What It Does Locas.dev takes a single input – a natural language question like:
“Is Downtown Chicago a good place to buy land?”
“Would Miami Beach be a good spot for opening a restaurant?”
Then it fetches real-world geo-tagged data like school ratings, hospitals, police stations, restaurants, air quality, transit nodes, green space, and even pollen levels — processes it using a custom-built LLM + rules engine — and responds with a structured, explainable breakdown.
Each answer includes:
Ratings across dimensions like services, amenities, transport, environment, and competition
Google Maps links to nearby landmarks
Human-readable insights with pros and cons
A clear summary and final verdict
Tech Stack Built with Python, LangChain, and a custom vector lookup engine
LLM-driven reasoning with deterministic safety layers
Google Maps, OpenStreetMap, and public datasets for grounding
Fully stateless – just one GET request per analysis
Why I Built It As someone working in AI and education, I constantly run into the challenge of selecting the “right place” for programs, centers, or even partnerships. Existing solutions are either too manual, map-based, or lack narrative reasoning. I wanted something that thinks about a location the way a human researcher or consultant would — but on-demand, in seconds.
Try It There are a few public demos on the homepage — feel free to test it with your own questions.
→ https://locas.dev
Would love feedback from fellow HNers! I'm especially curious about:
Potential use cases you see
What features you'd want next (e.g., scoring your own address, neighborhood comparisons, investment suggestions)
Integrations (Slack, Notion, etc.)
Thanks for reading! – Azhar