ByteNex›Articles›Declarative Flows vs Imperative Scripts
Declarative Flows vs Imperative Scripts
A declarative flow names the tools and the order. The kernel executes the flow; the flow does not execute itself.
What does a flow contain?
The tools, their order, and their contracts. The kernel executes the flow. The flow itself is data, not executable code.
Why declarative?
Because the flow is data the kernel can verify, not code that runs on its own. This separation allows the kernel to maintain control.
Can a flow call an undeclared tool?
No. An undeclared tool is rejected before execution. A tool that is not declared has no expected outcome to verify against.
How are tools executed?
The flow specifies the order. The kernel executes them sequentially or in parallel based on the flow rules.
What happens if a step fails?
Verification catches the failure. The AST engine repairs the tool if possible, and the cycle updates its strategy weights based on the failed verdict.
Related
- Agent Kernel Architecture: The Complete Guide
A comprehensive guide to the components of an agent kernel: the cycle, the AST engine, the cognitive memory engine, and the isolated context.
- Agent Kernel vs Agent Framework
A framework supplies components a developer assembles. A kernel runs the cycle itself, verifies outcomes, and updates its own weights.