All Projects
AI / Full StackProduction
Bol Mitra
WhatsApp-to-CRM Bridge SaaS
PocketBaseGoJavaScriptWhatsApp WebOpenRouterDockerCaddyTailwind CSSDaisyUIRazorpay
Overview
An integration platform that connects a user's existing WhatsApp (Web) number to CRM/ERP systems (ERPNext, Zoho, HubSpot) via a node-based visual flow builder and outbound webhooks. Features include a multi-number shared inbox, WhatsApp group monitoring, message-to-ticket conversion, and AI-driven chat analytics (sentiment, summarization, custom extraction) via OpenRouter.
3+
Supported CRMs (integrations)
3
AI Analytics Types (sentiment, summarization, extraction)
The Problem
Businesses rely on WhatsApp for customer communication but lack a way to bridge those conversations into their CRM/ERP systems. Manual entry is error-prone and doesn't scale.
What I Built
A full-stack SaaS that authenticates WhatsApp via QR login, provides a visual node-based flow builder for routing messages to CRM webhooks, and runs AI analytics on conversations. Multi-number shared inbox and group monitoring let teams collaborate. Outbound webhooks push events to ERPNext, Zoho, HubSpot, and custom endpoints.
Architecture
PocketBase (Go) backend with custom JS hooks for scheduled jobs, webhook dispatch, and Razorpay billing. Containerized WhatsApp session bridge with per-number isolation and a reaper service for cleanup. Caddy reverse proxy. Docker Compose orchestration. Tailwind CSS + DaisyUI frontend served statically.
Technical Decisions
- Used PocketBase for auth, database, and realtime — extending it with JS hooks for business logic.
- Built a containerized WhatsApp bridge (QR login) with per-number isolation to avoid session conflicts.
- Added a reaper service to clean up stale sessions and prevent resource leaks.
- Chose OpenRouter for AI analytics to support multiple LLM providers without vendor lock-in.
- Node-based visual flow builder lets non-technical users define routing logic without code.
WhatsApp session management
Maintaining stable WhatsApp Web sessions per number, handling QR re-auth, and cleaning up abandoned sessions required a dedicated bridge container and reaper service.
Multi-tenant webhook dispatch
Dispatching webhooks reliably at scale with retries, idempotency, and per-tenant rate limiting across diverse CRM endpoints.
AI chat analytics accuracy
Prompt engineering for sentiment, summarization, and custom extraction to produce consistent structured output from unstructured chat logs.
My Contribution
- Architected and built the PocketBase backend with custom JS hooks.
- Developed the WhatsApp session bridge container and reaper service.
- Built the visual flow builder and webhook dispatch system.
- Integrated OpenRouter for AI-driven chat analytics (sentiment, summarization, extraction).
- Set up Docker Compose, Caddy, and Drone CI for production deployment.