All Classes and Interfaces
Class
Description
Everything the runtime traced under the agent commands of one session, read through
TelemetryReader.getAgentTrace(String).Represents the result of a Consumer handling a message when run through the testkit.
Consumer Testkit for use in unit tests for Consumers.
One evaluation case: the command sent to the agent and the evaluators to run against the agent's
reply and tool calls.
Turns recorded production interactions into cases.
The label used in the report to group an evaluator's results.
One check over one case's evidence.
What an evaluator found on one turn.
The built-in evaluators, and the labels used in the report to group evaluator results.
Allows to simulate publishing messages for the purposes of testing incoming message flow.
Allows to assert published messages for the purposes of testing outgoing message flow.
Represents the result of an EventSourcedEntity handling a command when run in through the
testkit.
EventSourced Testkit for use in unit tests for EventSourced entities.
Cases bound to the agent under test, ready to run.
The cases of an experiment, before the agent under test is named.
Runs cases against an agent and collects the results.
One case's evidence and results.
The batch outcome: per-case results and the gate's verdict.
What a batch run must satisfy, checked over all case results.
One guardrail evaluation during a turn.
One turn: the input the agent was given, its reply as text and what the runtime traced while
producing it.
Decides how well a reply meets a criterion given as a sentence.
The judge's answer.
The agent behind
Judge.modelBased(akka.javasdk.testkit.TestKit): sends the model one request and reads back a Judge.Verdict.Represents the result of a KeyValueEntity handling a command when run in through the testkit.
KeyValueEntity Testkit for use in unit tests for Value entities.
Registry of mocked gRPC services for a running
TestKit.Registry of mocked HTTP services for a running
TestKit.This trait is meant to allow for unit testing when a service has cross-component or cross-service
calls.
A
Judge that asks a model through JudgeAgent.One model call made while answering a turn.
Configuration for a named object-storage bucket used in integration tests / dev mode.
Credentials for a GCS-backed bucket.
GCS service-account key file credentials.
GCP Application Default Credentials (workload identity / ADC).
Not part of the public API.
How the bucket is addressed at the endpoint.
Credentials for an S3-backed bucket.
AWS named-profile credentials (reads from
~/.aws/credentials).Static AWS credentials (access key + secret).
AWS workload-identity / instance-profile credentials.
One tool call as recorded in production: the tool, the arguments the agent passed, and the result
as recorded JSON.
Helper class for serializing and deserializing objects for testing schema migration.
Reads the spans the runtime exports to the in-memory span exporter of a running TestKit.
Testkit for running services locally.
Settings for testkit.
A JUnit 5 "Jupiter" Extension for
TestKit, which automatically manages the lifecycle of
the testkit.This class provided the necessary infrastructure to run integration test for projects built with
the Java SDK.
A
ModelProvider implementation for testing purposes that does not use a real AI model.Represents an AI response, which can include a message and/or list of tool invocation requests.
Factory methods for
TestModelProvider.ToolInvocationRequest instances that correspond to the internal
tools exposed by the autonomous agent runtime to the LLM.Participant reference for scripted conversations, with type and reference name.
A step in a scripted conversation, specifying speaker and guidance.
Specifies a team member type and count for a
create_team tool invocation.Represents an input message.
Represents a tool invocation request with a name and arguments.
Represents a tool result.
Represents a user message.
Base class for building reply configurations for specific input predicates.
Specialized reply builder for handling tool result messages.
Test implementation of
NotificationPublisher that collects all published notifications
for later assertions.Represents the result of an Action handling a command when run in through the testkit.
TimedAction Testkit for use in unit tests for TimedActions.
Which injected stub serves which tool, for cases derived from recordings.
Loads the recorded result of one call into its stub.
One tool call: the name, the deserialized arguments and, when recorded, the result or the error.