The Workflow Model
We treat workflows as engineering systems, not scripts. Every automation follows this strict 5-stage architecture to ensure predictability.
1. Trigger Layer
The trigger defines strictly 'when' a workflow runs. It does not perform logic; it merely accepts the signal.
- ▪ Must be explicit (Webhook, Schedule, or Event).
- ▪ Should be idempotent (safe to retry).
- ▪ No business logic allowed here.
2. Validation Layer
All incoming data is treated as 'hostile' until validated. If inputs are missing or malformed, we reject them here, protecting downstream systems.
- ▪ Check required fields & data types.
- ▪ Verify authentication/signatures.
- ▪ Reject & Log invalid payloads immediately.
3. Decision & Routing
This layer determines 'where' data goes based on rules. This is where Human-in-the-Loop logic lives.
- ▪ Rule-based routing (no guessing).
- ▪ Handle edge cases explicitly.
- ▪ Route to 'Human Review' if confidence is low.
4. Action Layer
The side-effect. This is where we actually write to a database, send an email, or call an API.
- ▪ Actions must be atomic.
- ▪ Assume external APIs will fail (timeouts).
- ▪ No hidden logic inside the action node.
5. Observation Layer
If you can't see it fail, you can't trust it. We log success, failure, and error payloads for every run.
- ▪ Log execution ID and status.
- ▪ Capture error payloads.
- ▪ Metrics: Volume, Latency, Failure Rate.
The Fabalos Standard
"If a workflow cannot be explained using this architecture, it is not ready for production. We design for failure, assuming that inputs will be wrong and APIs will break."