Module agentforge4j.testkit

Package com.agentforge4j.testkit.harness

Class WorkflowTestHarness

java.lang.Object

com.agentforge4j.testkit.harness.WorkflowTestHarness

public final class WorkflowTestHarness
extends Object

The bootstrap-facing harness adapter: assembles an AgentForge4j runtime with the deterministic fake LLM provider active, runs a workflow, and returns a WorkflowRunResult holding the final state plus the captured event stream and files.

This is the single class in the testkit permitted to depend on agentforge4j.bootstrap; the assertion and capture layers stay assembly-agnostic. Fake is activated through the only seam that accepts it — an explicit LlmClientResolver over a FakeLlmClient backed by a run-agnostic StaticFakeResponseSource, which avoids any pre-start run-id registration timing problem.

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static final class	WorkflowTestHarness.Builder	Fluent builder for WorkflowTestHarness.

Method Summary

Modifier and Type	Method	Description
static WorkflowTestHarness.Builder	builder()	Starts a new harness builder.
WorkflowRunResult	run(String workflowId)	Assembles a fake-backed runtime, runs the given workflow to completion or to its first suspended state, and returns the captured result.
WorkflowRunResult	run(String workflowId, List<GateResponse> responses)	Assembles a fake-backed runtime, starts the given workflow, and drives it forward by draining one response at each human-in-the-loop pause (input / review / step-approval / escalation) until the responses are exhausted.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Method Details

builder

public static WorkflowTestHarness.Builder builder()

Starts a new harness builder.

Returns:

a fresh builder

run

public WorkflowRunResult run(String workflowId)

Assembles a fake-backed runtime, runs the given workflow to completion or to its first suspended state, and returns the captured result. Equivalent to run(String, List) with no scripted responses — the run is left at its first human-in-the-loop pause.

Parameters:

workflowId - id of a workflow present in the configured fixtures; must not be blank

Returns:

the run result: final state plus captured events and files

run

public WorkflowRunResult run(String workflowId,
 List<GateResponse> responses)

Assembles a fake-backed runtime, starts the given workflow, and drives it forward by draining one response at each human-in-the-loop pause (input / review / step-approval / escalation) until the responses are exhausted. When the queue is exhausted at a pause, the run is left at that pause so a scenario may assert a meaningful pending state. A response left over after the run has already reached a terminal state is a scripting error — the scenario queued more responses than the run paused for — and fails loudly rather than being silently dropped, symmetric with the pause-mismatch check.

Parameters:

workflowId - id of a workflow present in the configured fixtures; must not be blank

responses - scripted human responses, one consumed per pause in order; must not be null (may be empty)

Returns:

the run result: the last observed state plus captured events and files

Throws:

IllegalStateException - if a queued response does not match the pause the run is at, or if the run reaches a terminal state while scripted responses remain unconsumed