Event Sourcing and CQRS
Methodologies and patterns for business logic where system state is rebuilt from events.
Product
Our own technology platform
The challenge
The task the client came with and why it could not be solved before.
Assembling each project from off-the-shelf libraries means inheriting their limits. We built our own foundation so that we set the architecture instead of accommodating someone else's.
Mechanics
The mechanics the product is built from.
Methodologies and patterns for business logic where system state is rebuilt from events.
Functionality is added and changed without pausing the running version or its services.
Deployment across servers for geographic proximity to the client.
Connection to any external service — exchanging data through requests or information feeds.
Pipeline
The path of a request through the platform: from an external system to an answer assembled from the event journal.
External systems connect over HTTP or WebSocket. The entry point is the router: TLS, request distribution, rate limiting.
Logic is split into small services. Each owns its part and does not call its neighbours directly.
A service does not overwrite state silently — it records an event. The journal restores both the current view of the data and the full history of operations.
The broker delivers it to the other services: queues and topics, guaranteed delivery, reprocessing after a failure.
Reading and writing are separated: read models are assembled in advance and updated from events, without competing with operations.
Reads are served from pre-assembled views — fast and without locking writes.
Where it sits
Odin is the bottom layer: both web products and client services stand on it.
The perimeter runs on our own servers and can span several locations, so data sits closer to users.
Message broker, storages, router, orchestration, monitoring. Assembled into a configuration that fits the task.
Company products and client services: Udeu, Ambitions, back offices, concierge systems.
Outcome
What those platform properties give a particular product.
Services are tied by contracts and events, not by a shared codebase: changing one does not drag the others along.
Two versions of a service run at the same time, so rollout and rollback happen without stopping service.
State is restored from the event journal: investigating what happened needs no separate system.
Read models are assembled in advance and updated from events, so reports do not get in the way of operations.
The configuration is chosen for the task — from a single-server installation to a distributed perimeter.
Several locations with data replication and traffic switching between them.
Technology
The platform, the stack and the devices the product runs on.
Technologies
Platforms
Links
The product website and the materials about it.
Live demo
Demo access is provided on request.
Track record
Projects where this product is already running.
We will deploy it white-label, tune the processes and onboard your team.