Crista Online
Travel ecosystem: AI chat, map, routes, and RAG over POI — frontend, multiple backends, Docker stack.
Challenge
Build a trip-planning product: users describe a journey in chat, get place recommendations and routes on a map. Data must come from many sources (OSM, KudaGo, MKRF), and answers must rely on an up-to-date POI base via RAG.
Solution
Assembled an ecosystem: React/Vite UI, ai_agent (FastAPI + Swagger), data_backend ETL for 83 Russian regions and 90 world cities, vectorization_backend with Chroma, placesweb graph, and suitcase for the user “suitcase”. Local docker-compose with pgAdmin, Grafana, Prometheus; production on crista.online with separate API domains.
Outcome
Working product: AI chat, inspiration board, Leaflet map, API health checks, and monitoring. Service-oriented architecture — each backend deploys and documents independently.
Stack
Gallery
