Skip to content

Chakra Commerce — Enterprise Modernization

A reference architecture for modernizing enterprise monoliths to cloud-native microservices — where humans define what correct looks like and AI agents do the building.

The Thesis

Most modernization projects fail at one of two places: they either begin implementation before the domain is understood, or they distribute domain knowledge across PRs and conversations where it can't be consumed systematically.

This reference architecture inverts both problems.

Humans author contracts — precise, versioned specifications of SLAs, business invariants, and event schemas. These live in contracts/ and are the only source of truth the system ever needs to trust.

AI agents implement from those contracts — every service skeleton, Helm chart, CI pipeline, and Prometheus alert is derivable from the contracts by an agent given a task spec. The line between what humans wrote and what agents built is explicit and structural.

SLA measurement is woven in from day one — not added after the fact. Every SLA target in contracts/slas/ flows through a deterministic script to observability/slos/ and observability/alerts/. The first commit includes the alert that would page on-call if the SLA were breached.

At a Glance

  • 10 Architecture Patterns

    From Strangler Fig migration to Leave-and-Layer, Event Sourcing, CQRS, Saga Orchestration, Choreography, Circuit Breaker, Bulkhead, Sidecar, and Outbox — each backed by an ADR.

    Browse the Pattern Catalogue

  • 15 Architecture Decision Records

    Every architectural choice is an accepted ADR with context, decision, consequences, and cross-references. No decisions in Slack threads or wikis.

    Browse ADRs

  • How This Was Built

    All 6 AI dev model personas — Human Domain Expert, Documentation Agent, Script Authoring Agent, Script Executor, Implementation Agent, and Compliance Agent — contributed to this project. See the full breakdown.

    How This Was Built

  • SLA → Alert in one pipeline

    A YAML file in contracts/slas/ is the only place a target is written. A deterministic script derives SLO definitions and multi-window burn rate alerts. 20/20 coverage checks pass in CI.

    See the SLA Pipeline

The Modernization Journey

Chakra Commerce began as a Java EE monolith. The modernization extracts one bounded context at a time using the Strangler Fig pattern, never pausing the business.

Phase What changes Key pattern Status
1 — Facade Kong deployed in front of monolith Strangler Fig Done
2 — Customers Customers service extracted Strangler Fig + CDC Done
3 — Inventory Inventory extracted with CQRS Strangler Fig + CDC + CQRS In progress
4 — Orders Orders extracted with event sourcing Event Sourcing + Saga Planned
5 — Decommission Monolith retired — Planned

The Warehouse Management System runs in parallel throughout — never modified, accessed via a Leave-and-Layer proxy.

Read the full migration strategy

The Modernization Challenge

Chakra Commerce represents a class of enterprise problem found in nearly every organisation that built software between 2000 and 2015: a monolith that was the right choice at the time, now limiting the team's ability to move independently, measure reliability, or scale specific workloads.

This reference architecture addresses four specific challenges:

  • Extraction without downtime: The strangler fig pattern lets each bounded context go live independently, with the monolith continuing to serve traffic throughout.
  • SLA accountability during migration: SLA targets are defined before extraction begins. Every migration phase has observable go/no-go criteria tied to those SLAs.
  • Data consistency across split storage: CDC pipelines and the event bus replace the shared database as the integration mechanism between newly extracted services.
  • Legacy system integration: The Leave-and-Layer pattern keeps the on-premises Warehouse Management System untouched while the rest of the architecture modernises around it.

Read the full migration strategy

Repository Structure

contracts/          Human-authored source of truth: SLAs, invariants, event schemas
docs/               Architecture docs, ADRs, DDD models, migration plans, runbooks
ai-agents/          Task specs for LLM agents and deterministic scripts
services/           TypeScript service skeletons built from contracts/ by AI agents
infrastructure/     Terraform modules, Helm charts, Kubernetes manifests
observability/      SLO definitions, burn-rate alerts, Grafana dashboards, OTEL config
tooling/            Scripts that transform structured inputs into artifacts
api/                OpenAPI 3.1 + AsyncAPI 3.0 contracts

Browse the source

Every page in this documentation links to the relevant source files in the GitHub repository.