Aurevia System Blueprint
AI workflow architecture for a premium real estate demand engine.
This page maps how Aurevia receives inbound property demand, enriches it through AI and retrieval, executes workflow actions, and exposes the results through a premium operations command center.
Inbound Sources
Website, WhatsApp, Email
Orchestration
Extraction, retrieval, replies
Action Layer
Lead capture, follow-up, routing
Data Layer
PostgreSQL and Qdrant state
AI Workflow Overview
Inbound messages are normalized into a consistent event shape.
The AI layer extracts intent, budget, location, and property preferences.
The orchestrator chooses retrieval and action tools based on user context.
Suggested properties, insights, follow-ups, and escalations are written back to the operating system.
Inbound Channels
Captures direct property interest from landing pages, forms, and on-site chat flows.
Handles conversational demand from high-intent prospects who prefer fast mobile interaction.
Processes slower, detail-rich inquiries and keeps them inside the same operating workflow.
AI Intelligence Layer
- Intent detection and journey classification
- Entity extraction for budget, location, property type, and timing
- RAG retrieval against indexed brokerage knowledge
- LLM orchestration for grounded replies and next-action decisions
Action Layer
- Property suggestion lookup
- Lead creation and enrichment
- Follow-up scheduling
- High-value escalation routing
- Analytics and operational logging
Data Layer
PostgreSQL
Operational CRM state for leads, conversations, events, and workflow records.
Qdrant
Vector knowledge index for retrieval-augmented answers and document grounding.
Frontend Runtime
Premium Next.js dashboard and landing surfaces, including demo-safe fallback mode.
Stack Summary
Production-shaped foundations across interface, API, AI, and data.
| Layer | Summary |
|---|---|
| Frontend | Next.js 14, App Router, TypeScript, Tailwind CSS |
| Backend | FastAPI, Pydantic, SQLAlchemy, Alembic |
| AI | OpenAI GPT-4o orchestration, embeddings, prompt-driven extraction |
| Data | PostgreSQL, Qdrant, portable SQLite support for selected deployments |
| Deployment | Vercel, Render, Railway, Hugging Face Space workflow |