Travel Research Toolkit

I built travel-search tools to help an assistant compare concrete options against constraints such as destination, dates and budget. The practical challenge is recovering the exact offer behind a result, including its selected date and room.

Python · Playwright · Chromium · Structured web data · HTTP APIs

Inspect the rendered offer

The browser-based search loads the travel provider with the requested filters, waits for rendered offer cards and extracts prices and destination links. The links retain the provider’s offer identifier so an attractive result can be opened as the specific option that was examined.

I also explored a lighter HTTP-only parser. The provider’s initial HTML exposes generic promotions while the requested filters depend on JavaScript, so browser execution is needed for the filtered comparison.

Keep the assistant connected to evidence

The toolkit brings structured search results into the planning workflow: actual offers to compare, with constraints and links retained. That is a useful boundary between language-model assistance and the tool’s job of obtaining current information.

This is part of how I use agentic workflows in everyday life. I build small tools where a conversational assistant needs a more precise view of the outside world, then keep the decision grounded in the information the tools return.