Documentation
¶
Overview ¶
Package driver runs a workflow type for one scenario against a prepared env and returns the artifact to judge. New workflow types plug in by registering a WorkflowDriver; thread_mode_initial is the first.
Index ¶
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type JobExecution ¶
type JobExecution struct{}
JobExecution drives a workspace Job against a seeded case and judges what the job produced: its run outcome (success/failure + summary), the case state after the run, the actions present, and the tool-call trajectory. The job is run through the real JobRunner, so this faithfully exercises the production executor (simple / planexec), source resolution, and tool wiring.
type Registry ¶
type Registry struct {
// contains filtered or unexported fields
}
Registry maps workflow kinds to drivers.
func Default ¶
func Default() *Registry
Default returns a registry with all built-in drivers registered.
func (*Registry) Lookup ¶
func (r *Registry) Lookup(kind string) (WorkflowDriver, bool)
Lookup returns the driver for a kind.
func (*Registry) Register ¶
func (r *Registry) Register(d WorkflowDriver)
Register adds a driver. Panics on duplicate kind (a programming error).
type ThreadInitial ¶
type ThreadInitial struct{}
ThreadInitial drives the thread-mode initial sequence: a top-level post starts the initialization (create) agent. The agent investigates and may ask the user a question instead of creating a case immediately; when it does, the simulator answers and the answer is injected as a thread reply that resumes the create agent, looping up to the persona's MaxAnswerTurns until the agent commits a case. The produced case is the artifact.
type WorkflowDriver ¶
type WorkflowDriver interface {
Kind() string
Run(ctx context.Context, e *env.Env, sc *scenario.Scenario, sim evaltype.Simulator) (evaltype.Artifact, error)
}
WorkflowDriver runs one workflow type for a scenario against a prepared env and returns the produced artifact. It may call sim.Answer when the agent asks the user a question.
Source Files