BackendProject 5 of 5
Crave Connect
A Zomato-style food app with an intelligent recommendation system. A friend and I both worked on the backend before we dropped it.
- Kind
- Side project, dropped
- When
- May to August 2025
- With
- With a friend
- My part
- Backend, with my friend
01 / The idea
A food app in the style of Zomato, with an intelligent way to recommend what to eat.
02 / What we built
The backend. A FastAPI service over a PostgreSQL database of 19 tables: restaurants, menus, orders, delivery, payments, reviews and more.
03 / The smart part
Every menu item gets an embedding stored with pgvector. A craving typed in plain words is turned into a search over those embeddings, and the best matches come back as recommendations.
04 / What happened
My friend and I both worked on the backend. We dropped the project. The last commit is from August 2025.
Code
# backend/app/api/queries/queries.py
@router.post("/resolve", response_model=MenuItemListResponse)
def resolve_query(query: QueryCreate, db: Session = Depends(get_db)):
query_obj = queries_repo.create(db, obj_in=query)
# embed the query and find the closest menu items
menu_item_ids, context = resolve_query_gemini_top_k(
db, user_id=query.user_id, query_text=query.query_text
)
menu_items = [menu_item_repo.get(db, id=mid) for mid in menu_item_ids]
for mid in menu_item_ids:
recommendation_repo.create(db, obj_in=RecommendationCreate(
query_id=query_obj.id,
menu_item_id=mid,
confidence_score=context['confidences'].get(mid, 1.0),
meta={},
))
return MenuItemListResponse(data=menu_items)