the search query
• Scraping delay: achieving a balance between delivering an accurate result quickly without sacrificing data

Accomplishments that we're proud of

• Pipelining multiple APIs and ensuring accurate and useful responses
• Building a working MVP in a short amount of time

What we learned

• Combining multiple APIs to create a cohesive program
• Cleaning API data to ensure information is usable

What's next for TrendCast

• Including more sources to scrape for more accurate coverage
• Expand coverage past tech-only news
• Add more voices for variety.

Built With

• apify
• cartesia
• fastapi
• openai
• pydantic
• python
• react
• typescript