Knowledge base & RAG
AI answers draw on a knowledge base the team can manage, using relevant business information as context.
Case notes / 03
RAHO CONTROL ROOM
AI-Powered
WhatsApp Chatbot.
Helping customer service teams answer recurring questions using business knowledge, while keeping the team in control.
Knowledge base & RAGHuman handoffDurable outbox
Behind the interface
From architecture to control room.
Project documentation and screens from RAHO Control Room.
Project overview
Open full size (new tab)
WhatsApp chatbot documentation: the Rule → RAG → AI flow, with an outbox and worker for message delivery.
Contact details, conversations, and some operational data have been blurred in the screenshots.
Context & contribution
Automation that works
with the team.
The need
Customer service teams in companies and small businesses face recurring customer questions. This chatbot is designed to help answer them with relevant business information, while providing a way to hand conversations over to a person.
Teams can add a knowledge base. Retrieval-Augmented Generation (RAG) uses this information as context for AI answers, alongside rule-based message handling.
My role
I worked as the Fullstack Developer & System Designer, developing the application interface and backend, and designing the overall system architecture.
- A dashboard interface for operations and conversation management.
- A backend connecting chatbot logic, the knowledge base, and WhatsApp services.
- An architecture separating response processing, delivery queues, and operator workflows.
Key features
From questions to follow-up.
Five features connecting AI automation and customer service.
Rule-based message handling
Deterministic rules work alongside AI responses in the chatbot flow.
Shared inbox & human handoff
Customer service teams can browse conversations, take over from AI, and send manual replies from one inbox.
Durable outbox & worker
Messages are sent through a queue and processed by a background worker, with support for delivery retries.
WhatsApp alerts for customer service
When AI creates a new handoff, a question summary is sent to customer service via WhatsApp so the team can follow up.
AI helps answer.
The team can always take over.
System foundations
Every service has a role.
Responsibilities across the system, from the dashboard to WhatsApp message delivery.
- Next.js · React · TypeScript
- Dashboard interface for administration and operations.
- Express
- API for authentication, chatbot routing, RAG, and business logic.
- PostgreSQL · Prisma
- Application data storage and access.
- Redis · BullMQ
- Job queues and delivery retries.
- Baileys · Worker
- WhatsApp messaging services and outgoing message processing.
- MinIO
- Document and file storage.
- OpenAI-compatible AI provider
- AI models supporting chatbot responses.