Accommodation Booking Agentic AI

My current housing-request workflow still relies on e-mails and copy-pasting between a project manager, an admin, and our corporate travel engine. I want that entire front half of the process—finding suitable places to stay—run hands-free through an agentic AI. Here is the flow I need replicated and improved: • A project manager supplies job-site address, start and end dates, and any special notes. • The AI instantly searches our travel partner engine for hotels within the prescribed radius, then, when the stay exceeds a month, switches to or includes AirBNB/VRBO style rentals. • While searching it must weigh distance from the site, total cost, room type, and amenity requirements exactly as a human would. • It returns the three best options with a succinct comparison table (price, distance, key amenities). • Once the PM selects one, the existing admin or a second agent can complete the actual booking through the same engine or the rental platform. Success for me is simple: 1. I type the project details into a chat or webhook and receive three ranked options in under two minutes. 2. The search logic demonstrably respects the distance, cost, and amenity priorities. 3. The hand-off to booking is structured so that later we can fully automate it with minimal code changes. We prefer to utilize Gemma 4 as the model, but beyond that, the rest of the tech stack is up to you. Our current environment runs in DigitalOcean and Azure. DigitalOcean probably makes the most sense for development-to-production, but open to changes. When you respond, tell me which agent framework you’d start with and how you would test that the recommendations are actually optimal for distance, cost, and amenities.

Project ID: 40457224