Build a Board Game Recommender for Your Own Collection
Quickchat published a step-by-step tutorial for building a board game recommender that only suggests games you own. The example assistant, Shelf Night, reads your collection from a Google Sheet, then filters it by tonight's player count, available time and preferences.
The sheet has a Collection tab with a stable ID, ownership, player range and your own planning time for each game. A Plays tab holds the log. In the walkthrough, four players with an hour get a shortlist with both Azul and Cascadia. After Cascadia is logged as played, asking to avoid recent repeats leaves only Azul. Rows marked unowned or missing player or time data stay out of the results.
You connect Google Sheets through Quickchat's Actions. The integration asks for the limited drive.file scope and creates its own spreadsheet, so it can't see other files in your Drive. The first version needs only a read-only API Action that pulls the Collection range. Adding play logging comes next. The assistant records a play only after you confirm it, and the tutorial has you check the saved row as well as the reply.
A starter kit has the two spreadsheet templates, the full prompt, three Action recipes and a test checklist. They're reference files you copy in by hand, not an automatic Agent import. A BoardGameGeek collection export can seed the sheet, but the tutorial does that manually and doesn't use BGG's API or keep the two in sync. You'll need a Quickchat account and a Google account. Conversations and Simulation Testing use AI credits, so check your plan's allowances.