Your First Project
Build a small architecture model that is useful for real review work
A good first project is small but honest. Model the elements people actually discuss, add the most important relationships, map one critical workflow, and review the result before adding detail.
A Useful Starting Shape
- 5 to 10 named systems or containers
- The important static relationships between them
- One Architecture View with a specific audience and purpose
- One critical workflow
- Data ownership for the most important information
- One deployment view when runtime topology matters
1. Name Elements For Humans
Use names like “Checkout API”, “Orders Service”, and “Payment Gateway”. Keep implementation evidence in descriptions.
| Name | Description evidence |
|---|---|
| Checkout API | Source module: apps/checkout-api |
| Orders Service | Source module: services/orders |
| Kafka Events | Provisioning definition: infra/kafka |
Model deeper levels only when they answer a real question. Archflow supports System → Container → Component → Block, with blocks available for additional decomposition.
2. Declare Static Relationships
Relationships describe the intended architecture. Add enough to explain the design before using workflows to record runtime behavior.
3. Map One Critical Workflow
Choose a flow such as “Place order” or “Authenticate customer.” Map its ordered steps and connect them to the participating architecture elements.
4. Review Evidence
Use Complexity, Blast Radius, and Coupling as review prompts:
- Is a dense dependency cluster intentional?
- Which workflows degrade when an element fails?
- Does workflow co-activity reveal a missing static relationship?
- Is the selected analysis scope appropriate?
Then review the resulting items under Decisions.
5. Document The Conclusion
Capture durable conclusions in an Architecture Document or the Architecture Handbook. Add supporting sources to Project Knowledge and approve only the memories or insights that should influence future Archie work.
6. Create A Version
Create a named version when this small model becomes an agreed baseline. Future experiments can then be compared or merged without overwriting the reviewed state.