Building administration
Properties, households, committee contacts, payment records, and announcements.
01 / BUILDING MANAGEMENT & AI
VaadBot.
One building.
A more connected experience.
A platform for Israeli building committees and residents, connecting building administration, shared information, resident communication, and AI-assisted access to knowledge.
DOMAINBuilding management
FOUNDATIONC# / ASP.NET Core 10
ARCHITECTUREMulti-tenant application + AI services
THE IDEA
Less scattered information.
More shared context.
01 — THE PRODUCT
A shared place for
building life.
Building information often lives across conversations, documents, and individual committee members. VaadBot brings those workflows into one complete, building-scoped application.
Resident communication
WhatsApp Cloud API connector and application services for resident-facing interactions.
Shared spaces & documents
Shared-area management and property-scoped documents and knowledge.
AI-assisted knowledge
Gateway-mediated chat and document-ingestion stages, backed by retrieval and indexing components.
Administration, communication, shared documents, and AI-assisted knowledge work together in a connected experience for committees and residents.
02 — THE TOOLKIT
Technology
with a purpose.
C# / ASP.NET Core 10
Management endpoints, the WhatsApp connector, and independently hosted AI services.
React / TypeScript / Vite
The management web interface and the separate public marketing frontend.
SQL Server / Entity Framework Core
Application persistence and durable coordination for AI stage handoffs.
RabbitMQ
AI workflow notifications carrying persisted operation identifiers between stages.
Qdrant / MinIO
Vector search and indexing components, plus object-storage integration for AI document artifacts.
OpenAI / Ollama
Explicit chat and embedding provider integrations selected through configuration.
WhatsApp Cloud API
A dedicated connector for webhook-driven conversations.
03 — SYSTEM ARCHITECTURE
Connected at the boundaries.
The platform connects the web application, WhatsApp connector, and AI services. AI callers enter through the Gateway; pipeline coordination is persisted in SQL Server, with RabbitMQ notifications between stages.
CHAT FLOW
Gateway → query preparation → query embedding → context retrieval → answering → Gateway.
INGESTION FLOW
Gateway → content preparation → content embedding → knowledge indexing → Gateway.
Both AI pipelines run through dedicated C# services, with persistent operation state and coordinated handoffs from the Gateway to the final result.
04 — ENGINEERING NOTES
The details
behind the diagram.
Building-scoped access
Management actions use the current user’s property context. Retrieval resolves authorized knowledge and compatible index generations before querying vectors.
Durable state, small notifications
SQL Server owns AI coordination state. Stage handlers check their processed flags; RabbitMQ publishers carry the operation ID instead of transporting document content.
Provider identity is explicit
Provider configuration and embedding compatibility remain application concerns. Qdrant supplies vector storage and search rather than defining model identity.
EXPLORE THE OTHER PROJECT
AutoBroker