RAG / SQL / LangGraph
Aker RAG
A property-scoped AI platform for property-management Q&A across deterministic SQL, governed LLM-to-SQL, and RAG.
Technical Highlights
- Ingests rent-roll spreadsheet snapshots into MySQL for structured property data.
- Routes SQL, RAG, and hybrid requests through a LangGraph orchestrator.
- Uses deterministic SQL templates first, then governed LLM-to-SQL when needed.
- Validates generated SQL with allowlists, property-code scope checks, limits, and safety rules.
- Retrieves property-filtered website snippets from Qdrant and returns markdown, UI blocks, and citations.
Stack
Python, FastAPI, MySQL, Qdrant, LangGraph, Gemini, OpenAI, xAI Grok, Docker, and pytest.