kandev/apps/backend/internal/plugins/host_data_test.go

package plugins

import (
	"context"
	"errors"
	"fmt"
	"testing"
	"time"

	agentsettingsdto "github.com/kandev/kandev/internal/agent/settings/dto"
	analyticsmodels "github.com/kandev/kandev/internal/analytics/models"
	"github.com/kandev/kandev/internal/plugins/manifest"
	taskmodels "github.com/kandev/kandev/internal/task/models"
	"github.com/kandev/kandev/internal/task/repository/repoerrors"
	wfmodels "github.com/kandev/kandev/internal/workflow/models"
	v1 "github.com/kandev/kandev/pkg/api/v1"
	"github.com/kandev/kandev/pkg/pluginsdk"
	"google.golang.org/grpc/codes"
	"google.golang.org/grpc/status"
)

// ── fakes for the narrow Host data API interfaces ───────────────────────

type fakeTaskDataSource struct {
	workspaces       []*taskmodels.Workspace
	tasksByWorkspace map[string][]*taskmodels.Task
	tasksByID        map[string]*taskmodels.Task
	repositories     map[string][]*taskmodels.Repository
	sessionsByTask   map[string][]*taskmodels.TaskSession
	executorRunning  map[string]*taskmodels.ExecutorRunning

	// gotIncludeArchived records the includeArchived flag of every
	// ListTasksByWorkspace call, in call order.
	gotIncludeArchived []bool

	// executorRunningCalls counts GetExecutorRunningBySessionID calls, so
	// tests can prove Sessions().List only resolves ACPSessionID for the
	// page it actually returns, not every session it fetched.
	executorRunningCalls int

	// listTasksByWorkspaceCalls counts ListTasksByWorkspace calls, so tests
	// can prove a workspace with more tasks than a single page issues
	// multiple calls instead of returning a truncated first page.
	listTasksByWorkspaceCalls int
}

func (f *fakeTaskDataSource) ListWorkspaces(context.Context) ([]*taskmodels.Workspace, error) {
	return f.workspaces, nil
}

// ListTasksByWorkspace mirrors the real repository's page/pageSize semantics
// (1-based page, offset = (page-1)*pageSize, total = the full unpaginated
// count) so tests can prove callers that need every task actually loop
// pagination to completion instead of assuming a single page has it all.
func (f *fakeTaskDataSource) ListTasksByWorkspace(
	_ context.Context, workspaceID, _, _, _ string, page, pageSize int, _ string, includeArchived, _, _, _ bool,
) ([]*taskmodels.Task, int, error) {
	f.gotIncludeArchived = append(f.gotIncludeArchived, includeArchived)
	f.listTasksByWorkspaceCalls++
	all := f.tasksByWorkspace[workspaceID]
	total := len(all)
	offset := (page - 1) * pageSize
	if offset < 0 || offset >= total {
		return []*taskmodels.Task{}, total, nil
	}
	end := offset + pageSize
	if end > total {
		end = total
	}
	return all[offset:end], total, nil
}

func (f *fakeTaskDataSource) GetTask(_ context.Context, id string) (*taskmodels.Task, error) {
	task, ok := f.tasksByID[id]
	if !ok {
		return nil, repoerrors.ErrTaskNotFound
	}
	return task, nil
}

func (f *fakeTaskDataSource) ListRepositories(_ context.Context, workspaceID string) ([]*taskmodels.Repository, error) {
	return f.repositories[workspaceID], nil
}

func (f *fakeTaskDataSource) ListTaskSessions(_ context.Context, taskID string) ([]*taskmodels.TaskSession, error) {
	return f.sessionsByTask[taskID], nil
}

func (f *fakeTaskDataSource) GetExecutorRunningBySessionID(_ context.Context, sessionID string) (*taskmodels.ExecutorRunning, error) {
	f.executorRunningCalls++
	running, ok := f.executorRunning[sessionID]
	if !ok {
		return nil, taskmodels.ErrExecutorRunningNotFound
	}
	return running, nil
}

type fakeWorkflowLister struct {
	workflows map[string][]*taskmodels.Workflow
}

func (f *fakeWorkflowLister) ListWorkflows(_ context.Context, workspaceID string, _ bool) ([]*taskmodels.Workflow, error) {
	return f.workflows[workspaceID], nil
}

type fakeWorkflowStepLister struct {
	steps map[string][]*wfmodels.WorkflowStep
}

func (f *fakeWorkflowStepLister) ListStepsByWorkflow(_ context.Context, workflowID string) ([]*wfmodels.WorkflowStep, error) {
	return f.steps[workflowID], nil
}

type fakeAgentProfileDataSource struct {
	resp *agentsettingsdto.ListAgentsResponse
}

func (f *fakeAgentProfileDataSource) ListAgents(context.Context) (*agentsettingsdto.ListAgentsResponse, error) {
	return f.resp, nil
}

type fakeSessionCodeStatsSource struct {
	calls      int
	lastFilter analyticsmodels.SessionCodeStatsFilter
	stats      []*analyticsmodels.SessionCodeStats
	err        error
}

func (f *fakeSessionCodeStatsSource) ListSessionCodeStats(
	_ context.Context, filter analyticsmodels.SessionCodeStatsFilter,
) ([]*analyticsmodels.SessionCodeStats, error) {
	f.calls++
	f.lastFilter = filter
	if f.err != nil {
		return nil, f.err
	}
	return f.stats, nil
}

type fakeMessageDataSource struct {
	calls      int
	lastFilter taskmodels.PluginMessageFilter
	messages   []*taskmodels.Message
	err        error
}

func (f *fakeMessageDataSource) ListMessagesForPlugin(
	_ context.Context, filter taskmodels.PluginMessageFilter,
) ([]*taskmodels.Message, error) {
	f.calls++
	f.lastFilter = filter
	if f.err != nil {
		return nil, f.err
	}
	return f.messages, nil
}

// fakeTaskWriter records CreateTask/UpdateTask inputs and returns canned
// tasks, standing in for the backendapp adapter over internal/task/service.
type fakeTaskWriter struct {
	lastCreate  TaskCreateInput
	lastUpdate  TaskUpdateInput
	createCalls int
	updateCalls int
	created     *taskmodels.Task
	updated     *taskmodels.Task
	createErr   error
	updateErr   error
}

func (f *fakeTaskWriter) CreateTask(_ context.Context, in TaskCreateInput) (*taskmodels.Task, error) {
	f.createCalls++
	f.lastCreate = in
	if f.createErr != nil {
		return nil, f.createErr
	}
	if f.created != nil {
		return f.created, nil
	}
	return &taskmodels.Task{ID: "task-created", WorkspaceID: in.WorkspaceID, WorkflowID: in.WorkflowID, Title: in.Title}, nil
}

func (f *fakeTaskWriter) UpdateTask(_ context.Context, in TaskUpdateInput) (*taskmodels.Task, error) {
	f.updateCalls++
	f.lastUpdate = in
	if f.updateErr != nil {
		return nil, f.updateErr
	}
	if f.updated != nil {
		return f.updated, nil
	}
	return &taskmodels.Task{ID: in.ID, Title: "updated"}, nil
}

// fakeMessenger records SendMessage calls, standing in for the backendapp
// adapter over the task service + orchestrator delivery path.
type fakeMessenger struct {
	calls       int
	lastTaskID  string
	lastSession string
	lastText    string
	lastSource  string
	result      PluginMessageResult
	err         error
}

func (f *fakeMessenger) SendMessage(_ context.Context, taskID, sessionID, text, source string) (PluginMessageResult, error) {
	f.calls++
	f.lastTaskID, f.lastSession, f.lastText, f.lastSource = taskID, sessionID, text, source
	if f.err != nil {
		return PluginMessageResult{}, f.err
	}
	if f.result == (PluginMessageResult{}) {
		return PluginMessageResult{SessionID: "session-x", Status: "sent"}, nil
	}
	return f.result, nil
}

// fakeTaskStarter records StartTask calls behind CreateTask's start_agent.
type fakeTaskStarter struct {
	calls  int
	lastID string
	err    error
}

func (f *fakeTaskStarter) StartTask(_ context.Context, taskID string) error {
	f.calls++
	f.lastID = taskID
	return f.err
}

// testDataHost bundles a pluginHost with every fake it was wired from, so
// tests can both drive Host calls and assert against the fakes' recorded
// state.
type testDataHost struct {
	host       *pluginHost
	tasks      *fakeTaskDataSource
	workflows  *fakeWorkflowLister
	steps      *fakeWorkflowStepLister
	profiles   *fakeAgentProfileDataSource
	codeStats  *fakeSessionCodeStatsSource
	messages   *fakeMessageDataSource
	utilAgents *fakeUtilityAgentSource
	utilRun    *fakeUtilityRunner
	taskWriter *fakeTaskWriter
	messenger  *fakeMessenger
	starter    *fakeTaskStarter
}

// newTestDataHost builds a fully-wired pluginHost (every Host data API
// dependency set, even if a given test's capabilities don't grant every
// resource) so each test only needs to vary caps.
func newTestDataHost(caps manifest.Capabilities) *testDataHost {
	d := &testDataHost{
		tasks:      &fakeTaskDataSource{},
		workflows:  &fakeWorkflowLister{},
		steps:      &fakeWorkflowStepLister{},
		profiles:   &fakeAgentProfileDataSource{resp: &agentsettingsdto.ListAgentsResponse{}},
		codeStats:  &fakeSessionCodeStatsSource{},
		messages:   &fakeMessageDataSource{},
		utilAgents: &fakeUtilityAgentSource{},
		utilRun:    &fakeUtilityRunner{text: "ok"},
		taskWriter: &fakeTaskWriter{},
		messenger:  &fakeMessenger{},
		starter:    &fakeTaskStarter{},
	}
	d.host = &pluginHost{
		pluginID:         "p1",
		capabilities:     caps,
		taskData:         d.tasks,
		workflows:        d.workflows,
		workflowSteps:    d.steps,
		agentProfiles:    d.profiles,
		sessionCodeStats: d.codeStats,
		messageData:      d.messages,
		taskWriter:       d.taskWriter,
		configs:          &fakeConfigReader{configs: map[string]any{utilityAgentConfigKey: "utility-agent-42"}},
		utilityDeps: func() (utilityAgentSource, utilityRunner) {
			return d.utilAgents, d.utilRun
		},
		writeDeps: func() (taskMessenger, taskStarter) {
			return d.messenger, d.starter
		},
	}
	return d
}

// ── capability gating: denied without api_read:<resource> ──────────────

func TestPluginHost_Tasks_DeniedWithoutCapability(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{})
	_, _, err := d.host.Tasks().List(context.Background(), pluginsdk.TaskFilter{}, pluginsdk.Page{})
	assertPermissionDenied(t, err, "api_read:tasks")

	_, err = d.host.Tasks().Get(context.Background(), "task-1")
	assertPermissionDenied(t, err, "api_read:tasks")
}

func TestPluginHost_Sessions_DeniedWithoutCapability(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{})
	_, _, err := d.host.Sessions().List(context.Background(), pluginsdk.SessionFilter{}, pluginsdk.Page{})
	assertPermissionDenied(t, err, "api_read:sessions")

	_, _, err = d.host.Sessions().CodeStats(context.Background(), pluginsdk.SessionFilter{}, pluginsdk.Page{})
	assertPermissionDenied(t, err, "api_read:sessions")
}

func TestPluginHost_Workspaces_DeniedWithoutCapability(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{})
	_, _, err := d.host.Workspaces().List(context.Background(), pluginsdk.Page{})
	assertPermissionDenied(t, err, "api_read:workspaces")
}

func TestPluginHost_Workflows_DeniedWithoutCapability(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{})
	_, _, err := d.host.Workflows().List(context.Background(), "ws-1", pluginsdk.Page{})
	assertPermissionDenied(t, err, "api_read:workflows")

	_, err = d.host.Workflows().ListSteps(context.Background(), "wf-1")
	assertPermissionDenied(t, err, "api_read:workflows")
}

func TestPluginHost_AgentProfiles_DeniedWithoutCapability(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{})
	_, _, err := d.host.AgentProfiles().List(context.Background(), pluginsdk.Page{})
	assertPermissionDenied(t, err, "api_read:agent_profiles")
}

func TestPluginHost_Repositories_DeniedWithoutCapability(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{})
	_, _, err := d.host.Repositories().List(context.Background(), "ws-1", pluginsdk.Page{})
	assertPermissionDenied(t, err, "api_read:repositories")
}

func TestPluginHost_Messages_DeniedWithoutCapability(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{})
	_, _, err := d.host.Messages().List(context.Background(), pluginsdk.MessageFilter{}, pluginsdk.Page{})
	assertPermissionDenied(t, err, "api_read:messages")
}

// ── capability gating: succeeds with api_read:<resource> ───────────────

func TestPluginHost_Tasks_SucceedsWithCapability(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"tasks"}})
	d.tasks.workspaces = []*taskmodels.Workspace{{ID: "ws-1"}}
	d.tasks.tasksByWorkspace = map[string][]*taskmodels.Task{
		"ws-1": {{ID: "task-1", WorkspaceID: "ws-1", WorkflowID: "wf-1", Title: "Task 1", State: v1.TaskStateTODO}},
	}
	d.tasks.tasksByID = map[string]*taskmodels.Task{"task-1": d.tasks.tasksByWorkspace["ws-1"][0]}

	tasks, info, err := d.host.Tasks().List(context.Background(), pluginsdk.TaskFilter{}, pluginsdk.Page{})
	if err != nil {
		t.Fatalf("List() unexpected error: %v", err)
	}
	if len(tasks) != 1 || tasks[0].ID != "task-1" {
		t.Fatalf("List() = %+v, want one task-1", tasks)
	}
	if info == nil || info.HasMore {
		t.Fatalf("PageInfo = %+v, want HasMore=false", info)
	}

	got, err := d.host.Tasks().Get(context.Background(), "task-1")
	if err != nil {
		t.Fatalf("Get() unexpected error: %v", err)
	}
	if got == nil || got.Title != "Task 1" {
		t.Fatalf("Get() = %+v, want Task 1", got)
	}

	// A missing task returns a gRPC NotFound error directly, matching what a
	// real plugin observes over the wire (grpcHostServer.GetTask forwards
	// this error as-is) rather than a (nil, nil) success that only the
	// in-process caller would ever see.
	got, err = d.host.Tasks().Get(context.Background(), "no-such-task")
	if got != nil {
		t.Fatalf("Get() for missing task = (%+v, %v), want (nil, NotFound)", got, err)
	}
	if status.Code(err) != codes.NotFound {
		t.Fatalf("Get() for missing task error = %v, want codes.NotFound", err)
	}
}

func TestPluginHost_Workspaces_SucceedsWithCapability(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"workspaces"}})
	d.tasks.workspaces = []*taskmodels.Workspace{{ID: "ws-1", Name: "Workspace 1", OwnerID: "user-1"}}

	workspaces, _, err := d.host.Workspaces().List(context.Background(), pluginsdk.Page{})
	if err != nil {
		t.Fatalf("List() unexpected error: %v", err)
	}
	if len(workspaces) != 1 || workspaces[0].Name != "Workspace 1" {
		t.Fatalf("List() = %+v, want one Workspace 1", workspaces)
	}
}

func TestPluginHost_Workflows_SucceedsWithCapability(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"workflows"}})
	d.workflows.workflows = map[string][]*taskmodels.Workflow{
		"ws-1": {{ID: "wf-1", WorkspaceID: "ws-1", Name: "Default"}},
	}
	d.steps.steps = map[string][]*wfmodels.WorkflowStep{
		"wf-1": {{ID: "step-1", WorkflowID: "wf-1", Name: "Todo", Position: 0, StageType: wfmodels.StageType("work")}},
	}

	workflows, _, err := d.host.Workflows().List(context.Background(), "ws-1", pluginsdk.Page{})
	if err != nil {
		t.Fatalf("List() unexpected error: %v", err)
	}
	if len(workflows) != 1 || workflows[0].ID != "wf-1" {
		t.Fatalf("List() = %+v, want one wf-1", workflows)
	}

	steps, err := d.host.Workflows().ListSteps(context.Background(), "wf-1")
	if err != nil {
		t.Fatalf("ListSteps() unexpected error: %v", err)
	}
	if len(steps) != 1 || steps[0].StageType != "work" {
		t.Fatalf("ListSteps() = %+v, want one step with StageType=work", steps)
	}
}

func TestPluginHost_AgentProfiles_SucceedsWithCapability(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"agent_profiles"}})
	d.profiles.resp = &agentsettingsdto.ListAgentsResponse{
		Agents: []agentsettingsdto.AgentDTO{
			{
				ID: "agent-1",
				Profiles: []agentsettingsdto.AgentProfileDTO{
					{ID: "profile-1", AgentID: "agent-1", Name: "Default", AgentDisplayName: "Claude", Model: "claude-x", Mode: "code"},
				},
			},
		},
	}

	profiles, _, err := d.host.AgentProfiles().List(context.Background(), pluginsdk.Page{})
	if err != nil {
		t.Fatalf("List() unexpected error: %v", err)
	}
	if len(profiles) != 1 || profiles[0].DisplayName != "Claude" || profiles[0].Model != "claude-x" {
		t.Fatalf("List() = %+v, want one Claude/claude-x profile", profiles)
	}
}

func TestPluginHost_Repositories_SucceedsWithCapability(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"repositories"}})
	branch := "main"
	d.tasks.repositories = map[string][]*taskmodels.Repository{
		"ws-1": {{ID: "repo-1", WorkspaceID: "ws-1", Name: "kandev", DefaultBranch: branch}},
	}

	repos, _, err := d.host.Repositories().List(context.Background(), "ws-1", pluginsdk.Page{})
	if err != nil {
		t.Fatalf("List() unexpected error: %v", err)
	}
	if len(repos) != 1 || repos[0].Name != "kandev" || repos[0].DefaultBranch == nil || *repos[0].DefaultBranch != "main" {
		t.Fatalf("List() = %+v, want one kandev repo on main", repos)
	}
}

// TestFetchTasksForWorkspaces_DoesNotTruncateBeyondPageSize proves a
// workspace with more tasks than a single taskFetchPageSize page is fully
// enumerated, not silently cut off at the first page.
func TestFetchTasksForWorkspaces_DoesNotTruncateBeyondPageSize(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"tasks"}})
	const total = taskFetchPageSize + 1
	tasks := make([]*taskmodels.Task, total)
	for i := 0; i < total; i++ {
		tasks[i] = &taskmodels.Task{ID: fmt.Sprintf("task-%04d", i), WorkspaceID: "ws-1"}
	}
	d.tasks.tasksByWorkspace = map[string][]*taskmodels.Task{"ws-1": tasks}

	got, err := d.host.fetchTasksForWorkspaces(context.Background(), []string{"ws-1"}, false, false)
	if err != nil {
		t.Fatalf("fetchTasksForWorkspaces() unexpected error: %v", err)
	}
	if len(got) != total {
		t.Fatalf("fetchTasksForWorkspaces() returned %d tasks, want %d (must not silently truncate at taskFetchPageSize)", len(got), total)
	}
	if d.tasks.listTasksByWorkspaceCalls < 2 {
		t.Fatalf("ListTasksByWorkspace called %d times, want >= 2 (a single %d-task page can't hold %d tasks)",
			d.tasks.listTasksByWorkspaceCalls, taskFetchPageSize, total)
	}
}

// TestPluginHost_Tasks_List_DoesNotTruncateBeyondPageSize is the
// end-to-end version of the above: Tasks().List's HasMore/pagination must
// stay accurate even when the underlying per-workspace fetch spans more than
// one taskFetchPageSize page.
func TestPluginHost_Tasks_List_DoesNotTruncateBeyondPageSize(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"tasks"}})
	const total = taskFetchPageSize + 1
	tasks := make([]*taskmodels.Task, total)
	for i := 0; i < total; i++ {
		tasks[i] = &taskmodels.Task{ID: fmt.Sprintf("task-%04d", i), WorkspaceID: "ws-1"}
	}
	d.tasks.tasksByWorkspace = map[string][]*taskmodels.Task{"ws-1": tasks}

	// maxPageLimit caps a single RPC page at 200 items, so walk every page
	// via cursor and count the total returned across the whole read.
	seen := 0
	cursor := ""
	for {
		page, info, err := d.host.Tasks().List(context.Background(), pluginsdk.TaskFilter{WorkspaceIDs: []string{"ws-1"}}, pluginsdk.Page{Limit: 200, Cursor: cursor})
		if err != nil {
			t.Fatalf("List() unexpected error: %v", err)
		}
		seen += len(page)
		if info == nil || !info.HasMore {
			break
		}
		cursor = info.NextCursor
	}
	if seen != total {
		t.Fatalf("Tasks().List() across all pages returned %d tasks, want %d (must not silently truncate at taskFetchPageSize)", seen, total)
	}
}

// ── Session DTO mapping: acp_session_id sourcing ────────────────────────

func TestPluginHost_Sessions_ACPSessionIDFromMetadata(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"sessions"}})
	d.tasks.workspaces = []*taskmodels.Workspace{{ID: "ws-1"}}
	d.tasks.tasksByWorkspace = map[string][]*taskmodels.Task{"ws-1": {{ID: "task-1", WorkspaceID: "ws-1"}}}
	d.tasks.sessionsByTask = map[string][]*taskmodels.TaskSession{
		"task-1": {{
			ID:        "session-1",
			TaskID:    "task-1",
			State:     taskmodels.TaskSessionStateRunning,
			StartedAt: time.Now(),
			Metadata:  map[string]any{"acp": map[string]any{"session_id": "acp-from-metadata"}},
		}},
	}
	// Also seed an executors_running row so the test proves metadata wins
	// over the fallback when both are present.
	d.tasks.executorRunning = map[string]*taskmodels.ExecutorRunning{
		"session-1": {SessionID: "session-1", ResumeToken: "acp-from-fallback"},
	}

	sessions, _, err := d.host.Sessions().List(context.Background(), pluginsdk.SessionFilter{}, pluginsdk.Page{})
	if err != nil {
		t.Fatalf("List() unexpected error: %v", err)
	}
	if len(sessions) != 1 || sessions[0].ACPSessionID != "acp-from-metadata" {
		t.Fatalf("List() = %+v, want ACPSessionID=acp-from-metadata", sessions)
	}
}

func TestPluginHost_Sessions_ACPSessionIDFallsBackToExecutorRunning(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"sessions"}})
	d.tasks.workspaces = []*taskmodels.Workspace{{ID: "ws-1"}}
	d.tasks.tasksByWorkspace = map[string][]*taskmodels.Task{"ws-1": {{ID: "task-1", WorkspaceID: "ws-1"}}}
	d.tasks.sessionsByTask = map[string][]*taskmodels.TaskSession{
		"task-1": {{ID: "session-1", TaskID: "task-1", State: taskmodels.TaskSessionStateRunning, StartedAt: time.Now()}},
	}
	d.tasks.executorRunning = map[string]*taskmodels.ExecutorRunning{
		"session-1": {SessionID: "session-1", ResumeToken: "acp-from-fallback"},
	}

	sessions, _, err := d.host.Sessions().List(context.Background(), pluginsdk.SessionFilter{}, pluginsdk.Page{})
	if err != nil {
		t.Fatalf("List() unexpected error: %v", err)
	}
	if len(sessions) != 1 || sessions[0].ACPSessionID != "acp-from-fallback" {
		t.Fatalf("List() = %+v, want ACPSessionID=acp-from-fallback", sessions)
	}
}

func TestPluginHost_Sessions_ACPSessionIDEmptyWhenNeitherSourceHasIt(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"sessions"}})
	d.tasks.workspaces = []*taskmodels.Workspace{{ID: "ws-1"}}
	d.tasks.tasksByWorkspace = map[string][]*taskmodels.Task{"ws-1": {{ID: "task-1", WorkspaceID: "ws-1"}}}
	d.tasks.sessionsByTask = map[string][]*taskmodels.TaskSession{
		"task-1": {{ID: "session-1", TaskID: "task-1", State: taskmodels.TaskSessionStateRunning, StartedAt: time.Now()}},
	}
	// No executorRunning entry for session-1: GetExecutorRunningBySessionID
	// returns ErrExecutorRunningNotFound, which must be swallowed.

	sessions, _, err := d.host.Sessions().List(context.Background(), pluginsdk.SessionFilter{}, pluginsdk.Page{})
	if err != nil {
		t.Fatalf("List() unexpected error: %v", err)
	}
	if len(sessions) != 1 || sessions[0].ACPSessionID != "" {
		t.Fatalf("List() = %+v, want empty ACPSessionID", sessions)
	}
}

// TestPluginHost_Sessions_PaginatesBeforeResolvingACPSessionID proves
// Sessions().List resolves ACPSessionID (a per-item DB call via
// resolveACPSessionID's GetExecutorRunningBySessionID fallback) only for the
// page it returns, not for every session it fetched — pagination must happen
// on the raw, already-sorted sessions BEFORE the per-item DTO conversion, or
// a Page{Limit: 2} read over 5 sessions would issue 5 lookups instead of 2.
func TestPluginHost_Sessions_PaginatesBeforeResolvingACPSessionID(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"sessions"}})
	d.tasks.workspaces = []*taskmodels.Workspace{{ID: "ws-1"}}
	d.tasks.tasksByWorkspace = map[string][]*taskmodels.Task{"ws-1": {{ID: "task-1", WorkspaceID: "ws-1"}}}

	const total = 5
	sessions := make([]*taskmodels.TaskSession, total)
	base := time.Date(2026, 1, 10, 0, 0, 0, 0, time.UTC)
	for i := 0; i < total; i++ {
		// No Metadata: resolveACPSessionID always falls through to
		// GetExecutorRunningBySessionID for every one of these sessions.
		sessions[i] = &taskmodels.TaskSession{
			ID:        fmt.Sprintf("session-%d", i),
			TaskID:    "task-1",
			State:     taskmodels.TaskSessionStateRunning,
			StartedAt: base.Add(time.Duration(total-i) * time.Hour), // session-0 newest
		}
	}
	d.tasks.sessionsByTask = map[string][]*taskmodels.TaskSession{"task-1": sessions}

	const limit = 2
	page, info, err := d.host.Sessions().List(context.Background(), pluginsdk.SessionFilter{}, pluginsdk.Page{Limit: limit})
	if err != nil {
		t.Fatalf("List() unexpected error: %v", err)
	}
	if len(page) != limit {
		t.Fatalf("List() returned %d sessions, want %d", len(page), limit)
	}
	if info == nil || !info.HasMore {
		t.Fatalf("PageInfo = %+v, want HasMore=true", info)
	}
	if d.tasks.executorRunningCalls != limit {
		t.Fatalf("GetExecutorRunningBySessionID called %d times, want %d (only the returned page, not all %d fetched sessions)",
			d.tasks.executorRunningCalls, limit, total)
	}
}

// TestPluginHost_SessionsCodeStats_DelegatesToAnalyticsService proves
// Sessions().CodeStats reaches the injected analytics service with the
// filter translated 1:1, and maps its results back unchanged (ADR 0043(b):
// SessionCodeStats is a stable, computed shape returned as-is).
func TestPluginHost_SessionsCodeStats_DelegatesToAnalyticsService(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"sessions"}})
	d.codeStats.stats = []*analyticsmodels.SessionCodeStats{
		{SessionID: "session-1", LinesAddedCommitted: 10, LinesDeletedCommitted: 2, LinesAddedPeakPending: 5, LinesDeletedPeakPending: 1},
	}

	filter := pluginsdk.SessionFilter{TaskIDs: []string{"task-1"}, WorkspaceIDs: []string{"ws-1"}, States: []string{"RUNNING"}}
	stats, info, err := d.host.Sessions().CodeStats(context.Background(), filter, pluginsdk.Page{Limit: 10})
	if err != nil {
		t.Fatalf("CodeStats() unexpected error: %v", err)
	}
	if d.codeStats.calls != 1 {
		t.Fatalf("analytics service called %d times, want 1", d.codeStats.calls)
	}
	if len(d.codeStats.lastFilter.TaskIDs) != 1 || d.codeStats.lastFilter.TaskIDs[0] != "task-1" {
		t.Errorf("filter.TaskIDs = %v, want [task-1]", d.codeStats.lastFilter.TaskIDs)
	}
	if len(d.codeStats.lastFilter.WorkspaceIDs) != 1 || d.codeStats.lastFilter.WorkspaceIDs[0] != "ws-1" {
		t.Errorf("filter.WorkspaceIDs = %v, want [ws-1]", d.codeStats.lastFilter.WorkspaceIDs)
	}
	if len(d.codeStats.lastFilter.States) != 1 || d.codeStats.lastFilter.States[0] != "RUNNING" {
		t.Errorf("filter.States = %v, want [RUNNING]", d.codeStats.lastFilter.States)
	}
	// Limit is requested as limit+1 (the HasMore probe row) — see
	// sessionReader.CodeStats' doc comment.
	if d.codeStats.lastFilter.Limit != 11 {
		t.Errorf("filter.Limit = %d, want 11 (requested 10 + 1 probe row)", d.codeStats.lastFilter.Limit)
	}
	if len(stats) != 1 || stats[0].SessionID != "session-1" || stats[0].LinesAddedCommitted != 10 {
		t.Fatalf("CodeStats() = %+v, want session-1 passed through unchanged", stats)
	}
	if info == nil || info.HasMore {
		t.Fatalf("PageInfo = %+v, want HasMore=false", info)
	}
}

// TestPluginHost_SessionsCodeStats_HasMoreWhenExtraProbeRowReturned proves
// the limit+1 probe-row trick correctly reports HasMore and trims the extra
// row before returning results.
func TestPluginHost_SessionsCodeStats_HasMoreWhenExtraProbeRowReturned(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"sessions"}})
	d.codeStats.stats = []*analyticsmodels.SessionCodeStats{
		{SessionID: "session-1"}, {SessionID: "session-2"},
	}

	stats, info, err := d.host.Sessions().CodeStats(context.Background(), pluginsdk.SessionFilter{}, pluginsdk.Page{Limit: 1})
	if err != nil {
		t.Fatalf("CodeStats() unexpected error: %v", err)
	}
	if len(stats) != 1 || stats[0].SessionID != "session-1" {
		t.Fatalf("CodeStats() = %+v, want exactly one trimmed result", stats)
	}
	if info == nil || !info.HasMore || info.NextCursor != "1" {
		t.Fatalf("PageInfo = %+v, want HasMore=true NextCursor=1", info)
	}
}

func TestPluginHost_SessionsCodeStats_PropagatesAnalyticsError(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"sessions"}})
	wantErr := errors.New("boom")
	d.codeStats.err = wantErr

	_, _, err := d.host.Sessions().CodeStats(context.Background(), pluginsdk.SessionFilter{}, pluginsdk.Page{})
	if !errors.Is(err, wantErr) {
		t.Fatalf("CodeStats() error = %v, want %v", err, wantErr)
	}
}

// TestPluginHost_Sessions_TaskIDsScopedToWorkspaceIDs proves SessionFilter's
// TaskIDs and WorkspaceIDs are ANDed together: a task id that resolves to a
// task outside the requested workspaces must not leak its sessions, even
// though the id itself was explicitly requested.
func TestPluginHost_Sessions_TaskIDsScopedToWorkspaceIDs(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"sessions"}})
	task1 := &taskmodels.Task{ID: "task-1", WorkspaceID: "ws-1"}
	task2 := &taskmodels.Task{ID: "task-2", WorkspaceID: "ws-2"}
	d.tasks.tasksByID = map[string]*taskmodels.Task{"task-1": task1, "task-2": task2}
	d.tasks.sessionsByTask = map[string][]*taskmodels.TaskSession{
		"task-1": {{ID: "session-1", TaskID: "task-1", State: taskmodels.TaskSessionStateRunning, StartedAt: time.Now()}},
		"task-2": {{ID: "session-2", TaskID: "task-2", State: taskmodels.TaskSessionStateRunning, StartedAt: time.Now()}},
	}

	filter := pluginsdk.SessionFilter{TaskIDs: []string{"task-1", "task-2"}, WorkspaceIDs: []string{"ws-1"}}
	sessions, _, err := d.host.Sessions().List(context.Background(), filter, pluginsdk.Page{})
	if err != nil {
		t.Fatalf("List() unexpected error: %v", err)
	}
	if len(sessions) != 1 || sessions[0].ID != "session-1" {
		t.Fatalf("List() = %+v, want only session-1 (task-2 is outside the requested workspace)", sessions)
	}
}

// ── Stable sort: deterministic pagination across equal timestamps ───────

// TestSortTasksNewestFirst_TiesBrokenByID proves sortTasksNewestFirst orders
// equal-CreatedAt tasks deterministically by ID, not by whatever order
// sort.Slice's unstable algorithm happens to leave them in — an offset-based
// paginated read of tasks sharing a CreatedAt second (a plausible seed/batch
// scenario) must return the same order on every call, or successive pages
// can skip or duplicate a task.
func TestSortTasksNewestFirst_TiesBrokenByID(t *testing.T) {
	same := time.Date(2026, 1, 1, 0, 0, 0, 0, time.UTC)
	tasks := []*taskmodels.Task{
		{ID: "c", CreatedAt: same},
		{ID: "a", CreatedAt: same},
		{ID: "b", CreatedAt: same},
	}

	sortTasksNewestFirst(tasks)

	got := []string{tasks[0].ID, tasks[1].ID, tasks[2].ID}
	want := []string{"a", "b", "c"}
	for i := range want {
		if got[i] != want[i] {
			t.Fatalf("sortTasksNewestFirst() order = %v, want %v (equal CreatedAt must tie-break by ID)", got, want)
		}
	}
}

// TestSortSessionsNewestFirst_TiesBrokenByID mirrors
// TestSortTasksNewestFirst_TiesBrokenByID for sessions (StartedAt instead of
// CreatedAt).
func TestSortSessionsNewestFirst_TiesBrokenByID(t *testing.T) {
	same := time.Date(2026, 1, 1, 0, 0, 0, 0, time.UTC)
	sessions := []*taskmodels.TaskSession{
		{ID: "c", StartedAt: same},
		{ID: "a", StartedAt: same},
		{ID: "b", StartedAt: same},
	}

	sortSessionsNewestFirst(sessions)

	got := []string{sessions[0].ID, sessions[1].ID, sessions[2].ID}
	want := []string{"a", "b", "c"}
	for i := range want {
		if got[i] != want[i] {
			t.Fatalf("sortSessionsNewestFirst() order = %v, want %v (equal StartedAt must tie-break by ID)", got, want)
		}
	}
}

// ── DTO mapping ──────────────────────────────────────────────────────────

func TestTaskModelToDTO_MapsFields(t *testing.T) {
	created := time.Date(2026, 1, 2, 3, 4, 5, 0, time.UTC)
	task := &taskmodels.Task{
		ID:          "task-1",
		WorkspaceID: "ws-1",
		WorkflowID:  "wf-1",
		Title:       "Fix bug",
		Description: "details",
		State:       v1.TaskStateInProgress,
		Priority:    "high",
		Origin:      "agent_created",
		CreatedAt:   created,
		UpdatedAt:   created,
		ParentID:    "parent-1",
		Identifier:  "KAN-1",
		IsEphemeral: false,
		Repositories: []*taskmodels.TaskRepository{
			{ID: "tr-1", RepositoryID: "repo-1", BaseBranch: "main", Position: 0},
		},
		Metadata: map[string]any{"k": "v"},
	}

	dto := taskModelToDTO(task)

	if dto.ID != "task-1" || dto.State != "IN_PROGRESS" || dto.CreatedBy != "agent_created" {
		t.Fatalf("taskModelToDTO() = %+v, unexpected core fields", dto)
	}
	if dto.CreatedAt != created.Format(time.RFC3339) {
		t.Errorf("CreatedAt = %q, want RFC3339 %q", dto.CreatedAt, created.Format(time.RFC3339))
	}
	if dto.ParentID == nil || *dto.ParentID != "parent-1" {
		t.Errorf("ParentID = %v, want parent-1", dto.ParentID)
	}
	if len(dto.Repositories) != 1 || dto.Repositories[0].RepositoryID != "repo-1" {
		t.Errorf("Repositories = %+v, want one repo-1", dto.Repositories)
	}
	if dto.Metadata["k"] != "v" {
		t.Errorf("Metadata = %+v, want k=v", dto.Metadata)
	}
}

func TestTaskModelToDTO_EmptyParentIDIsNil(t *testing.T) {
	dto := taskModelToDTO(&taskmodels.Task{ID: "task-1"})
	if dto.ParentID != nil {
		t.Fatalf("ParentID = %v, want nil for a root task", dto.ParentID)
	}
}

// ── Archived-task visibility ────────────────────────────────────────────

// Sessions().List must fetch tasks WITH archived included (an archived
// task's sessions are still real sessions), while Tasks().List keeps
// archived tasks out of plugin-visible work items.
func TestPluginHost_ArchivedTaskFetchFlags(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"sessions", "tasks"}})
	d.tasks.workspaces = []*taskmodels.Workspace{{ID: "ws-1"}}
	d.tasks.tasksByWorkspace = map[string][]*taskmodels.Task{"ws-1": {{ID: "task-1", WorkspaceID: "ws-1"}}}

	if _, _, err := d.host.Sessions().List(context.Background(), pluginsdk.SessionFilter{}, pluginsdk.Page{}); err != nil {
		t.Fatalf("Sessions().List unexpected error: %v", err)
	}
	if _, _, err := d.host.Tasks().List(context.Background(), pluginsdk.TaskFilter{}, pluginsdk.Page{}); err != nil {
		t.Fatalf("Tasks().List unexpected error: %v", err)
	}

	want := []bool{true, false} // sessions read first (archived in), tasks read second (archived out)
	if len(d.tasks.gotIncludeArchived) != 2 ||
		d.tasks.gotIncludeArchived[0] != want[0] || d.tasks.gotIncludeArchived[1] != want[1] {
		t.Fatalf("includeArchived per call = %v, want %v", d.tasks.gotIncludeArchived, want)
	}
}

// ── Messages reader ─────────────────────────────────────────────────────

func TestPluginHost_Messages_SucceedsAndStripsSystemContent(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"messages"}})
	created := time.Date(2026, 7, 20, 9, 30, 0, 0, time.UTC)
	d.messages.messages = []*taskmodels.Message{
		{
			ID:            "m1",
			TaskSessionID: "s1",
			TaskID:        "t1",
			TurnID:        "turn1",
			AuthorType:    taskmodels.MessageAuthorUser,
			Content:       "<kandev-system>secret prompt</kandev-system>Please summarize yesterday.",
			Type:          taskmodels.MessageTypeMessage,
			CreatedAt:     created,
		},
		{
			ID:            "m2",
			TaskSessionID: "s1",
			TaskID:        "t1",
			TurnID:        "turn1",
			AuthorType:    taskmodels.MessageAuthorAgent,
			Content:       "Here is the summary.",
			Type:          "", // empty type must default to "message"
			CreatedAt:     created.Add(time.Minute),
		},
	}

	msgs, info, err := d.host.Messages().List(context.Background(), pluginsdk.MessageFilter{
		SessionIDs: []string{"s1"},
	}, pluginsdk.Page{Limit: 50})
	if err != nil {
		t.Fatalf("List() unexpected error: %v", err)
	}
	if len(msgs) != 2 {
		t.Fatalf("List() len = %d, want 2", len(msgs))
	}
	if msgs[0].Content != "Please summarize yesterday." {
		t.Fatalf("content[0] = %q, want system block stripped", msgs[0].Content)
	}
	if msgs[0].AuthorType != "user" || msgs[1].AuthorType != "agent" {
		t.Fatalf("author types = %q/%q, want user/agent", msgs[0].AuthorType, msgs[1].AuthorType)
	}
	if msgs[1].Type != "message" {
		t.Fatalf("type[1] = %q, want defaulted to message", msgs[1].Type)
	}
	if msgs[0].SessionID != "s1" || msgs[0].TaskID != "t1" || msgs[0].TurnID != "turn1" {
		t.Fatalf("ids = %+v, want session/task/turn set", msgs[0])
	}
	if msgs[0].CreatedAt != "2026-07-20T09:30:00Z" {
		t.Fatalf("created_at = %q, want RFC3339", msgs[0].CreatedAt)
	}
	if info == nil || info.HasMore {
		t.Fatalf("PageInfo = %+v, want HasMore=false", info)
	}
}

func TestPluginHost_Messages_PassesFilterAndTimeRange(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"messages"}})
	since := "2026-07-20T00:00:00Z"
	until := "2026-07-21T00:00:00Z"

	_, _, err := d.host.Messages().List(context.Background(), pluginsdk.MessageFilter{
		SessionIDs: []string{"s1", "s2"},
		TaskIDs:    []string{"t1"},
		Types:      []string{"message", "content"},
		Since:      &since,
		Until:      &until,
	}, pluginsdk.Page{Limit: 10})
	if err != nil {
		t.Fatalf("List() unexpected error: %v", err)
	}
	got := d.messages.lastFilter
	if len(got.SessionIDs) != 2 || len(got.TaskIDs) != 1 || len(got.Types) != 2 {
		t.Fatalf("filter passthrough = %+v", got)
	}
	if got.Since == nil || !got.Since.Equal(time.Date(2026, 7, 20, 0, 0, 0, 0, time.UTC)) {
		t.Fatalf("since = %v, want parsed 2026-07-20", got.Since)
	}
	if got.Until == nil || !got.Until.Equal(time.Date(2026, 7, 21, 0, 0, 0, 0, time.UTC)) {
		t.Fatalf("until = %v, want parsed 2026-07-21", got.Until)
	}
	// Reader requests one extra row past the page limit to derive HasMore.
	if got.Limit != 11 {
		t.Fatalf("data-source Limit = %d, want page-limit+1 (11)", got.Limit)
	}
}

func TestPluginHost_Messages_PaginatesWithHasMore(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"messages"}})
	base := time.Date(2026, 7, 20, 9, 0, 0, 0, time.UTC)
	// Data source returns limit+1 rows (3) for a page limit of 2 → HasMore.
	for i := 0; i < 3; i++ {
		d.messages.messages = append(d.messages.messages, &taskmodels.Message{
			ID:         fmt.Sprintf("m%d", i),
			AuthorType: taskmodels.MessageAuthorUser,
			Content:    "hi",
			CreatedAt:  base.Add(time.Duration(i) * time.Minute),
		})
	}

	msgs, info, err := d.host.Messages().List(context.Background(), pluginsdk.MessageFilter{}, pluginsdk.Page{Limit: 2})
	if err != nil {
		t.Fatalf("List() unexpected error: %v", err)
	}
	if len(msgs) != 2 {
		t.Fatalf("page len = %d, want 2 (trimmed from 3)", len(msgs))
	}
	if info == nil || !info.HasMore || info.NextCursor != "2" {
		t.Fatalf("PageInfo = %+v, want HasMore=true NextCursor=2", info)
	}
}

func TestPluginHost_Messages_TooManyFilterValuesIsInvalidArgument(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"messages"}})
	sessionIDs := make([]string, maxMessageFilterValues+1)
	for i := range sessionIDs {
		sessionIDs[i] = fmt.Sprintf("s%d", i)
	}
	_, _, err := d.host.Messages().List(context.Background(), pluginsdk.MessageFilter{SessionIDs: sessionIDs}, pluginsdk.Page{})
	st, ok := status.FromError(err)
	if !ok || st.Code() != codes.InvalidArgument {
		t.Fatalf("err = %v, want InvalidArgument", err)
	}
	if d.messages.calls != 0 {
		t.Fatalf("data source called %d times, want 0 (rejected before query)", d.messages.calls)
	}
}

func TestPluginHost_Messages_InvalidTimeIsInvalidArgument(t *testing.T) {
	d := newTestDataHost(manifest.Capabilities{APIRead: []string{"messages"}})
	bad := "not-a-timestamp"
	_, _, err := d.host.Messages().List(context.Background(), pluginsdk.MessageFilter{Since: &bad}, pluginsdk.Page{})
	st, ok := status.FromError(err)
	if !ok || st.Code() != codes.InvalidArgument {
		t.Fatalf("err = %v, want InvalidArgument", err)
	}
	if d.messages.calls != 0 {
		t.Fatalf("data source called %d times, want 0 (rejected before query)", d.messages.calls)
	}
}