A niche search engine for movie-focused podcast episodes
A developer shared a search engine that finds podcast episodes actually about a specific movie. They said existing podcast search often returns mixed keyword results instead of true topic matches. The service classifies whether an episode is about a movie, either as a full episode or a real segment, instead of only mentioning it. The creator said the catalog is incomplete and the classifier can be confidently wrong.
Key points
- The product is called impdb.dev.
- The creator says the corpus has about 1.1 million podcast episodes.
- Clear movie title matches are handled directly, while ambiguous cases go to an LLM.
- The stated stack includes Hono SSR, Cloudflare Workers, and D1 databases.
- The creator openly says the catalog is incomplete and the classifier sometimes makes wrong calls.
Quick term guide
- search engine
- A website like Google or Bing that helps you find information on the internet.
- classifier
- A system that sorts an input or task into a type so the tool can decide what to do.
- pain point
- A specific problem or frustration that people experience repeatedly and want solved
- feedback loops
- Repeated cycles of checking a result, correcting it, and trying again.
- feedback loop
- A cycle where the output of a process is fed back as input so the system can correct and improve itself
- Cloudflare Workers
- A service for running web code without managing your own server.
- Cloudflare
- A service that protects websites and manages web traffic.
- databases
- Places where an app stores information such as users, orders, or content.