Flow Definition

The flow definition captures WHAT the integration does and HOW it’s implemented.

Creating a Flow

Run /camel-flow <flow-name> in your AI assistant. You’ll be guided through:

1. Flow Intent - What data is processed and what is the goal
2. Source System - Where data comes from and which Camel component handles it
3. Processing Steps - EIPs required (filter, split, aggregate, transform, etc.)
4. Sink System - Where data goes and which Camel component handles it
5. Error Handling - Dead Letter Channel and retry strategy

Then, only if relevant, the agent asks follow-up questions for:

• Data transformation - Schemas for DataMapper/XSLT generation; unmarshal only as a fallback
• Circuit Breaker - Only if an external HTTP/REST service is involved
• Idempotent Consumer - Only if consuming from a message broker or deduplication is needed
• Transactions - Only if writing to more than one external system
• Performance, Security, Monitoring - Only if mentioned during the interview

Flow File

The flow is saved to .camel-kit/flows/<flow-name>/flow.md.

Example flow diagram:

    flowchart LR
    subgraph Source
        K[("kafka:orders")]
    end
    subgraph "Processing (EIPs)"
        E1["unmarshal"]
        E2["validate"]
        E3["filter"]
    end
    subgraph Sink
        DB[("sql:INSERT")]
    end
    subgraph Error
        DLQ[("kafka:orders-dlq")]
    end

    K --> E1 --> E2 --> E3 --> DB
    E1 -.->|error| DLQ
    E2 -.->|error| DLQ
  

Best Practices

Constitution v2.0 — Six Enforced Rules

The constitution (.camel-kit/constitution.md) enforces exactly six rules on every generated route:

All other design guidance (error handling strategy, retry policy, circuit breaker, transactions, idempotency, throttling, Kubernetes, data format choices) is applied context-specifically during /camel-flow and /camel-migrate flow design — not enforced globally.