Battery distributor, 2026
WhatsApp CRM with AI
A CRM for a battery distributor that centralizes customers and orders, sends automated WhatsApp messages and uses AI with semantic search to understand conversation history.
Stack: NestJS, React 19, PostgreSQL + pgvector, Prisma, Redis, Evolution API
The problem
Sales and support happened on WhatsApp, but customers, orders and conversations were scattered across phones and spreadsheets. There was no single place to see who a customer is, what they bought and what was said.
What we built
- Customer and order management in a web dashboard;
- Automated WhatsApp messages tied to order events;
- A dedicated WhatsApp service, isolated from the rest of the system;
- An AI assistant with access to history, using semantic search ("find the conversations about pricing");
- Clear success and error feedback on every dashboard action.
The engineering decisions that matter
- Embeddings in the same database as relational data (PostgreSQL with pgvector): AI context and records live together, with no second database to maintain.
- Redis queues and caching to keep message throughput up without slowing the dashboard.
- WhatsApp protocol isolated in its own Evolution API instance, so channel instability never takes the CRM down.
| Layer | Technology |
|---|---|
| Backend | NestJS, Prisma |
| Data | PostgreSQL + pgvector, Redis |
| Evolution API | |
| Dashboard | React 19, Vite, Material UI |
See how we approach web applications and AI.