executus/run/palette_test.go

package run_test

import (
	"context"
	"encoding/json"
	"testing"

	"gitea.stevedudenhoeffer.com/steve/majordomo/llm"
	"gitea.stevedudenhoeffer.com/steve/majordomo/provider/fake"

	"gitea.stevedudenhoeffer.com/steve/executus/run"
	"gitea.stevedudenhoeffer.com/steve/executus/tool"
)

// recordingPalette captures the delegation call it received.
type recordingPalette struct {
	gotName, gotCaller, gotParent string
	gotInputs                     map[string]any
}

func (p *recordingPalette) ResolveSkill(context.Context, string, string) (string, error) {
	return "", nil
}
func (p *recordingPalette) InvokeSkill(_ context.Context, callerID, _, name string, inputs map[string]any, parentRunID string) (string, string, string, error) {
	p.gotName, p.gotCaller, p.gotParent, p.gotInputs = name, callerID, parentRunID, inputs
	return "the skill output", "child-run-1", "ok", nil
}
func (p *recordingPalette) ResolveAgent(context.Context, string, string) (string, error) {
	return "", nil
}
func (p *recordingPalette) InvokeAgent(context.Context, string, string, string, string, string, string, string, []string, func(context.Context, string, string)) (string, string, string, error) {
	return "", "", "ok", nil
}

// TestPaletteDelegation: an agent with a SkillPalette gets a skill__<name> tool;
// the model calls it, the executor routes it through run.Ports.Palette as a
// child of the current run, and the result flows back into the loop.
func TestPaletteDelegation(t *testing.T) {
	pal := &recordingPalette{}

	fp := fake.New("fake")
	fp.Enqueue("m",
		fake.ReplyWith(llm.Response{ToolCalls: []llm.ToolCall{{
			ID:        "c1",
			Name:      "skill__helper",
			Arguments: json.RawMessage(`{"inputs":{"q":"hi"}}`),
		}}}),
		fake.Reply("delegated and done"),
	)
	m, err := fp.Model("m")
	if err != nil {
		t.Fatal(err)
	}

	ex := run.New(run.Config{
		Registry: tool.NewRegistry(),
		Models:   func(ctx context.Context, _ string) (context.Context, llm.Model, error) { return ctx, m, nil },
		Ports:    run.Ports{Palette: pal},
	})

	res := ex.Run(context.Background(),
		run.RunnableAgent{ID: "a1", Name: "boss", ModelTier: "m", SkillPalette: []string{"helper"}},
		tool.Invocation{RunID: "parent-run", CallerID: "caller-7", ChannelID: "chan"},
		"delegate please")
	if res.Err != nil {
		t.Fatalf("run error: %v", res.Err)
	}
	if res.Output != "delegated and done" {
		t.Errorf("output = %q", res.Output)
	}
	if pal.gotName != "helper" {
		t.Errorf("InvokeSkill name = %q, want helper", pal.gotName)
	}
	if pal.gotCaller != "caller-7" {
		t.Errorf("InvokeSkill caller = %q, want caller-7", pal.gotCaller)
	}
	if pal.gotParent != "parent-run" {
		t.Errorf("InvokeSkill parentRunID = %q, want parent-run (child of the current run)", pal.gotParent)
	}
	if pal.gotInputs["q"] != "hi" {
		t.Errorf("InvokeSkill inputs = %+v, want q=hi", pal.gotInputs)
	}
}

// TestNoPaletteNoDelegationTools: nil PaletteSource → no delegation tools, run
// still works (the agent just has no skill__/agent__ tools).
func TestNoPaletteNoDelegationTools(t *testing.T) {
	fp := fake.New("fake")
	fp.Enqueue("m", fake.Reply("ok"))
	m, _ := fp.Model("m")
	ex := run.New(run.Config{
		Registry: tool.NewRegistry(),
		Models:   func(ctx context.Context, _ string) (context.Context, llm.Model, error) { return ctx, m, nil },
	})
	res := ex.Run(context.Background(),
		run.RunnableAgent{Name: "x", ModelTier: "m", SkillPalette: []string{"helper"}},
		tool.Invocation{RunID: "r"}, "hi")
	if res.Err != nil || res.Output != "ok" {
		t.Fatalf("nil-palette run failed: %v / %q", res.Err, res.Output)
	}
}

// TestDelegationDedupeAndEmptySkip: empty + duplicate palette names are skipped,
// not turned into "skill__"/duplicate tools that error at box.Add (gadfly C0).
func TestDelegationDedupeAndEmptySkip(t *testing.T) {
	pal := &recordingPalette{}
	fp := fake.New("fake")
	fp.Enqueue("m", fake.Reply("ok"))
	m, _ := fp.Model("m")
	ex := run.New(run.Config{
		Registry: tool.NewRegistry(),
		Models:   func(ctx context.Context, _ string) (context.Context, llm.Model, error) { return ctx, m, nil },
		Ports:    run.Ports{Palette: pal},
	})
	// "" (empty) and a duplicate "helper" must not break the build.
	res := ex.Run(context.Background(),
		run.RunnableAgent{Name: "x", ModelTier: "m", SkillPalette: []string{"helper", "", "helper"}},
		tool.Invocation{RunID: "r"}, "hi")
	if res.Err != nil {
		t.Fatalf("empty/duplicate palette names should be skipped, not error: %v", res.Err)
	}
	if res.Output != "ok" {
		t.Fatalf("output = %q", res.Output)
	}
}