OAI / Software and integration
AI assistant for documents, data and IoT
A wiki and chat platform connecting operational documents, databases and IoT telemetry through an orchestrator shared by the web interface and messaging.
/ context
The engineering problem
The project needed an access point for operational knowledge, domain data and IoT measurements while retaining the existing management systems. Assistant responses had to derive from the sources and tools available in the user’s context, with an execution path that could be reconstructed.
01
A platform for organizing knowledge
We built a multi-tenant wiki with chat, a React PWA frontend in TypeScript and a FastAPI backend. The interface supports multiple languages through i18next; authentication and application connections use JWT and OAuth. PostgreSQL provides the database layer, with SQLAlchemy and Alembic migrations to manage schema evolution. The monorepo structure brings together the frontend, services and shared components, making the connections between operational documentation and application functions explicit.
02
From the question to the tools
The LangGraph runtime classifies the request domain and builds a plan by selecting tools from a catalog. Execution is sequential; the collected results feed the synthesis and can trigger limited replanning. We integrated wiki search, read-only domain database access through SQL composition with tenant context, and a ThingsBoard connection through an MCP agent. OpenRouter supports model calls for conversation and decisions within the flow.
03
Channels, integrations and deployment
Web chat and the WhatsApp channel use the same orchestrator, keeping the connectors within a shared runtime. Calendar integration includes synchronization and controlled write operations; the APIs also deliver conversation events through SSE. The project includes the data schema and migrations, deployment configuration with Nginx HTTPS and systemd units, and technical documentation. Planning tools and delivery channels complete the scope of the application integrations.
/ output
What the work includes
- Multi-tenant wiki and chat
- Runtime with a tool catalog
- SQL, IoT and calendar connectors
- Messaging-channel integration
- Data schema, migrations and deployment configuration
/ contact
A system to build?
Start with functions, interfaces and constraints. Tell us about your project.