CASE STUDY · CRM · TELEPHONY · AI

Zvonex CRM: sales, telephony, and AI in one workspace

A production-ready multi-tenant CRM where managers work with contacts, calls, tasks, messages, and AI insights in one auditable timeline.

Build something similar ↗View architecture

Zvonex CRM platform overview — Selected project material · anonymized where required

65schema migrations

64test files

4demonstrated workflows

SIP / WebRTClive calling

THE CHALLENGE

What had to remain reliable.

Live calls, contact state, billing, AI jobs, and a real-time interface can all complete at different times. Duplicate terminal events, repeated payment webhooks, delayed recordings, and unavailable integrations must not corrupt a tenant's data or financial state.

THE SOLUTION

How the system was designed.

The system uses account-scoped RBAC, idempotent call finalization and payment webhooks, a ledger-backed billing domain, queued AI processing with locks and retries, and explicit recording retention. Telephony is connected to the contact timeline instead of living as a disconnected tool.

DELIVERED

The working scope.

A contact-centered CRM with companies, owners, stages, tags, tasks, calls, email, messages, and an auditable history.
Asterisk 20 telephony with SIP/WebRTC, AMI events, extensions, trunks, access rules, live call state, recordings, and reporting.
Speaker-aware transcription, quality and risk analysis, objection extraction, suggested next actions, and company-knowledge RAG.
Auditable billing for plans, seats, modules, balances, trials, invoices, and idempotent automated payments.
Separate CRM, administration, and onboarding interfaces with importing, deduplication, bulk actions, and role-based access.

ARCHITECTURE

The main system boundaries and data flow.

[ Vue 3 CRM / Admin / Onboarding ]
                 | REST + live events
                 v
             [ NestJS API ]
        +---------+---------+--------------+
        v         v         v              v
 [ CRM domain ] [ Billing ] [ Integrations ] [ AI jobs ]
        |         |         |              |
        +---------+---------+--------------+
                          v
                    [ PostgreSQL ]

[ Browser WebRTC ] <-> [ Asterisk 20 / SIP ]
          |                    | AMI / recordings
          +----------+---------+
                     v
        [ call record + contact timeline ]
                     v
 [ STT -> transcript -> RAG context -> AI insight ]

OUTCOMES

What the architecture enables.

One working context for sales and calls

Failure-safe call and billing lifecycle

AI insight connected to source recordings and company knowledge

Operational controls for both companies and platform administrators

STACKNestJSTypeScriptVue 3PostgreSQLAsterisk 20SIPWebRTCRAGDocker

YOUR PROJECT CAN START WITH A SHORT TECHNICAL REVIEW

Show me the workflow, integration points, and current bottleneck.

Discuss the project ↗