Parents who want a sports club, a music course or a youth group for their child have a hard time finding one. A city website lists its clubs, but only with names and links. The details parents actually need (what is offered, for which ages, on which day and at what time) are spread over many club websites, each built differently and often out of date. Kidas collects this information from cities in Germany (currently only Wörth am Rhein). It reads the city's club directory, follows each club's own website and uses an LLM to pull out the activities for children and teenagers. Parents can then browse every offer, ask questions in a chat, and let an AI planner put together a week of activities for several children that fits their family's schedule.
Searchable activity list for everybody: filter by age, weekday, category and keyword. The list is rendered on the
server, so a search can be shared as a link and works without JavaScript.
Chat grounded in real data. Answers only describe activities that are in the database. When a detail such as a
time or a price is missing, the chat says so and does not make one up.
AI weekly planner for several children. It takes their ages, interests and constraints from the profile ("school
until 4 pm") into account, never schedules two activities at the same time for one child, and saves several
alternative plans.
Parents can report offers. A form takes a description or a photo of a flyer, an LLM reads it, and the offer is published after a review on an admin page.
Privacy by design. Sign-in by magic link, no passwords. The planner's model never sees the children's names: they
are replaced by "Kind 1", "Kind 2" before any model call and restored only in the reply. Parents can export or
delete their account, old conversations are deleted automatically, and clubs can ask to be removed.
LangGraph , LangChain, LangSmith
OpenAI chat and embedding models
PostgreSQL full-text search, pgvector, Reciprocal Rank Fusion
Python, FastAPI, Pydantic, psycopg, httpx, BeautifulSoup, uv
Next.js 16 (App Router), React 19, TypeScript, Tailwind CSS, shadcn/ui
Supabase
pytest, Ragas, Playwright
Google Cloud Run, Vercel, GitHub Actions, Docker
Streamlit
