Open-source AI real-estate search. 320 GitHub stars and 120 forks, MIT licensed. RAG + ChromaDB + 16 LLM providers + 9 languages. The OSS sibling of PropVector AI.
AleksNeStu/ai-real-estate-assistant
Maintained by @AleksNeStu on GitHub · on LinkedIn.
Updated:
Real-estate listings are public data. The search layer on top of them shouldn't be a walled garden. OpenStore is the public, working version of our real-estate stack — and the credibility proof for PropVector AI.
Index your own listing data into ChromaDB. Hybrid RAG keeps keyword and semantic matches together.
Search the way a buyer talks, in any of nine languages, through any of sixteen LLM providers.
Mortgage, ROI, CMA, total cost of ownership and a multi-year price forecast next to each result.
MIT licensed. Clone it and run it yourself.
Point it at your own listing data — the index stays on your machine.
You want a working reference for RAG over property data, with the reranking and provider switching already done.
You need to prototype semantic search on your own data before you commit to a vendor.
You want the search self-hosted, with the data staying on your infrastructure.
OpenAI, Anthropic, Google, Grok, DeepSeek, and local Ollama. Switch providers without code changes.
Vector search with MMR reranking. Retrieval-augmented generation pulls live data into the prompt.
English, Polish, Russian, German, Spanish, Italian, Portuguese, Turkish, Ukrainian. Native UI + native LLM prompts.
Ask like "2-bedroom apartment in Kraków under 500k". Automatic filter extraction from the query.
Mortgage calculator, rent-vs-buy, ROI, TCO, CMA reports. AI property valuation with 1y / 3y / 5y / 10y forecasts.
Property clustering, area analytics, neighborhood one-liner summaries on property detail pages.
OpenStore ships the core architecture. PropVector AI is the commercially hardened version: live MLS data feeds, accounts, favorites, saved searches, AI CRM, enterprise features. Same primitives, more constraints.
Three signals buyers can verify before they start a conversation.
OpenStore is an index you run yourself: point it at your own listing data and the index stays on your machine. The source is public, so you can read how the index is built before you run it — AleksNeStu/ai-real-estate-assistant on GitHub.
The person you contact stays close to the work, from written scope through operation. See how we work.
OpenStore is MIT-licensed and public, with 320 GitHub stars and 120 forks. Inspect the foundation.
Yes. It is MIT licensed and public on GitHub. You can run it, modify it and ship it inside your own product.
OpenStore ships the core architecture: vector search, MMR reranking, 16 LLM providers, 9 languages and the financial tools. PropVector AI adds live MLS feeds, accounts, favourites, saved searches, an AI CRM and an uptime SLA.
Wherever you run it. OpenStore is self-hosted, so the listing data and the vector index stay on your infrastructure.
OpenStore supports 16 LLM providers behind one interface: OpenAI, Anthropic, Google, Grok, DeepSeek and a local Ollama install. You can switch between them without changing code, and a local Ollama install means the model runs on your own machine. The project is written in Python and TypeScript, with FastAPI and Next.js, and it is MIT licensed on GitHub. Retrieval is hybrid RAG over ChromaDB, with vector search and MMR reranking, and one index serves all 9 languages, each with native UI and native LLM prompts.
Issues, PRs, and feature requests welcome on GitHub.
Open issues →