feat(state-store): ListSteps, dashboard run detail reads it; drop tasks.stage

docs/prds/execution-tree-state-machine.md: step_executions (0017) is now the
durable per-step record; the dashboard's run detail (GET /api/tasks/{id}/
attempts) still folded stage_start/stage_finish events into a rail by hand,
and tasks.stage duplicated what StartStep/FinishStep already record properly.

Added the missing read:
- ListSteps(execution_id, attempt) -- attempt 0 lists every attempt of the
  execution, oldest first, so the rail renders every attempt from one call.
  New storecontract.StepReader (separate from StepRecorder: this is a
  dashboard/API/dispatch action per the PRD's wire contract, never
  task-scoped-grant callable), postgresdb.ListStepExecutions, staterpc
  client/server/values.go plumbing, wired through state_store.go's Deps and
  archieui/compose.go's Server.Steps the same optional-capability pattern
  every other dashboard read uses.
- internal/webui/api_tasks_attempts.go: stageViewsFromSteps replaces
  foldStages/finishStage/stageOutcome. Attempt bounds, event_count and which
  attempt numbers exist stay sourced from events (an attempt can carry
  events -- agent calls -- with no step), only the stage list itself moves.
  The wire JSON contract is unchanged, so no frontend change was needed;
  `task ui` confirms ui/dist still matches a fresh build.
- A running step past the task's own current attempt reads interrupted, not
  running: RecoverStale (archie-core, still open per #1214) is what corrects
  a crashed run's stale "running" row, and until it does this is the honest
  read rather than trusting a row that may be stale.

Dropped tasks.stage and every reader, no fallback:
- Migration 0018 drops the column; tasks.sql's INSERT/UPDATE/RETURNING lists
  and the sqlc-generated code no longer carry it.
- workflow.Task.Stage is gone; workflow.Run's in-flight "which stage is this"
  bookkeeping moves to TaskContext.Stage (execution-time only, never
  persisted -- park/finish's event tagging is its only reader).
- pb.Task.stage (wire field 10) is `[deprecated = true]`, not deleted (never
  reuse a field number); staterpc's taskProto/taskValue no longer map it.
- taskactions.Task.Stage and the retry event's "previous_stage" data field
  are gone: the operator-action package had no reader for the column beyond
  echoing it.
- The chat /tasks command's per-task "workflow/stage" display drops to
  workflow-only (ChatTaskSummary.Stage removed, workflowStage() deleted):
  it was a genuine reader with no replacement in scope for this change.

Read the archie-web-ui skill and the state-store/daemon docs/development
pages first, per instruction. Stayed out of cancel/stop code throughout
(Pi's concurrent CancelExecution work); the only shared files touched
(workflow.go, daemon-adjacent SQL) got the smallest possible diff.
This commit is contained in:
sam committed 2026-09-27 06:57:47 +10:00
1 parent ed12d4de78
commit 5f375f8a14
35 files changed
+1910 -1259

No files matched your search

-1
View File
@@ -166,7 +166,6 @@ func (a chatTaskListerAdapter) ListChatTasks(ctx context.Context, identity strin
Title: task.Title,
Status: task.Status,
Workflow: task.Workflow,
Stage: task.Stage,
PRNumber: task.PRNumber,
Attempt: task.Attempt,
ParkReason: task.ParkReason,
+3
View File
@@ -352,6 +352,9 @@ func (b *boot) stateStoreDeps(grants *staterpc.TaskGrants) staterpc.Deps {
if ec, ok := b.st.(storecontract.ExecutionCanceller); ok {
deps.Canceller = ec
}
if sr, ok := b.st.(storecontract.StepReader); ok {
deps.StepReader = sr
}
b.accessDeps(&deps)
if css, ok := b.st.(storecontract.ConfigSnapshotStore); ok {
deps.ConfigSnapshots = css
+3
View File
@@ -90,6 +90,9 @@ func compose(d deps) *webui.Server {
if snapshots, ok := d.Store.(storecontract.ConfigSnapshotStore); ok {
srv.ConfigSource = webui.RemoteConfigView(snapshots)
}
if steps, ok := d.Store.(storecontract.StepReader); ok {
srv.Steps = steps
}
// Channel lifecycle follows the same split as the config view: the process
// hosting the channels publishes, this one reads. Withholding it left
// /api/channels answering with nothing while channels were running
+25 -5
View File
@@ -20,6 +20,7 @@ import (
"github.com/samcharles93/archie-core/internal/domain/messaging"
"github.com/samcharles93/archie-core/internal/domain/storecontract"
"github.com/samcharles93/archie-core/internal/domain/taskactions"
"github.com/samcharles93/archie-core/internal/domain/workflow/task"
"github.com/samcharles93/archie-core/internal/events"
"github.com/samcharles93/archie-core/internal/gateway"
"github.com/samcharles93/archie-core/internal/infrastructure/gatewayrpc"
@@ -102,10 +103,15 @@ func serveGRPC(t *testing.T, register func(grpc.ServiceRegistrar)) (target strin
func TestUIServesDashboardAgainstRemoteContracts(t *testing.T) {
st := pgstore.Open(t)
defer st.Close()
seeded, err := st.EnqueueChatTask(t.Context(), "acme", "widget", "remote summary", "body", "implement", "")
if err != nil {
if _, err := st.EnqueueChatTask(t.Context(), "acme", "widget", "remote summary", "body", "implement", ""); err != nil {
t.Fatalf("seed task: %v", err)
}
// Claimed (not just enqueued): StartStep guards on the execution's own
// current attempt, which a merely-queued task does not have yet.
seeded, err := st.ClaimNext(t.Context())
if err != nil || seeded == nil {
t.Fatalf("claim seeded task: %+v %v", seeded, err)
}
// The daemon publishes the configuration page's projection; this process
// only renders it (archie-core-ymut). The forge host is published too: it
// is what the run-detail reads resolve their repository and pull-request
@@ -129,9 +135,23 @@ func TestUIServesDashboardAgainstRemoteContracts(t *testing.T) {
// worker's change capture -- attributed to one attempt, so the assertions
// further down read them back through the real client rather than a fake.
seedAt := time.Now().UTC()
prepareStepID, _, err := st.StartStep(t.Context(), task.StepStart{ExecutionID: seeded.ID, Attempt: 1, Kind: task.StepKindStage, Name: "prepare"})
if err != nil {
t.Fatalf("seed prepare step: %v", err)
}
if _, err := st.FinishStep(t.Context(), task.StepFinish{
StepID: prepareStepID, ExecutionID: seeded.ID, From: taskstate.StepRunning, To: taskstate.StepSucceeded,
}); err != nil {
t.Fatalf("finish prepare step: %v", err)
}
// Terminal, not left running: the attempt rail reads the task's own
// record for "is this still happening" (inFlight), and a run still
// marked running would report the attempt "running" regardless of its
// finished step, rather than the "ok" a completed run's own stages earn.
if err := st.Transition(t.Context(), seeded.ID, taskstate.Running, taskstate.Completed, "done"); err != nil {
t.Fatalf("complete seeded task: %v", err)
}
for _, e := range []events.Event{
{Kind: events.KindStageStart, TaskID: seeded.ID, Attempt: 1, Stage: "prepare", At: seedAt},
{Kind: events.KindStageFinish, TaskID: seeded.ID, Attempt: 1, Stage: "prepare", At: seedAt.Add(time.Second), Data: map[string]any{"duration_ms": 1000}},
{Kind: events.KindConfigCaptured, TaskID: seeded.ID, Attempt: 1, At: seedAt, Data: map[string]any{
"schema": events.ConfigCapturedSchema, "document": map[string]any{"bot_user": "archie"},
}},
@@ -150,7 +170,7 @@ func TestUIServesDashboardAgainstRemoteContracts(t *testing.T) {
stateTarget, stopState := serveGRPC(t, func(r grpc.ServiceRegistrar) {
eda := pgstore.EDA(t, nil)
staterpc.RegisterServer(r, staterpc.Deps{Tasks: st, Captures: eda, BindingDispatcher: eda, ConfigSnapshots: st, Log: slog.New(slog.DiscardHandler)})
staterpc.RegisterServer(r, staterpc.Deps{Tasks: st, Captures: eda, BindingDispatcher: eda, ConfigSnapshots: st, StepReader: st, Log: slog.New(slog.DiscardHandler)})
})
defer stopState()
chat := &fakeChat{sessions: []gateway.SessionContext{
File diff suppressed because it is too large. Load diff
@@ -43,6 +43,7 @@ const (
StateStoreService_StartStep_FullMethodName = "/state.v1.StateStoreService/StartStep"
StateStoreService_FinishStep_FullMethodName = "/state.v1.StateStoreService/FinishStep"
StateStoreService_CancelExecution_FullMethodName = "/state.v1.StateStoreService/CancelExecution"
StateStoreService_ListSteps_FullMethodName = "/state.v1.StateStoreService/ListSteps"
StateStoreService_Requeue_FullMethodName = "/state.v1.StateStoreService/Requeue"
StateStoreService_ParkTask_FullMethodName = "/state.v1.StateStoreService/ParkTask"
StateStoreService_RecoverStale_FullMethodName = "/state.v1.StateStoreService/RecoverStale"
@@ -160,6 +161,9 @@ type StateStoreServiceClient interface {
// task-scoped grant): a cancel is a dashboard, API and dispatch action
// decided by the Authorizer, not a worker call.
CancelExecution(ctx context.Context, in *CancelExecutionRequest, opts ...grpc.CallOption) (*CancelExecutionResponse, error)
// ListSteps is a dashboard/API/dispatch action (Authorizer: read on the
// run), not task-scoped-grant callable like StartStep/FinishStep.
ListSteps(ctx context.Context, in *ListStepsRequest, opts ...grpc.CallOption) (*ListStepsResponse, error)
Requeue(ctx context.Context, in *RequeueRequest, opts ...grpc.CallOption) (*RequeueResponse, error)
ParkTask(ctx context.Context, in *ParkTaskRequest, opts ...grpc.CallOption) (*ParkTaskResponse, error)
RecoverStale(ctx context.Context, in *RecoverStaleRequest, opts ...grpc.CallOption) (*RecoverStaleResponse, error)
@@ -533,6 +537,16 @@ func (c *stateStoreServiceClient) CancelExecution(ctx context.Context, in *Cance
return out, nil
}
func (c *stateStoreServiceClient) ListSteps(ctx context.Context, in *ListStepsRequest, opts ...grpc.CallOption) (*ListStepsResponse, error) {
cOpts := append([]grpc.CallOption{grpc.StaticMethod()}, opts...)
out := new(ListStepsResponse)
err := c.cc.Invoke(ctx, StateStoreService_ListSteps_FullMethodName, in, out, cOpts...)
if err != nil {
return nil, err
}
return out, nil
}
func (c *stateStoreServiceClient) Requeue(ctx context.Context, in *RequeueRequest, opts ...grpc.CallOption) (*RequeueResponse, error) {
cOpts := append([]grpc.CallOption{grpc.StaticMethod()}, opts...)
out := new(RequeueResponse)
@@ -1278,6 +1292,9 @@ type StateStoreServiceServer interface {
// task-scoped grant): a cancel is a dashboard, API and dispatch action
// decided by the Authorizer, not a worker call.
CancelExecution(context.Context, *CancelExecutionRequest) (*CancelExecutionResponse, error)
// ListSteps is a dashboard/API/dispatch action (Authorizer: read on the
// run), not task-scoped-grant callable like StartStep/FinishStep.
ListSteps(context.Context, *ListStepsRequest) (*ListStepsResponse, error)
Requeue(context.Context, *RequeueRequest) (*RequeueResponse, error)
ParkTask(context.Context, *ParkTaskRequest) (*ParkTaskResponse, error)
RecoverStale(context.Context, *RecoverStaleRequest) (*RecoverStaleResponse, error)
@@ -1483,6 +1500,9 @@ func (UnimplementedStateStoreServiceServer) FinishStep(context.Context, *FinishS
func (UnimplementedStateStoreServiceServer) CancelExecution(context.Context, *CancelExecutionRequest) (*CancelExecutionResponse, error) {
return nil, status.Error(codes.Unimplemented, "method CancelExecution not implemented")
}
func (UnimplementedStateStoreServiceServer) ListSteps(context.Context, *ListStepsRequest) (*ListStepsResponse, error) {
return nil, status.Error(codes.Unimplemented, "method ListSteps not implemented")
}
func (UnimplementedStateStoreServiceServer) Requeue(context.Context, *RequeueRequest) (*RequeueResponse, error) {
return nil, status.Error(codes.Unimplemented, "method Requeue not implemented")
}
@@ -2137,6 +2157,24 @@ func _StateStoreService_CancelExecution_Handler(srv interface{}, ctx context.Con
return interceptor(ctx, in, info, handler)
}
func _StateStoreService_ListSteps_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
in := new(ListStepsRequest)
if err := dec(in); err != nil {
return nil, err
}
if interceptor == nil {
return srv.(StateStoreServiceServer).ListSteps(ctx, in)
}
info := &grpc.UnaryServerInfo{
Server: srv,
FullMethod: StateStoreService_ListSteps_FullMethodName,
}
handler := func(ctx context.Context, req interface{}) (interface{}, error) {
return srv.(StateStoreServiceServer).ListSteps(ctx, req.(*ListStepsRequest))
}
return interceptor(ctx, in, info, handler)
}
func _StateStoreService_Requeue_Handler(srv interface{}, ctx context.Context, dec func(interface{}) error, interceptor grpc.UnaryServerInterceptor) (interface{}, error) {
in := new(RequeueRequest)
if err := dec(in); err != nil {
@@ -3425,6 +3463,10 @@ var StateStoreService_ServiceDesc = grpc.ServiceDesc{
MethodName: "CancelExecution",
Handler: _StateStoreService_CancelExecution_Handler,
},
{
MethodName: "ListSteps",
Handler: _StateStoreService_ListSteps_Handler,
},
{
MethodName: "Requeue",
Handler: _StateStoreService_Requeue_Handler,
@@ -90,6 +90,15 @@ type ExecutionCanceller interface {
CancelExecution(ctx context.Context, taskID int64, reason, to string) ([]int64, error)
}
// StepReader lists one execution's recorded steps -- the dashboard run
// detail's read (docs/prds/execution-tree-state-machine.md). Separate from
// StepRecorder: this is a dashboard/API/dispatch action (Authorizer: read on
// the run), never task-scoped-grant callable the way StartStep/FinishStep
// are.
type StepReader interface {
ListSteps(ctx context.Context, executionID int64, attempt int) ([]task.StepExecution, error)
}
// TaskArchiver removes one terminal task's local record with an optimistic
// status guard. It is separate from the already broad lifecycle contract so
// consumers that only run tasks do not acquire an operator-only capability.
+5 -5
View File
@@ -20,10 +20,10 @@ var (
// Task is the state needed to authorize and apply an operator action.
type Task struct {
ID int64
Owner, Repo, Identity, Status, Stage, ParkReason string
IssueNumber, RetryCount int
ForgeBacked bool
ID int64
Owner, Repo, Identity, Status, ParkReason string
IssueNumber, RetryCount int
ForgeBacked bool
// Attempt is the run the operator acted on. It is stamped onto the event
// this action records so an intervention is attributable to the run it
// changed course -- without it a retry or a stop is indistinguishable
@@ -256,7 +256,7 @@ func (s Service) applyRetry(ctx context.Context, task *Task, actor Actor, o outc
}
err := s.Store.RetryTask(ctx, task.ID, "parked", "")
o.event.Kind, o.event.Detail = events.KindTaskRetried, actor.describe("retried")
o.event.Data = map[string]any{"retry_count": task.RetryCount + 1, "previous_stage": task.Stage, "previous_reason": task.ParkReason}
o.event.Data = map[string]any{"retry_count": task.RetryCount + 1, "previous_reason": task.ParkReason}
return o, err
}
+4 -3
View File
@@ -161,8 +161,9 @@ func captureTaskContext(t *testing.T, trees Trees, store Store) *TaskContext {
return &TaskContext{
Task: &Task{
ID: 11, Attempt: 2, Owner: "acme", Repo: "widget", IssueNumber: 7,
Stage: "commit-push", Branch: "feat/7-widget",
Branch: "feat/7-widget",
},
Stage: "commit-push",
Repo: config.Repo{Owner: "acme", Name: "widget", Base: "main"},
Trees: trees,
Store: store,
@@ -278,7 +279,7 @@ func TestStageCommitCapturesTheChangedFiles(t *testing.T) {
trees := &capturingTrees{commitAllChanged: true, stats: sampleStats()}
store := &recordingStore{}
tc := captureTaskContext(t, trees, store)
tc.Task.Stage = "commit-repro"
tc.Stage = "commit-repro"
if err := StageCommit("commit-repro", func(*TaskContext) string { return "msg" }).Run(context.Background(), tc); err != nil {
t.Fatalf("StageCommit.Run() = %v, want nil", err)
@@ -430,7 +431,7 @@ func TestOpenPRCapturesThePullRequestNumber(t *testing.T) {
forgeClient := &fakeForge{}
tc := captureTaskContext(t, trees, store)
tc.Forge = forgeClient
tc.Task.Stage = "open-pr"
tc.Stage = "open-pr"
tc.Task.Title = "widget fails"
if err := OpenPR(context.Background(), tc, "body"); err != nil {
+1 -1
View File
@@ -71,7 +71,7 @@ func (tc *TaskContext) captureChanges(ctx context.Context, after string) task.Ch
// Durable, never bus-only: the worktree this was measured from is gone by
// the time anyone reads it, so a capture that only reached the live feed
// would be lost with the process.
if err := tc.EmitDurable(ctx, events.KindChangesCaptured, tc.Task.Stage, "", changeCaptureData(tc, after, stats)); err != nil {
if err := tc.EmitDurable(ctx, events.KindChangesCaptured, tc.Stage, "", changeCaptureData(tc, after, stats)); err != nil {
tc.Log.Warn("change capture not persisted", "captured_after", after, "err", err)
}
return stats
+24 -1
View File
@@ -1,6 +1,10 @@
package task
import "github.com/samcharles93/archie-core/internal/taskstate"
import (
"time"
"github.com/samcharles93/archie-core/internal/taskstate"
)
// The step-execution vocabulary of docs/prds/execution-tree-state-machine.md:
// one stage run, or one agent or workflow.call step an execution's tree
@@ -57,3 +61,22 @@ type StepFinish struct {
// the agent runtime records its own calls' usage.
TokensUsed int64
}
// StepExecution is one recorded step: ListSteps' read, the dashboard run
// detail's authoritative source (docs/prds/execution-tree-state-machine.md;
// replaces folding stage_start/stage_finish events, which tasks.stage used to
// back). ParentID is 0 for a stage at the tree's root.
type StepExecution struct {
ID int64
ExecutionID int64
Attempt int
ParentID int64
Depth int
Kind string
Name string
Status taskstate.StepStatus
Detail string
TokensUsed int64
StartedAt time.Time
FinishedAt time.Time
}
-1
View File
@@ -53,7 +53,6 @@ type Task struct {
WorkflowDefinitionVersion int64 `json:"workflow_definition_version"`
WorkflowDefinitionDigest string `json:"workflow_definition_digest"`
WorkflowDefinitionYAML string `json:"workflow_definition_yaml"`
Stage string `json:"stage"`
Branch string `json:"branch"`
Plan string `json:"plan"`
Notes string `json:"notes"`
+9 -4
View File
@@ -109,6 +109,11 @@ type TaskContext struct {
// Dir/Branch are set by the prepare step.
Dir string
Branch string
// Stage is the name of the stage currently running, kept here (not
// persisted on Task -- docs/prds/execution-tree-state-machine.md drops
// tasks.stage in favour of the durable StepExecution record StartStep
// writes) purely so park/finish can tag the event they emit with it.
Stage string
// BuildSummary is the builder agent's finish summary -- the PR body.
BuildSummary string
// BuildNoChanges is set when the builder returned StatusPassed but
@@ -302,7 +307,7 @@ func Run(ctx context.Context, wf Workflow, tc *TaskContext) {
log := tc.Log.With("workflow", wf.Name, "repo", tc.Repo.FullName(), "issue", t.IssueNumber)
for _, stage := range wf.Stages {
t.Stage = stage.Name
tc.Stage = stage.Name
_ = tc.Store.Update(ctx, t)
// Every stage is a StepExecution (docs/prds/execution-tree-state-machine.md):
// the store records pending -> running and writes the stage_start event
@@ -407,7 +412,7 @@ func finish(ctx context.Context, tc *TaskContext, log *slog.Logger) {
}
_ = tc.Store.Update(ctx, t)
_ = tc.Store.Transition(ctx, t.ID, StatusRunning, tc.Outcome.Status, tc.Outcome.Detail)
tc.Emit(events.KindOutcome, t.Stage, tc.Outcome.Detail, map[string]any{"status": tc.Outcome.Status})
tc.Emit(events.KindOutcome, tc.Stage, tc.Outcome.Detail, map[string]any{"status": tc.Outcome.Status})
log.Info("workflow finished", "status", tc.Outcome.Status)
}
@@ -416,12 +421,12 @@ func park(ctx context.Context, tc *TaskContext, reason string) {
t.ParkReason = reason
_ = tc.Store.Update(ctx, t)
_ = tc.Store.Transition(ctx, t.ID, StatusRunning, StatusParked, reason)
tc.Emit(events.KindParked, t.Stage, reason, nil)
tc.Emit(events.KindParked, tc.Stage, reason, nil)
// The run's own log is the surface an operator downloads to answer "why did
// this park?", and the event above lands on the timeline, which the log
// never carries. This is the one place a park's reason is final, so it is
// the one place that records it -- the stage names itself in the reason.
tc.Log.Error("task parked", "stage", t.Stage, "reason", reason)
tc.Log.Error("task parked", "stage", tc.Stage, "reason", reason)
}
func clip(s string, n int) string {
+2 -18
View File
@@ -712,8 +712,8 @@ func formatTasks(tasks []ChatTaskSummary, now time.Time) string {
}
fmt.Fprintf(&b, "%s #%d %s%s", icon, t.ID, t.Title, reportLineBreak)
fmt.Fprintf(&b, " %s", label)
if stage := workflowStage(t.Workflow, t.Stage); stage != "" {
fmt.Fprintf(&b, " · %s", stage)
if t.Workflow != "" {
fmt.Fprintf(&b, " · %s", t.Workflow)
}
if t.Attempt > 1 {
fmt.Fprintf(&b, " · attempt %d", t.Attempt)
@@ -729,22 +729,6 @@ func formatTasks(tasks []ChatTaskSummary, now time.Time) string {
return strings.TrimSpace(b.String())
}
// workflowStage joins a task's workflow and stage for display, degrading
// gracefully when either is unknown (e.g. a task that hasn't started its
// first stage yet).
func workflowStage(workflow, stage string) string {
switch {
case workflow == "" && stage == "":
return ""
case stage == "":
return workflow
case workflow == "":
return stage
default:
return workflow + "/" + stage
}
}
type taskStateDisplay struct {
icon string
label string
+6 -6
View File
@@ -250,11 +250,11 @@ func TestFormatTasks(t *testing.T) {
want string
}{
{
name: "running task shows workflow/stage and age",
name: "running task shows workflow and age",
tasks: []ChatTaskSummary{
{ID: 42, Title: "Fix login bug", Status: taskstate.Running, Workflow: "tdd", Stage: "implement", UpdatedAt: now.Add(-3 * time.Minute)},
{ID: 42, Title: "Fix login bug", Status: taskstate.Running, Workflow: "tdd", UpdatedAt: now.Add(-3 * time.Minute)},
},
want: "🗂 Archie tasks\n\n▶ #42 Fix login bug \n Running · tdd/implement · updated 3m ago",
want: "🗂 Archie tasks\n\n▶ #42 Fix login bug \n Running · tdd · updated 3m ago",
},
{
// The criterion this command was built for: a bare "Running: 1"
@@ -262,10 +262,10 @@ func TestFormatTasks(t *testing.T) {
// failure loop, and each task's own age is what tells them apart.
name: "a stuck running task does not read like one making progress",
tasks: []ChatTaskSummary{
{ID: 5, Title: "Wedged", Status: taskstate.Running, Workflow: "tdd", Stage: "implement", UpdatedAt: now.Add(-47 * time.Minute)},
{ID: 6, Title: "Working", Status: taskstate.Running, Workflow: "tdd", Stage: "implement", UpdatedAt: now.Add(-20 * time.Second)},
{ID: 5, Title: "Wedged", Status: taskstate.Running, Workflow: "tdd", UpdatedAt: now.Add(-47 * time.Minute)},
{ID: 6, Title: "Working", Status: taskstate.Running, Workflow: "tdd", UpdatedAt: now.Add(-20 * time.Second)},
},
want: "🗂 Archie tasks\n\n▶ #5 Wedged \n Running · tdd/implement · updated 47m ago\n\n▶ #6 Working \n Running · tdd/implement · updated just now",
want: "🗂 Archie tasks\n\n▶ #5 Wedged \n Running · tdd · updated 47m ago\n\n▶ #6 Working \n Running · tdd · updated just now",
},
{
name: "parked task shows its reason",
-1
View File
@@ -35,7 +35,6 @@ type ChatTaskSummary struct {
Title string `json:"title"`
Status string `json:"status"`
Workflow string `json:"workflow,omitempty"`
Stage string `json:"stage,omitempty"`
PRNumber int `json:"pr_number,omitempty"`
// Attempt is the task's current retry attempt. 0 or 1 means it has never
// been retried; the /tasks command only surfaces this once it is
@@ -0,0 +1,13 @@
-- tasks.stage was the one mutable column recording "which stage is this task
-- on", overwritten as workflow.Run advanced and carrying no history: a failed
-- attempt could not say which step failed or what ran before it without
-- replaying the event stream (docs/prds/execution-tree-state-machine.md,
-- "Problem"). step_executions (0017) is now the durable per-step record the
-- dashboard's run detail reads (ListSteps), so this column has no remaining
-- reader.
-- +goose Up
ALTER TABLE tasks DROP COLUMN stage;
-- +goose Down
ALTER TABLE tasks ADD COLUMN stage text NOT NULL DEFAULT '';
@@ -24,7 +24,7 @@ func (q *Queries) CallStatusDetail(ctx context.Context, taskID int64) (string, e
const enqueueCallTask = `-- name: EnqueueCallTask :one
WITH caller AS (
SELECT t.id, t.owner, t.repo, t.issue_number, t.title, t.body, t.labels, t.status, t.workflow, t.stage, t.branch, t.plan, t.notes, t.pr_number, t.tokens_used, t.iterations, t.attempt, t.park_reason, t.watch_comment_id, t.park_class, t.remediation_rounds, t.retry_count, t.source, t.identity, t.binding_id, t.binding_version, t.review_payload, t.workflow_definition_version, t.workflow_definition_digest, t.workflow_definition_yaml, t.created_at, t.updated_at, t.review_cursor, t.inputs, t.org_id, t.workspace_id, t.call_parent_task_id, t.call_depth FROM tasks t WHERE t.id = $1 FOR UPDATE
SELECT t.id, t.owner, t.repo, t.issue_number, t.title, t.body, t.labels, t.status, t.workflow, t.branch, t.plan, t.notes, t.pr_number, t.tokens_used, t.iterations, t.attempt, t.park_reason, t.watch_comment_id, t.park_class, t.remediation_rounds, t.retry_count, t.source, t.identity, t.binding_id, t.binding_version, t.review_payload, t.workflow_definition_version, t.workflow_definition_digest, t.workflow_definition_yaml, t.created_at, t.updated_at, t.review_cursor, t.inputs, t.org_id, t.workspace_id, t.call_parent_task_id, t.call_depth FROM tasks t WHERE t.id = $1 FOR UPDATE
)
INSERT INTO tasks (owner, repo, issue_number, title, body, labels, workflow, source, identity, org_id, workspace_id, inputs, call_parent_task_id, call_depth)
SELECT
@@ -41,7 +41,7 @@ SELECT
caller.id, caller.call_depth + 1
FROM caller
WHERE caller.status = 'running' AND caller.call_depth + 1 <= $7::int
RETURNING id, owner, repo, issue_number, title, body, labels, status, workflow, stage, branch, plan, notes, pr_number, tokens_used, iterations, attempt, park_reason, watch_comment_id, park_class, remediation_rounds, retry_count, source, identity, binding_id, binding_version, review_payload, workflow_definition_version, workflow_definition_digest, workflow_definition_yaml, created_at, updated_at, review_cursor, inputs, org_id, workspace_id, call_parent_task_id, call_depth
RETURNING id, owner, repo, issue_number, title, body, labels, status, workflow, branch, plan, notes, pr_number, tokens_used, iterations, attempt, park_reason, watch_comment_id, park_class, remediation_rounds, retry_count, source, identity, binding_id, binding_version, review_payload, workflow_definition_version, workflow_definition_digest, workflow_definition_yaml, created_at, updated_at, review_cursor, inputs, org_id, workspace_id, call_parent_task_id, call_depth
`
type EnqueueCallTaskParams struct {
@@ -84,7 +84,6 @@ func (q *Queries) EnqueueCallTask(ctx context.Context, arg EnqueueCallTaskParams
&i.Labels,
&i.Status,
&i.Workflow,
&i.Stage,
&i.Branch,
&i.Plan,
&i.Notes,
@@ -348,7 +348,6 @@ type Task struct {
Labels string
Status string
Workflow string
Stage string
Branch string
Plan string
Notes string
@@ -157,6 +157,69 @@ func (q *Queries) InterruptAttemptSteps(ctx context.Context, arg InterruptAttemp
return items, nil
}
const listStepExecutions = `-- name: ListStepExecutions :many
SELECT id, execution_id, attempt, parent_id, depth, kind, name, status, detail, tokens_used, started_at, finished_at
FROM step_executions
WHERE execution_id = $1::bigint AND ($2::bigint = 0 OR attempt = $2::bigint)
ORDER BY attempt, id
`
type ListStepExecutionsParams struct {
ExecutionID int64
Attempt int64
}
type ListStepExecutionsRow struct {
ID int64
ExecutionID int64
Attempt int64
ParentID pgtype.Int8
Depth int32
Kind string
Name string
Status string
Detail string
TokensUsed int64
StartedAt pgtype.Timestamptz
FinishedAt pgtype.Timestamptz
}
// attempt = 0 lists every attempt of the execution, oldest first; the index
// (execution_id, attempt, id) makes both this and the single-attempt form a
// straight index scan.
func (q *Queries) ListStepExecutions(ctx context.Context, arg ListStepExecutionsParams) ([]ListStepExecutionsRow, error) {
rows, err := q.db.Query(ctx, listStepExecutions, arg.ExecutionID, arg.Attempt)
if err != nil {
return nil, err
}
defer rows.Close()
var items []ListStepExecutionsRow
for rows.Next() {
var i ListStepExecutionsRow
if err := rows.Scan(
&i.ID,
&i.ExecutionID,
&i.Attempt,
&i.ParentID,
&i.Depth,
&i.Kind,
&i.Name,
&i.Status,
&i.Detail,
&i.TokensUsed,
&i.StartedAt,
&i.FinishedAt,
); err != nil {
return nil, err
}
items = append(items, i)
}
if err := rows.Err(); err != nil {
return nil, err
}
return items, nil
}
const lockExecutionForStep = `-- name: LockExecutionForStep :one
SELECT id, status, org_id, workspace_id, owner, repo, issue_number, workflow, attempt
@@ -28,7 +28,7 @@ func (q *Queries) ArchiveTaskDelete(ctx context.Context, arg ArchiveTaskDeletePa
}
const beginRemediationTask = `-- name: BeginRemediationTask :execrows
UPDATE tasks SET status = 'queued', workflow = 'remediate', stage = '', park_reason = '', park_class = 'needs_human', review_payload = $2, updated_at = now()
UPDATE tasks SET status = 'queued', workflow = 'remediate', park_reason = '', park_class = 'needs_human', review_payload = $2, updated_at = now()
WHERE id = $1 AND status = 'pr_open'
`
@@ -48,7 +48,7 @@ func (q *Queries) BeginRemediationTask(ctx context.Context, arg BeginRemediation
const claimByIssue = `-- name: ClaimByIssue :one
UPDATE tasks SET status = 'running', attempt = attempt + 1, updated_at = now()
WHERE owner = $1 AND repo = $2 AND issue_number = $3 AND status = 'queued'
RETURNING id, owner, repo, issue_number, title, body, labels, status, workflow, stage, branch, plan, notes, pr_number, tokens_used, iterations, attempt, park_reason, watch_comment_id, park_class, remediation_rounds, retry_count, source, identity, binding_id, binding_version, review_payload, workflow_definition_version, workflow_definition_digest, workflow_definition_yaml, created_at, updated_at, review_cursor, inputs, org_id, workspace_id, call_parent_task_id, call_depth
RETURNING id, owner, repo, issue_number, title, body, labels, status, workflow, branch, plan, notes, pr_number, tokens_used, iterations, attempt, park_reason, watch_comment_id, park_class, remediation_rounds, retry_count, source, identity, binding_id, binding_version, review_payload, workflow_definition_version, workflow_definition_digest, workflow_definition_yaml, created_at, updated_at, review_cursor, inputs, org_id, workspace_id, call_parent_task_id, call_depth
`
type ClaimByIssueParams struct {
@@ -70,7 +70,6 @@ func (q *Queries) ClaimByIssue(ctx context.Context, arg ClaimByIssueParams) (Tas
&i.Labels,
&i.Status,
&i.Workflow,
&i.Stage,
&i.Branch,
&i.Plan,
&i.Notes,
@@ -113,7 +112,7 @@ WHERE id = (
FOR UPDATE SKIP LOCKED
LIMIT 1
)
RETURNING id, owner, repo, issue_number, title, body, labels, status, workflow, stage, branch, plan, notes, pr_number, tokens_used, iterations, attempt, park_reason, watch_comment_id, park_class, remediation_rounds, retry_count, source, identity, binding_id, binding_version, review_payload, workflow_definition_version, workflow_definition_digest, workflow_definition_yaml, created_at, updated_at, review_cursor, inputs, org_id, workspace_id, call_parent_task_id, call_depth
RETURNING id, owner, repo, issue_number, title, body, labels, status, workflow, branch, plan, notes, pr_number, tokens_used, iterations, attempt, park_reason, watch_comment_id, park_class, remediation_rounds, retry_count, source, identity, binding_id, binding_version, review_payload, workflow_definition_version, workflow_definition_digest, workflow_definition_yaml, created_at, updated_at, review_cursor, inputs, org_id, workspace_id, call_parent_task_id, call_depth
`
// FOR UPDATE SKIP LOCKED replaces the SQLite single-writer assumption: two
@@ -132,7 +131,6 @@ func (q *Queries) ClaimNextTask(ctx context.Context) (Task, error) {
&i.Labels,
&i.Status,
&i.Workflow,
&i.Stage,
&i.Branch,
&i.Plan,
&i.Notes,
@@ -270,7 +268,7 @@ VALUES (
'default'
)
)
RETURNING id, owner, repo, issue_number, title, body, labels, status, workflow, stage, branch, plan, notes, pr_number, tokens_used, iterations, attempt, park_reason, watch_comment_id, park_class, remediation_rounds, retry_count, source, identity, binding_id, binding_version, review_payload, workflow_definition_version, workflow_definition_digest, workflow_definition_yaml, created_at, updated_at, review_cursor, inputs, org_id, workspace_id, call_parent_task_id, call_depth
RETURNING id, owner, repo, issue_number, title, body, labels, status, workflow, branch, plan, notes, pr_number, tokens_used, iterations, attempt, park_reason, watch_comment_id, park_class, remediation_rounds, retry_count, source, identity, binding_id, binding_version, review_payload, workflow_definition_version, workflow_definition_digest, workflow_definition_yaml, created_at, updated_at, review_cursor, inputs, org_id, workspace_id, call_parent_task_id, call_depth
`
type InsertChatTaskParams struct {
@@ -308,7 +306,6 @@ func (q *Queries) InsertChatTask(ctx context.Context, arg InsertChatTaskParams)
&i.Labels,
&i.Status,
&i.Workflow,
&i.Stage,
&i.Branch,
&i.Plan,
&i.Notes,
@@ -415,7 +412,7 @@ func (q *Queries) ListOpenPRs(ctx context.Context) ([]ListOpenPRsRow, error) {
}
const listTaskSummaries = `-- name: ListTaskSummaries :many
SELECT id, owner, repo, issue_number, title, status, workflow, stage,
SELECT id, owner, repo, issue_number, title, status, workflow,
pr_number, tokens_used, iterations, attempt, park_reason, retry_count,
created_at, updated_at, plan, source, identity, binding_id, binding_version
FROM tasks ORDER BY updated_at DESC LIMIT $1
@@ -429,7 +426,6 @@ type ListTaskSummariesRow struct {
Title string
Status string
Workflow string
Stage string
PrNumber int64
TokensUsed int64
Iterations int64
@@ -465,7 +461,6 @@ func (q *Queries) ListTaskSummaries(ctx context.Context, limit int32) ([]ListTas
&i.Title,
&i.Status,
&i.Workflow,
&i.Stage,
&i.PrNumber,
&i.TokensUsed,
&i.Iterations,
@@ -545,7 +540,7 @@ func (q *Queries) RecoverStaleTasks(ctx context.Context) (int64, error) {
const requeueTask = `-- name: RequeueTask :execrows
UPDATE tasks SET status = 'queued',
workflow = CASE WHEN $1::text = '' THEN workflow ELSE $1::text END,
stage = '', park_reason = '', park_class = 'needs_human', updated_at = now()
park_reason = '', park_class = 'needs_human', updated_at = now()
WHERE id = $2 AND status = $3
`
@@ -566,7 +561,7 @@ func (q *Queries) RequeueTask(ctx context.Context, arg RequeueTaskParams) (int64
const retryTask = `-- name: RetryTask :execrows
UPDATE tasks SET status = 'queued', retry_count = retry_count + 1,
workflow = CASE WHEN $1::text = '' THEN workflow ELSE $1::text END,
stage = '', park_reason = '', park_class = 'needs_human', updated_at = now()
park_reason = '', park_class = 'needs_human', updated_at = now()
WHERE id = $2 AND status = $3
`
@@ -625,7 +620,7 @@ func (q *Queries) StampTaskBinding(ctx context.Context, arg StampTaskBindingPara
}
const taskByID = `-- name: TaskByID :one
SELECT id, owner, repo, issue_number, title, body, labels, status, workflow, stage, branch, plan, notes, pr_number, tokens_used, iterations, attempt, park_reason, watch_comment_id, park_class, remediation_rounds, retry_count, source, identity, binding_id, binding_version, review_payload, workflow_definition_version, workflow_definition_digest, workflow_definition_yaml, created_at, updated_at, review_cursor, inputs, org_id, workspace_id, call_parent_task_id, call_depth FROM tasks WHERE id = $1
SELECT id, owner, repo, issue_number, title, body, labels, status, workflow, branch, plan, notes, pr_number, tokens_used, iterations, attempt, park_reason, watch_comment_id, park_class, remediation_rounds, retry_count, source, identity, binding_id, binding_version, review_payload, workflow_definition_version, workflow_definition_digest, workflow_definition_yaml, created_at, updated_at, review_cursor, inputs, org_id, workspace_id, call_parent_task_id, call_depth FROM tasks WHERE id = $1
`
func (q *Queries) TaskByID(ctx context.Context, id int64) (Task, error) {
@@ -641,7 +636,6 @@ func (q *Queries) TaskByID(ctx context.Context, id int64) (Task, error) {
&i.Labels,
&i.Status,
&i.Workflow,
&i.Stage,
&i.Branch,
&i.Plan,
&i.Notes,
@@ -675,7 +669,7 @@ func (q *Queries) TaskByID(ctx context.Context, id int64) (Task, error) {
}
const taskByIssue = `-- name: TaskByIssue :one
SELECT id, owner, repo, issue_number, title, body, labels, status, workflow, stage, branch, plan, notes, pr_number, tokens_used, iterations, attempt, park_reason, watch_comment_id, park_class, remediation_rounds, retry_count, source, identity, binding_id, binding_version, review_payload, workflow_definition_version, workflow_definition_digest, workflow_definition_yaml, created_at, updated_at, review_cursor, inputs, org_id, workspace_id, call_parent_task_id, call_depth FROM tasks WHERE owner = $1 AND repo = $2 AND issue_number = $3
SELECT id, owner, repo, issue_number, title, body, labels, status, workflow, branch, plan, notes, pr_number, tokens_used, iterations, attempt, park_reason, watch_comment_id, park_class, remediation_rounds, retry_count, source, identity, binding_id, binding_version, review_payload, workflow_definition_version, workflow_definition_digest, workflow_definition_yaml, created_at, updated_at, review_cursor, inputs, org_id, workspace_id, call_parent_task_id, call_depth FROM tasks WHERE owner = $1 AND repo = $2 AND issue_number = $3
`
type TaskByIssueParams struct {
@@ -697,7 +691,6 @@ func (q *Queries) TaskByIssue(ctx context.Context, arg TaskByIssueParams) (Task,
&i.Labels,
&i.Status,
&i.Workflow,
&i.Stage,
&i.Branch,
&i.Plan,
&i.Notes,
@@ -731,7 +724,7 @@ func (q *Queries) TaskByIssue(ctx context.Context, arg TaskByIssueParams) (Task,
}
const taskByPR = `-- name: TaskByPR :one
SELECT id, owner, repo, issue_number, title, body, labels, status, workflow, stage, branch, plan, notes, pr_number, tokens_used, iterations, attempt, park_reason, watch_comment_id, park_class, remediation_rounds, retry_count, source, identity, binding_id, binding_version, review_payload, workflow_definition_version, workflow_definition_digest, workflow_definition_yaml, created_at, updated_at, review_cursor, inputs, org_id, workspace_id, call_parent_task_id, call_depth FROM tasks WHERE owner = $1 AND repo = $2 AND pr_number = $3 AND status = $4
SELECT id, owner, repo, issue_number, title, body, labels, status, workflow, branch, plan, notes, pr_number, tokens_used, iterations, attempt, park_reason, watch_comment_id, park_class, remediation_rounds, retry_count, source, identity, binding_id, binding_version, review_payload, workflow_definition_version, workflow_definition_digest, workflow_definition_yaml, created_at, updated_at, review_cursor, inputs, org_id, workspace_id, call_parent_task_id, call_depth FROM tasks WHERE owner = $1 AND repo = $2 AND pr_number = $3 AND status = $4
`
type TaskByPRParams struct {
@@ -762,7 +755,6 @@ func (q *Queries) TaskByPR(ctx context.Context, arg TaskByPRParams) (Task, error
&i.Labels,
&i.Status,
&i.Workflow,
&i.Stage,
&i.Branch,
&i.Plan,
&i.Notes,
@@ -847,11 +839,11 @@ func (q *Queries) UpdateReviewPayloadTask(ctx context.Context, arg UpdateReviewP
}
const updateTask = `-- name: UpdateTask :exec
UPDATE tasks SET workflow = $2, stage = $3, branch = $4, plan = $5, notes = $6,
pr_number = $7, tokens_used = $8, iterations = $9, park_reason = $10,
watch_comment_id = $11, retry_count = $12, remediation_rounds = $13,
review_payload = $14, workflow_definition_version = $15,
workflow_definition_digest = $16, workflow_definition_yaml = $17,
UPDATE tasks SET workflow = $2, branch = $3, plan = $4, notes = $5,
pr_number = $6, tokens_used = $7, iterations = $8, park_reason = $9,
watch_comment_id = $10, retry_count = $11, remediation_rounds = $12,
review_payload = $13, workflow_definition_version = $14,
workflow_definition_digest = $15, workflow_definition_yaml = $16,
updated_at = now()
WHERE id = $1
`
@@ -859,7 +851,6 @@ WHERE id = $1
type UpdateTaskParams struct {
ID int64
Workflow string
Stage string
Branch string
Plan string
Notes string
@@ -880,7 +871,6 @@ func (q *Queries) UpdateTask(ctx context.Context, arg UpdateTaskParams) error {
_, err := q.db.Exec(ctx, updateTask,
arg.ID,
arg.Workflow,
arg.Stage,
arg.Branch,
arg.Plan,
arg.Notes,
@@ -61,3 +61,12 @@ UPDATE step_executions
SET status = 'interrupted', finished_at = now()
WHERE execution_id = $1 AND attempt = $2 AND status = 'running'
RETURNING id, name;
-- name: ListStepExecutions :many
-- attempt = 0 lists every attempt of the execution, oldest first; the index
-- (execution_id, attempt, id) makes both this and the single-attempt form a
-- straight index scan.
SELECT id, execution_id, attempt, parent_id, depth, kind, name, status, detail, tokens_used, started_at, finished_at
FROM step_executions
WHERE execution_id = @execution_id::bigint AND (@attempt::bigint = 0 OR attempt = @attempt::bigint)
ORDER BY attempt, id;
@@ -29,7 +29,7 @@ SELECT * FROM tasks WHERE owner = $1 AND repo = $2 AND pr_number = $3 AND status
-- The dashboard's list. This projection is deliberately narrow: Plan gates
-- the "Decision required" panel and Source labels the row, and widening it
-- back would send rows without them.
SELECT id, owner, repo, issue_number, title, status, workflow, stage,
SELECT id, owner, repo, issue_number, title, status, workflow,
pr_number, tokens_used, iterations, attempt, park_reason, retry_count,
created_at, updated_at, plan, source, identity, binding_id, binding_version
FROM tasks ORDER BY updated_at DESC LIMIT $1;
@@ -65,11 +65,11 @@ RETURNING *;
UPDATE tasks SET binding_id = $2, binding_version = $3, inputs = $4 WHERE id = $1;
-- name: UpdateTask :exec
UPDATE tasks SET workflow = $2, stage = $3, branch = $4, plan = $5, notes = $6,
pr_number = $7, tokens_used = $8, iterations = $9, park_reason = $10,
watch_comment_id = $11, retry_count = $12, remediation_rounds = $13,
review_payload = $14, workflow_definition_version = $15,
workflow_definition_digest = $16, workflow_definition_yaml = $17,
UPDATE tasks SET workflow = $2, branch = $3, plan = $4, notes = $5,
pr_number = $6, tokens_used = $7, iterations = $8, park_reason = $9,
watch_comment_id = $10, retry_count = $11, remediation_rounds = $12,
review_payload = $13, workflow_definition_version = $14,
workflow_definition_digest = $15, workflow_definition_yaml = $16,
updated_at = now()
WHERE id = $1;
@@ -129,7 +129,7 @@ WHERE owner = $1 AND repo = $2 AND issue_number = $3 AND status = 'queued'
RETURNING *;
-- name: BeginRemediationTask :execrows
UPDATE tasks SET status = 'queued', workflow = 'remediate', stage = '', park_reason = '', park_class = 'needs_human', review_payload = $2, updated_at = now()
UPDATE tasks SET status = 'queued', workflow = 'remediate', park_reason = '', park_class = 'needs_human', review_payload = $2, updated_at = now()
WHERE id = $1 AND status = 'pr_open';
-- name: UpdateReviewPayloadTask :execrows
@@ -143,13 +143,13 @@ WHERE id = $1 AND status = 'pr_open';
-- name: RequeueTask :execrows
UPDATE tasks SET status = 'queued',
workflow = CASE WHEN @workflow::text = '' THEN workflow ELSE @workflow::text END,
stage = '', park_reason = '', park_class = 'needs_human', updated_at = now()
park_reason = '', park_class = 'needs_human', updated_at = now()
WHERE id = @id AND status = @from_status;
-- name: RetryTask :execrows
UPDATE tasks SET status = 'queued', retry_count = retry_count + 1,
workflow = CASE WHEN @workflow::text = '' THEN workflow ELSE @workflow::text END,
stage = '', park_reason = '', park_class = 'needs_human', updated_at = now()
park_reason = '', park_class = 'needs_human', updated_at = now()
WHERE id = @id AND status = @from_status;
-- name: ArchiveTaskDelete :execrows
+35
View File
@@ -224,6 +224,41 @@ func stepEventData(to taskstate.StepStatus, durationMS int64, detail string) map
return data
}
// ListSteps reads one execution's recorded steps, every attempt oldest first
// when attempt is 0, or just the one it names. It is the dashboard run
// detail's authoritative source (docs/prds/execution-tree-state-machine.md),
// replacing the fold over stage_start/stage_finish events tasks.stage used to
// back -- unlike StartStep/FinishStep, this is a plain read with no
// transition to guard, so it takes no transaction.
func (s *Store) ListSteps(ctx context.Context, executionID int64, attempt int) ([]task.StepExecution, error) {
rows, err := s.queries().ListStepExecutions(ctx, postgresdb.ListStepExecutionsParams{
ExecutionID: executionID, Attempt: int64(attempt),
})
if err != nil {
return nil, err
}
steps := make([]task.StepExecution, len(rows))
for i, r := range rows {
steps[i] = task.StepExecution{
ID: r.ID, ExecutionID: r.ExecutionID, Attempt: int(r.Attempt),
ParentID: nullableInt64(r.ParentID), Depth: int(r.Depth),
Kind: r.Kind, Name: r.Name, Status: taskstate.StepStatus(r.Status),
Detail: r.Detail, TokensUsed: r.TokensUsed,
StartedAt: r.StartedAt.Time, FinishedAt: r.FinishedAt.Time,
}
}
return steps, nil
}
// nullableInt64 reads a pgtype.Int8 the way stepParentID wrote it: absent
// means 0, a stage at the tree's root.
func nullableInt64(v pgtype.Int8) int64 {
if !v.Valid {
return 0
}
return v.Int64
}
// stepParentID encodes the nullable parent id an insert takes: 0 means "no
// parent" -- a stage at the tree's root -- which the column stores as NULL
// rather than as a reference to a row that does not exist.
@@ -237,6 +237,71 @@ func TestFinishStepRefusesStaleAndForeign(t *testing.T) {
}
}
// TestListStepsOrdersByAttemptThenID is the dashboard run detail's
// authoritative read (docs/prds/execution-tree-state-machine.md): every
// attempt in one call when attempt is 0, one attempt's steps when it names
// one, both ordered oldest first.
func TestListStepsOrdersByAttemptThenID(t *testing.T) {
s := storeFor(t)
execution := runningExecution(t, s)
firstStepID, _, err := s.StartStep(t.Context(), task.StepStart{
ExecutionID: execution.ID, Attempt: execution.Attempt, Kind: task.StepKindStage, Name: "plan",
})
if err != nil {
t.Fatalf("StartStep: %v", err)
}
if _, err := s.FinishStep(t.Context(), task.StepFinish{
StepID: firstStepID, ExecutionID: execution.ID,
From: taskstate.StepRunning, To: taskstate.StepSucceeded, Detail: "planned", TokensUsed: 42,
}); err != nil {
t.Fatalf("FinishStep: %v", err)
}
secondStepID, _, err := s.StartStep(t.Context(), task.StepStart{
ExecutionID: execution.ID, Attempt: execution.Attempt, Kind: task.StepKindStage, Name: "implement",
})
if err != nil {
t.Fatalf("StartStep: %v", err)
}
steps, err := s.ListSteps(t.Context(), execution.ID, execution.Attempt)
if err != nil {
t.Fatalf("ListSteps: %v", err)
}
if len(steps) != 2 {
t.Fatalf("ListSteps len = %d, want 2", len(steps))
}
if steps[0].ID != firstStepID || steps[0].Name != "plan" || steps[0].Status != taskstate.StepSucceeded ||
steps[0].Detail != "planned" || steps[0].TokensUsed != 42 || steps[0].StartedAt.IsZero() || steps[0].FinishedAt.IsZero() {
t.Fatalf("steps[0] = %+v, want the finished plan step", steps[0])
}
if steps[1].ID != secondStepID || steps[1].Name != "implement" || steps[1].Status != taskstate.StepRunning ||
steps[1].StartedAt.IsZero() || !steps[1].FinishedAt.IsZero() {
t.Fatalf("steps[1] = %+v, want the still-running implement step", steps[1])
}
// attempt 0 lists every attempt; there is only one here, so the result is
// the same, proving the "list everything" branch is reachable and correct
// rather than coincidentally equal to a hardcoded attempt.
all, err := s.ListSteps(t.Context(), execution.ID, 0)
if err != nil {
t.Fatalf("ListSteps(attempt=0): %v", err)
}
if len(all) != 2 {
t.Fatalf("ListSteps(attempt=0) len = %d, want 2", len(all))
}
// A different execution's steps never appear, however queried.
other := runningExecution(t, s)
otherSteps, err := s.ListSteps(t.Context(), other.ID, 0)
if err != nil {
t.Fatalf("ListSteps other execution: %v", err)
}
if len(otherSteps) != 0 {
t.Fatalf("ListSteps other execution = %+v, want none", otherSteps)
}
}
// A late FinishStep for a step another path already finished is stale, the
// PRD's named wire behaviour: errors.Is matches on the client too.
func TestFinishStepOffTablePairIsIllegalOverTheTable(t *testing.T) {
+1 -3
View File
@@ -93,7 +93,6 @@ func taskFromRow(t postgresdb.Task) *workflow.Task {
WorkflowDefinitionVersion: t.WorkflowDefinitionVersion,
WorkflowDefinitionDigest: t.WorkflowDefinitionDigest,
WorkflowDefinitionYAML: t.WorkflowDefinitionYaml,
Stage: t.Stage,
Branch: t.Branch,
Plan: t.Plan,
Notes: t.Notes,
@@ -282,7 +281,6 @@ func (s *Store) Update(ctx context.Context, t *workflow.Task) error {
return s.queries().UpdateTask(ctx, postgresdb.UpdateTaskParams{
ID: t.ID,
Workflow: t.Workflow,
Stage: t.Stage,
Branch: t.Branch,
Plan: t.Plan,
Notes: t.Notes,
@@ -527,7 +525,7 @@ func (s *Store) Tasks(ctx context.Context, limit int) ([]workflow.Task, error) {
for _, r := range rows {
tasks = append(tasks, workflow.Task{
ID: r.ID, Owner: r.Owner, Repo: r.Repo, IssueNumber: int(r.IssueNumber),
Title: r.Title, Status: r.Status, Workflow: r.Workflow, Stage: r.Stage,
Title: r.Title, Status: r.Status, Workflow: r.Workflow,
PRNumber: int(r.PrNumber), TokensUsed: int(r.TokensUsed), Iterations: int(r.Iterations),
Attempt: int(r.Attempt), ParkReason: r.ParkReason, RetryCount: int(r.RetryCount),
CreatedAt: r.CreatedAt, UpdatedAt: r.UpdatedAt, Plan: r.Plan, Source: r.Source,
@@ -155,6 +155,18 @@ func (c *Client) FinishStep(ctx context.Context, s task.StepFinish) (events.Even
return event, nil
}
func (c *Client) ListSteps(ctx context.Context, executionID int64, attempt int) ([]task.StepExecution, error) {
r, err := c.client.ListSteps(ctx, &pb.ListStepsRequest{ExecutionId: executionID, Attempt: int64(attempt)})
if err != nil {
return nil, unmapError(err)
}
steps := make([]task.StepExecution, len(r.Steps))
for i, s := range r.Steps {
steps[i] = stepExecutionValue(s)
}
return steps, nil
}
func (c *Client) Update(ctx context.Context, t *task.Task) error {
_, err := c.client.Update(ctx, &pb.UpdateRequest{Task: taskProto(t)})
return unmapError(err)
@@ -106,6 +106,11 @@ type Deps struct {
// an operator action's guarded write over the execution and its steps.
// Optional: nil answers the RPC with codes.Unavailable.
Canceller storecontract.ExecutionCanceller
// StepReader lists an execution's recorded steps for the dashboard/API
// (docs/prds/execution-tree-state-machine.md): a plain read, never
// task-scoped-grant callable. Optional: nil answers ListSteps with
// codes.Unavailable.
StepReader storecontract.StepReader
// PlaybookDispatcher is the side-effecting-action idempotency ledger
// (docs/prds/eda-playbook-engine.md gap 2). Optional: nil disables the
// two playbook dispatch RPCs with codes.Unavailable.
@@ -247,6 +252,17 @@ func (s *server) FinishStep(ctx context.Context, r *pb.FinishStepRequest) (*pb.F
return &pb.FinishStepResponse{Event: eventProto(event)}, nil
}
func (s *server) ListSteps(ctx context.Context, r *pb.ListStepsRequest) (*pb.ListStepsResponse, error) {
if s.deps.StepReader == nil {
return nil, status.Error(codes.Unavailable, "step reader unavailable")
}
steps, err := s.deps.StepReader.ListSteps(ctx, r.ExecutionId, int(r.Attempt))
if err != nil {
return nil, s.logErr("ListSteps", err)
}
return &pb.ListStepsResponse{Steps: mapValues(steps, stepExecutionProto)}, nil
}
func (s *server) Update(ctx context.Context, r *pb.UpdateRequest) (*pb.UpdateResponse, error) {
if r.Task == nil {
return nil, status.Error(codes.InvalidArgument, "task is required")
+24 -2
View File
@@ -22,6 +22,7 @@ import (
"github.com/samcharles93/archie-core/internal/domain/workflow/task"
"github.com/samcharles93/archie-core/internal/events"
"github.com/samcharles93/archie-core/internal/logging"
"github.com/samcharles93/archie-core/internal/taskstate"
)
func timestamp(t time.Time) *timestamppb.Timestamp {
@@ -65,7 +66,7 @@ func taskProto(t *task.Task) *pb.Task { //nolint:dupl // mirror-image field-by-f
return &pb.Task{
Id: t.ID, Owner: t.Owner, Repo: t.Repo, IssueNumber: int64(t.IssueNumber),
Title: t.Title, Body: t.Body, Labels: t.Labels, Status: t.Status,
Workflow: t.Workflow, Stage: t.Stage, Branch: t.Branch, Plan: t.Plan, Notes: t.Notes,
Workflow: t.Workflow, Branch: t.Branch, Plan: t.Plan, Notes: t.Notes,
PrNumber: int64(t.PRNumber), TokensUsed: int64(t.TokensUsed), Iterations: int64(t.Iterations),
Attempt: int64(t.Attempt), ParkReason: t.ParkReason, RetryCount: int64(t.RetryCount),
WatchCommentId: t.WatchCommentID, Source: t.Source, Identity: t.Identity,
@@ -92,7 +93,7 @@ func taskValue(t *pb.Task) *task.Task { //nolint:dupl // see taskProto above
return &task.Task{
ID: t.Id, Owner: t.Owner, Repo: t.Repo, IssueNumber: int(t.IssueNumber),
Title: t.Title, Body: t.Body, Labels: t.Labels, Status: t.Status,
Workflow: t.Workflow, Stage: t.Stage, Branch: t.Branch, Plan: t.Plan, Notes: t.Notes,
Workflow: t.Workflow, Branch: t.Branch, Plan: t.Plan, Notes: t.Notes,
PRNumber: int(t.PrNumber), TokensUsed: int(t.TokensUsed), Iterations: int(t.Iterations),
Attempt: int(t.Attempt), ParkReason: t.ParkReason, RetryCount: int(t.RetryCount),
WatchCommentID: t.WatchCommentId, Source: t.Source, Identity: t.Identity,
@@ -306,6 +307,27 @@ func workflowStatValue(w *pb.WorkflowStat) storecontract.WorkflowStat {
}
}
func stepExecutionProto(s task.StepExecution) *pb.StepExecution {
return &pb.StepExecution{
Id: s.ID, ExecutionId: s.ExecutionID, Attempt: int64(s.Attempt),
ParentId: s.ParentID, Depth: int32(s.Depth), Kind: s.Kind, Name: s.Name,
Status: string(s.Status), Detail: s.Detail, TokensUsed: s.TokensUsed,
StartedAt: timestamp(s.StartedAt), FinishedAt: timestamp(s.FinishedAt),
}
}
func stepExecutionValue(s *pb.StepExecution) task.StepExecution {
if s == nil {
return task.StepExecution{}
}
return task.StepExecution{
ID: s.Id, ExecutionID: s.ExecutionId, Attempt: int(s.Attempt),
ParentID: s.ParentId, Depth: int(s.Depth), Kind: s.Kind, Name: s.Name,
Status: taskstate.StepStatus(s.Status), Detail: s.Detail, TokensUsed: s.TokensUsed,
StartedAt: timeValue(s.StartedAt), FinishedAt: timeValue(s.FinishedAt),
}
}
func stageStatProto(s storecontract.StageStat) *pb.StageStat {
return &pb.StageStat{Workflow: s.Workflow, Stage: s.Stage, Runs: int64(s.Runs), AvgMs: int64(s.AvgMs), Errors: int64(s.Errors)}
}
+1 -1
View File
@@ -17,7 +17,7 @@ func (s Store) TaskByID(ctx context.Context, id int64) (*taskactions.Task, error
if err != nil || t == nil {
return nil, err
}
return &taskactions.Task{ID: t.ID, Owner: t.Owner, Repo: t.Repo, Identity: t.Identity, Status: t.Status, Stage: t.Stage, ParkReason: t.ParkReason, IssueNumber: t.IssueNumber, RetryCount: t.RetryCount, Attempt: t.Attempt, ForgeBacked: t.IsForgeBacked()}, nil
return &taskactions.Task{ID: t.ID, Owner: t.Owner, Repo: t.Repo, Identity: t.Identity, Status: t.Status, ParkReason: t.ParkReason, IssueNumber: t.IssueNumber, RetryCount: t.RetryCount, Attempt: t.Attempt, ForgeBacked: t.IsForgeBacked()}, nil
}
// CancelExecution delegates to the store's one cancel path
+78 -94
View File
@@ -9,6 +9,7 @@ import (
"github.com/samcharles93/archie-core/internal/domain/workflow/task"
"github.com/samcharles93/archie-core/internal/events"
"github.com/samcharles93/archie-core/internal/taskstate"
)
// Stage/attempt statuses the attempt rail reports. They are one vocabulary
@@ -66,7 +67,12 @@ type taskStageView struct {
}
// handleTaskAttempts serves the stage rail: one task's attempts, each with the
// stages it recorded.
// stages it recorded. Attempt bounds, event counts and which attempt numbers
// exist still come from the task's events (an attempt can carry events with
// no step, e.g. agent calls, so events remain the complete attempt index);
// the stages themselves are step_executions, the authoritative record
// (docs/prds/execution-tree-state-machine.md), not a fold over
// stage_start/stage_finish events.
func (s *Server) handleTaskAttempts(w http.ResponseWriter, r *http.Request) {
t, ok := s.taskByPathID(w, r)
if !ok {
@@ -77,7 +83,15 @@ func (s *Server) handleTaskAttempts(w http.ResponseWriter, r *http.Request) {
http.Error(w, err.Error(), http.StatusInternalServerError)
return
}
attempts, unattributed := attemptsFromEvents(evs, t.Status, t.Attempt)
var steps []task.StepExecution
if s.Steps != nil {
steps, err = s.Steps.ListSteps(r.Context(), t.ID, 0)
if err != nil {
http.Error(w, err.Error(), http.StatusInternalServerError)
return
}
}
attempts, unattributed := attemptsFromEvents(evs, steps, t.Status, t.Attempt)
writeJSON(w, taskAttemptsView{
TaskID: t.ID,
CurrentAttempt: t.Attempt,
@@ -143,7 +157,7 @@ func (s *Server) taskAttemptTarget(w http.ResponseWriter, r *http.Request) (*tas
// unattributed -- every row written before the column existed and every
// deliberately task-agnostic producer carries it -- so grouping them would
// present unrelated activity as attempt zero's rail.
func attemptsFromEvents(evs []events.Event, taskStatus string, currentAttempt int) ([]taskAttemptView, int) {
func attemptsFromEvents(evs []events.Event, steps []task.StepExecution, taskStatus string, currentAttempt int) ([]taskAttemptView, int) {
grouped := make(map[int][]events.Event)
numbers := make([]int, 0, 4)
unattributed := 0
@@ -161,21 +175,34 @@ func attemptsFromEvents(evs []events.Event, taskStatus string, currentAttempt in
attempts := make([]taskAttemptView, 0, len(numbers))
for _, number := range numbers {
attempts = append(attempts, attemptRail(number, grouped[number], taskStatus, currentAttempt))
attempts = append(attempts, attemptRail(number, grouped[number], stepsForAttempt(steps, number), taskStatus, currentAttempt))
}
return attempts, unattributed
}
// attemptRail derives one attempt's status, bounds and stages from its events.
func attemptRail(number int, evs []events.Event, taskStatus string, currentAttempt int) taskAttemptView {
// stepsForAttempt filters ListSteps' full-execution result to one attempt's
// steps, in the order the query already returns them (attempt, id).
func stepsForAttempt(steps []task.StepExecution, attempt int) []task.StepExecution {
var out []task.StepExecution
for _, s := range steps {
if s.Attempt == attempt {
out = append(out, s)
}
}
return out
}
// attemptRail derives one attempt's status and bounds from its events, and
// its stages from its recorded StepExecutions.
func attemptRail(number int, evs []events.Event, steps []task.StepExecution, taskStatus string, currentAttempt int) taskAttemptView {
// The task's own record is the only honest source for "is this still
// happening": an unpaired start on any other attempt ended without
// happening": a step still running on any other attempt ended without
// finishing -- parked, retried, crashed or restarted -- so it is reported
// as interrupted rather than left looking permanently in flight.
inFlight := number == currentAttempt && taskStatus == task.StatusRunning
first, last := attemptBounds(evs)
stages := foldStages(evs, inFlight)
stages := stageViewsFromSteps(steps, inFlight)
attempt := taskAttemptView{
Attempt: number,
Status: attemptStatus(stages, inFlight),
@@ -215,34 +242,57 @@ func attemptBounds(evs []events.Event) (first, last time.Time) {
return first, last
}
// foldStages builds an attempt's ordered stage occurrences. A stage start is
// folded as an OPEN stage (status empty) until a matching finish closes it;
// resolving what is left needs the task's own record, so an open stage is
// running only while the daemon is executing this attempt and interrupted
// otherwise.
func foldStages(evs []events.Event, inFlight bool) []taskStageView {
// stageViewsFromSteps builds an attempt's ordered stage occurrences directly
// from its recorded StepExecutions -- the authoritative record
// (docs/prds/execution-tree-state-machine.md), not a fold over
// stage_start/stage_finish events. Only kind "stage" steps are shown: agent
// and call steps are the tree's own detail, not this rail's.
func stageViewsFromSteps(steps []task.StepExecution, inFlight bool) []taskStageView {
stages := []taskStageView{}
for _, e := range evs {
switch e.Kind {
case events.KindStageStart:
started := e.At
stages = append(stages, taskStageView{Name: e.Stage, Seq: len(stages), StartedAt: &started})
case events.KindStageFinish:
stages = finishStage(stages, e)
}
}
for i := range stages {
if stages[i].Status != "" {
for _, s := range steps {
if s.Kind != task.StepKindStage {
continue
}
stages[i].Status = attemptStatusInterrupted
if inFlight {
stages[i].Status = attemptStatusRunning
view := taskStageView{Name: s.Name, Seq: len(stages), Status: mapStepStatus(s.Status, inFlight)}
if s.Status == taskstate.StepFailed {
view.Error = s.Detail
}
if !s.StartedAt.IsZero() {
started := s.StartedAt
view.StartedAt = &started
}
if !s.FinishedAt.IsZero() {
duration := s.FinishedAt.Sub(s.StartedAt).Milliseconds()
view.DurationMS = &duration
}
stages = append(stages, view)
}
return stages
}
// mapStepStatus reads the rail's status vocabulary off a step's own recorded
// status. A step still recorded running past this attempt -- the daemon
// crashed or restarted without marking it -- reads interrupted rather than
// running: RecoverStale is what corrects the row itself, and until it runs
// this is the honest read of a stale one.
func mapStepStatus(status taskstate.StepStatus, inFlight bool) string {
switch status {
case taskstate.StepSucceeded:
return attemptStatusOK
case taskstate.StepFailed:
return attemptStatusFailed
case taskstate.StepCancelled, taskstate.StepInterrupted:
return attemptStatusInterrupted
case taskstate.StepRunning:
if inFlight {
return attemptStatusRunning
}
return attemptStatusInterrupted
default:
return attemptStatusUnknown
}
}
// attemptStatus reads one status off an attempt's stages. The task's own record
// outranks the events: an attempt the daemon is executing right now is running
// whatever its events have recorded so far -- including the ones that have
@@ -267,72 +317,6 @@ func attemptStatus(stages []taskStageView, inFlight bool) string {
}
}
// finishStage applies one stage_finish event, closing the earliest still-open
// stage of the same name. A finish with no matching start still yields the
// stage -- with no start time rather than an invented one -- because the stage
// demonstrably ran and dropping it would hide a failure.
func finishStage(stages []taskStageView, e events.Event) []taskStageView {
status, errText := stageOutcome(e.Data)
recorded, hasRecorded := durationMillis(e.Data["duration_ms"])
idx := -1
for i := range stages {
if stages[i].Name == e.Stage && stages[i].Status == "" {
idx = i
break
}
}
if idx < 0 {
stages = append(stages, taskStageView{Name: e.Stage, Seq: len(stages)})
idx = len(stages) - 1
}
stages[idx].Status = status
stages[idx].Error = errText
switch {
case hasRecorded:
// The producer's own measurement of the stage body, preferred over the
// gap between two event timestamps.
stages[idx].DurationMS = &recorded
case stages[idx].StartedAt != nil && !e.At.IsZero():
span := e.At.Sub(*stages[idx].StartedAt).Milliseconds()
stages[idx].DurationMS = &span
}
return stages
}
// stageOutcome reads a stage_finish event's own verdict. An interrupted finish
// is interrupted even though it also carries the cancellation error text; a
// non-empty error is a failure; anything else is ok. There is no exit code
// anywhere in this system, so no stage is ever badged as passing -- the status
// reports how the stage ended, not how well it went.
func stageOutcome(data map[string]any) (status, errText string) {
errText, _ = data["error"].(string)
if interrupted, _ := data["interrupted"].(bool); interrupted {
return attemptStatusInterrupted, errText
}
if errText != "" {
return attemptStatusFailed, errText
}
return attemptStatusOK, ""
}
// durationMillis reads a millisecond duration a producer recorded in an event's
// data. A value that crossed JSON is float64 and one built in this process is
// an integer, so both spellings are accepted; anything else is reported as "not
// recorded" rather than as zero.
func durationMillis(v any) (int64, bool) {
switch n := v.(type) {
case int:
return int64(n), true
case int64:
return n, true
case float64:
return int64(n), true
default:
return 0, false
}
}
func anyStageStatus(stages []taskStageView, status string) bool {
for _, s := range stages {
if s.Status == status {
+226 -179
View File
@@ -1,6 +1,7 @@
package webui
import (
"context"
"encoding/json"
"fmt"
"net/http"
@@ -11,19 +12,29 @@ import (
"time"
"github.com/samcharles93/archie-core/internal/domain/workflow"
"github.com/samcharles93/archie-core/internal/domain/workflow/task"
"github.com/samcharles93/archie-core/internal/events"
"github.com/samcharles93/archie-core/internal/taskstate"
)
// stepWriter is StepRecorder, asserted against srv.Store here rather than
// widened onto storecontract.TaskStore: the dashboard never writes a step
// itself, only reads them (Server.Steps, storecontract.StepReader), and this
// test needs real rows to exercise the read against.
type stepWriter interface {
StartStep(ctx context.Context, s task.StepStart) (int64, events.Event, error)
FinishStep(ctx context.Context, s task.StepFinish) (events.Event, error)
}
// railBase is the fixed clock the rail fixtures hang off, so every timestamp
// and duration asserted below is a literal rather than a tolerance.
var railBase = time.Date(2026, 9, 18, 7, 0, 0, 0, time.UTC)
func stageStartEvent(taskID int64, attempt int, stage string, at time.Time) events.Event {
return events.Event{Kind: events.KindStageStart, TaskID: taskID, Attempt: attempt, Stage: stage, At: at}
}
func stageFinishEvent(taskID int64, attempt int, stage string, at time.Time, data map[string]any) events.Event {
return events.Event{Kind: events.KindStageFinish, TaskID: taskID, Attempt: attempt, Stage: stage, At: at, Data: data}
// attemptEvent stands in for the events that still drive attempt bounds and
// counts (attemptsFromEvents), independent of the stage rail those events
// used to also back.
func attemptEvent(taskID int64, attempt int, at time.Time) events.Event {
return events.Event{Kind: events.KindAgentFinish, TaskID: taskID, Attempt: attempt, At: at}
}
func insertTaskEvent(t *testing.T, srv *Server, ev events.Event) {
@@ -33,6 +44,16 @@ func insertTaskEvent(t *testing.T, srv *Server, ev events.Event) {
}
}
// stageStep is a fixture StepExecution the pure-function test builds by
// hand -- no store round trip, since stageViewsFromSteps operates on
// task.StepExecution values directly.
func stageStep(name string, attempt int, status taskstate.StepStatus, startedAt, finishedAt time.Time, detail string) task.StepExecution {
return task.StepExecution{
Attempt: attempt, Kind: task.StepKindStage, Name: name, Status: status,
StartedAt: startedAt, FinishedAt: finishedAt, Detail: detail,
}
}
// getTaskAPI issues one task-detail request through the real mux. A route that
// is not registered falls through to the SPA asset handler and answers 200 with
// an HTML page, so using the mux makes a missing registration fail the shape
@@ -91,26 +112,6 @@ func twoAttemptTask(t *testing.T, srv *Server) *workflow.Task {
// summarizeAttemptRail renders a folded rail as one line per attempt so a table
// case states the derivation it expects without a struct literal per stage:
// "attempt:status[name:status:duration[#error][@nostart],...] duration".
func summarizeAttemptRail(attempts []taskAttemptView) string {
lines := make([]string, 0, len(attempts))
for _, a := range attempts {
stages := make([]string, 0, len(a.Stages))
for _, s := range a.Stages {
stage := s.Name + ":" + s.Status + ":" + durationToken(s.DurationMS)
if s.StartedAt == nil {
stage += "@nostart"
}
if s.Error != "" {
stage += "#" + s.Error
}
stages = append(stages, stage)
}
line := strconv.Itoa(a.Attempt) + ":" + a.Status + "[" + strings.Join(stages, ",") + "] " + durationToken(a.DurationMS)
lines = append(lines, line)
}
return strings.Join(lines, " | ")
}
func durationToken(ms *int64) string {
if ms == nil {
return "?"
@@ -118,161 +119,158 @@ func durationToken(ms *int64) string {
return strconv.FormatInt(*ms, 10)
}
// TestAttemptsFromEventsDerivesEveryRailRule is the table behind R1: every
// status rule the attempt rail depends on, one case each, plus the cases that
// would otherwise be answered by inventing data (a finish with no start, an
// event set with no stage information, events carrying no attempt at all).
func TestAttemptsFromEventsDerivesEveryRailRule(t *testing.T) {
// summarizeStages renders a stage list the same way summarizeAttemptRail's
// inner loop does, for a test that exercises stageViewsFromSteps alone.
func summarizeStages(stages []taskStageView) string {
tokens := make([]string, 0, len(stages))
for _, s := range stages {
token := s.Name + ":" + s.Status + ":" + durationToken(s.DurationMS)
if s.StartedAt == nil {
token += "@nostart"
}
if s.Error != "" {
token += "#" + s.Error
}
tokens = append(tokens, token)
}
return strings.Join(tokens, ",")
}
// TestStageViewsFromStepsDerivesEveryRailRule is the table behind R1: every
// status rule the rail depends on, sourced from StepExecutions -- the
// authoritative record (docs/prds/execution-tree-state-machine.md) -- rather
// than a fold over stage_start/stage_finish events.
func TestStageViewsFromStepsDerivesEveryRailRule(t *testing.T) {
at := railBase.Add
tests := []struct {
name string
steps []task.StepExecution
inFlight bool
want string
}{
{
name: "a succeeded step is ok and reports its measured span",
steps: []task.StepExecution{stageStep("prepare", 1, taskstate.StepSucceeded, at(0), at(time.Second), "")},
want: "prepare:ok:1000",
},
{
name: "a failed step fails and carries its detail as the error",
steps: []task.StepExecution{stageStep("implement", 1, taskstate.StepFailed, at(0), at(2*time.Second), "builder exited 1")},
want: "implement:failed:2000#builder exited 1",
},
{
name: "a cancelled step reads interrupted",
steps: []task.StepExecution{stageStep("implement", 1, taskstate.StepCancelled, at(0), at(time.Second), "stopped by operator")},
want: "implement:interrupted:1000",
},
{
name: "an interrupted step reads interrupted",
steps: []task.StepExecution{stageStep("implement", 1, taskstate.StepInterrupted, at(0), at(time.Second), "")},
want: "implement:interrupted:1000",
},
{
name: "a running step on the task's current in-flight attempt is running",
steps: []task.StepExecution{stageStep("implement", 2, taskstate.StepRunning, at(0), time.Time{}, "")},
inFlight: true,
want: "implement:running:?",
},
{
name: "a running step past the task's live attempt is stale, read as interrupted",
steps: []task.StepExecution{stageStep("implement", 1, taskstate.StepRunning, at(0), time.Time{}, "")},
inFlight: false,
want: "implement:interrupted:?",
},
{
name: "a pending step (never observed in practice) reads unknown rather than a false verdict",
steps: []task.StepExecution{stageStep("implement", 1, taskstate.StepPending, time.Time{}, time.Time{}, "")},
want: "implement:unknown:?@nostart",
},
{
name: "a repeated stage name yields one rail entry per occurrence",
steps: []task.StepExecution{
stageStep("implement", 1, taskstate.StepSucceeded, at(0), at(time.Second), ""),
{Attempt: 1, Kind: task.StepKindStage, Name: "implement", Status: taskstate.StepRunning, StartedAt: at(2 * time.Second)},
},
inFlight: true,
want: "implement:ok:1000,implement:running:?",
},
{
name: "an agent or call step is not part of the stage rail",
steps: []task.StepExecution{
stageStep("implement", 1, taskstate.StepSucceeded, at(0), at(time.Second), ""),
{Attempt: 1, Kind: task.StepKindAgent, Name: "implement", Status: taskstate.StepSucceeded, StartedAt: at(0), FinishedAt: at(time.Second)},
{Attempt: 1, Kind: task.StepKindCall, Name: "callee", Status: taskstate.StepSucceeded, StartedAt: at(0), FinishedAt: at(time.Second)},
},
want: "implement:ok:1000",
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
if got := summarizeStages(stageViewsFromSteps(tt.steps, tt.inFlight)); got != tt.want {
t.Errorf("stages = %q, want %q", got, tt.want)
}
})
}
}
// TestAttemptsFromEventsIndexesAttemptsFromEventsAndUnattributedCount pins
// what attemptsFromEvents still owns now that stage derivation moved to
// steps: which attempt numbers exist, and the unattributed-event count --
// both read from the task's events, since an attempt can carry events (agent
// calls, say) with no recorded step at all.
func TestAttemptsFromEventsIndexesAttemptsFromEventsAndUnattributedCount(t *testing.T) {
at := railBase.Add
duration := func(ms int64) map[string]any { return map[string]any{"duration_ms": ms} }
tests := []struct {
name string
taskStatus string
currentAttempt int
evs []events.Event
want string
steps []task.StepExecution
wantAttempts []int
wantUnattributed int
}{
{
name: "a finished stage is ok and the attempt reports its measured span",
taskStatus: workflow.StatusParked, currentAttempt: 1,
evs: []events.Event{
stageStartEvent(1, 1, "prepare", at(0)),
stageFinishEvent(1, 1, "prepare", at(time.Second), duration(1000)),
},
want: "1:ok[prepare:ok:1000] 1000",
},
{
name: "a failed stage fails the attempt and carries the error text",
taskStatus: workflow.StatusParked, currentAttempt: 1,
evs: []events.Event{
stageStartEvent(1, 1, "implement", at(0)),
stageFinishEvent(1, 1, "implement", at(2*time.Second), map[string]any{
"duration_ms": int64(2000),
"error": "stage implement: builder exited 1",
}),
},
want: "1:failed[implement:failed:2000#stage implement: builder exited 1] 2000",
},
{
name: "an interrupted stage reads interrupted even though the error text is recorded",
taskStatus: workflow.StatusParked, currentAttempt: 1,
evs: []events.Event{
stageStartEvent(1, 1, "implement", at(0)),
stageFinishEvent(1, 1, "implement", at(time.Second), map[string]any{
"duration_ms": int64(1000),
"error": "context canceled",
"interrupted": true,
}),
},
want: "1:interrupted[implement:interrupted:1000#context canceled] 1000",
},
{
name: "a failed stage outranks an interrupted one",
taskStatus: workflow.StatusParked, currentAttempt: 1,
evs: []events.Event{
stageStartEvent(1, 1, "a", at(0)),
stageFinishEvent(1, 1, "a", at(time.Second), map[string]any{"duration_ms": int64(1000), "error": "boom"}),
stageStartEvent(1, 1, "b", at(2*time.Second)),
stageFinishEvent(1, 1, "b", at(3*time.Second), map[string]any{
"duration_ms": int64(1000), "error": "canceled", "interrupted": true,
}),
},
want: "1:failed[a:failed:1000#boom,b:interrupted:1000#canceled] 3000",
},
{
name: "an unpaired start on the task's current in-flight attempt is running",
name: "a later attempt does not absorb the earlier attempt's steps",
taskStatus: workflow.StatusRunning, currentAttempt: 2,
evs: []events.Event{stageStartEvent(1, 2, "implement", at(0))},
want: "2:running[implement:running:?] ?",
},
{
name: "an unpaired start on a superseded attempt is interrupted",
taskStatus: workflow.StatusRunning, currentAttempt: 2,
evs: []events.Event{stageStartEvent(1, 1, "implement", at(0))},
want: "1:interrupted[implement:interrupted:?] 0",
},
{
name: "an unpaired start on a parked task's own attempt is interrupted",
taskStatus: workflow.StatusParked, currentAttempt: 1,
evs: []events.Event{stageStartEvent(1, 1, "implement", at(0))},
want: "1:interrupted[implement:interrupted:?] 0",
},
{
name: "a finish with no start yields the stage without an invented start time",
taskStatus: workflow.StatusParked, currentAttempt: 1,
evs: []events.Event{
stageFinishEvent(1, 1, "commit", at(0), duration(4000)),
},
want: "1:ok[commit:ok:4000@nostart] 0",
},
{
name: "a finish without a recorded duration derives it from its own start",
taskStatus: workflow.StatusParked, currentAttempt: 1,
evs: []events.Event{
stageStartEvent(1, 1, "prepare", at(0)),
stageFinishEvent(1, 1, "prepare", at(10*time.Second), nil),
},
want: "1:ok[prepare:ok:10000] 10000",
},
{
name: "a repeated stage name yields one rail entry per occurrence",
taskStatus: workflow.StatusRunning, currentAttempt: 1,
evs: []events.Event{
stageStartEvent(1, 1, "implement", at(0)),
stageFinishEvent(1, 1, "implement", at(time.Second), duration(1000)),
stageStartEvent(1, 1, "implement", at(2*time.Second)),
},
want: "1:running[implement:ok:1000,implement:running:?] ?",
},
{
name: "a later attempt does not absorb the earlier attempt's stages",
taskStatus: workflow.StatusRunning, currentAttempt: 2,
evs: []events.Event{
stageStartEvent(1, 1, "prepare", at(0)),
stageFinishEvent(1, 1, "prepare", at(time.Second), duration(1000)),
stageStartEvent(1, 2, "implement", at(2*time.Second)),
},
want: "1:ok[prepare:ok:1000] 1000 | 2:running[implement:running:?] ?",
evs: []events.Event{attemptEvent(1, 1, at(0)), attemptEvent(1, 2, at(time.Second))},
steps: []task.StepExecution{stageStep("prepare", 1, taskstate.StepSucceeded, at(0), at(time.Second), "")},
wantAttempts: []int{1, 2},
},
{
name: "events carrying no attempt are counted, never grouped",
taskStatus: workflow.StatusQueued, currentAttempt: 0,
evs: []events.Event{
{Kind: events.KindStageStart, TaskID: 1, Stage: "prepare", At: at(0)},
{Kind: events.KindStageFinish, TaskID: 1, Stage: "prepare", At: at(time.Second), Data: duration(1000)},
},
evs: []events.Event{{Kind: events.KindAgentFinish, TaskID: 1, At: at(0)}, {Kind: events.KindAgentFinish, TaskID: 1, At: at(time.Second)}},
wantUnattributed: 2,
},
{
name: "a task with no events has no attempts and nothing unattributed",
taskStatus: workflow.StatusQueued, currentAttempt: 0,
want: "",
},
{
name: "an attempt whose events carry no stage information is unknown",
name: "an attempt with events but no recorded step still appears, with no stages",
taskStatus: workflow.StatusParked, currentAttempt: 1,
evs: []events.Event{
{Kind: events.KindAgentFinish, TaskID: 1, Attempt: 1, At: at(0), Data: map[string]any{"status": "ok"}},
},
want: "1:unknown[] ?",
},
{
name: "the task's own record keeps the current attempt running with no stage events",
taskStatus: workflow.StatusRunning, currentAttempt: 1,
evs: []events.Event{
{Kind: events.KindAgentFinish, TaskID: 1, Attempt: 1, At: at(0), Data: map[string]any{"status": "ok"}},
},
want: "1:running[] ?",
evs: []events.Event{attemptEvent(1, 1, at(0))},
wantAttempts: []int{1},
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {
attempts, unattributed := attemptsFromEvents(tt.evs, tt.taskStatus, tt.currentAttempt)
if got := summarizeAttemptRail(attempts); got != tt.want {
t.Errorf("rail = %q, want %q", got, tt.want)
attempts, unattributed := attemptsFromEvents(tt.evs, tt.steps, tt.taskStatus, tt.currentAttempt)
var got []int
for _, a := range attempts {
got = append(got, a.Attempt)
}
if len(got) != len(tt.wantAttempts) {
t.Fatalf("attempts = %v, want %v", got, tt.wantAttempts)
}
for i, want := range tt.wantAttempts {
if got[i] != want {
t.Errorf("attempts = %v, want %v", got, tt.wantAttempts)
}
}
if unattributed != tt.wantUnattributed {
t.Errorf("unattributed = %d, want %d", unattributed, tt.wantUnattributed)
@@ -287,17 +285,67 @@ func TestAttemptsFromEventsDerivesEveryRailRule(t *testing.T) {
// attempt 2's stages.
func TestHandleTaskAttemptsRendersBothAttemptsOfARetry(t *testing.T) {
srv := newTestServer(t)
current := twoAttemptTask(t, srv)
at := railBase.Add
ctx := t.Context()
steps, ok := srv.Store.(stepWriter)
if !ok {
t.Fatal("test store does not implement stepWriter")
}
insertTaskEvent(t, srv, stageStartEvent(current.ID, 1, "prepare", at(0)))
insertTaskEvent(t, srv, stageFinishEvent(current.ID, 1, "prepare", at(1200*time.Millisecond),
map[string]any{"duration_ms": int64(1200)}))
insertTaskEvent(t, srv, stageStartEvent(current.ID, 1, "implement", at(1323*time.Millisecond)))
insertTaskEvent(t, srv, stageFinishEvent(current.ID, 1, "implement", at(231323*time.Millisecond),
map[string]any{"duration_ms": int64(230000), "error": "stage implement: builder exited 1"}))
insertTaskEvent(t, srv, stageStartEvent(current.ID, 2, "implement", at(time.Hour)))
// Not twoAttemptTask: StartStep guards on the execution's own current
// attempt (a step named for a superseded attempt is a stale transition
// by design -- steps.go, "a caller naming one the execution no longer
// runs is describing a run that is over"), so attempt 1's steps have to
// be recorded while the task is actually on attempt 1, before the retry
// that moves it to attempt 2.
if _, err := srv.Store.EnqueueIssue(ctx, "acme", "widget", 1, "task", "", "", ""); err != nil {
t.Fatal(err)
}
claimed1, err := srv.Store.ClaimNext(ctx)
if err != nil || claimed1 == nil || claimed1.Attempt != 1 {
t.Fatalf("ClaimNext = (%+v, %v), want attempt 1", claimed1, err)
}
prepareID, _, err := steps.StartStep(ctx, task.StepStart{ExecutionID: claimed1.ID, Attempt: 1, Kind: task.StepKindStage, Name: "prepare"})
if err != nil {
t.Fatalf("StartStep prepare: %v", err)
}
if _, err := steps.FinishStep(ctx, task.StepFinish{
StepID: prepareID, ExecutionID: claimed1.ID, From: taskstate.StepRunning, To: taskstate.StepSucceeded,
}); err != nil {
t.Fatalf("FinishStep prepare: %v", err)
}
implementID, _, err := steps.StartStep(ctx, task.StepStart{ExecutionID: claimed1.ID, Attempt: 1, Kind: task.StepKindStage, Name: "implement"})
if err != nil {
t.Fatalf("StartStep implement: %v", err)
}
if _, err := steps.FinishStep(ctx, task.StepFinish{
StepID: implementID, ExecutionID: claimed1.ID, From: taskstate.StepRunning, To: taskstate.StepFailed,
Detail: "stage implement: builder exited 1",
}); err != nil {
t.Fatalf("FinishStep implement: %v", err)
}
if err := srv.Store.Transition(ctx, claimed1.ID, workflow.StatusRunning, workflow.StatusParked, "stage implement: builder exited 1"); err != nil {
t.Fatal(err)
}
if err := srv.Store.RetryTask(ctx, claimed1.ID, workflow.StatusParked, ""); err != nil {
t.Fatal(err)
}
current, err := srv.Store.ClaimNext(ctx)
if err != nil || current == nil || current.Attempt != 2 {
t.Fatalf("second ClaimNext = (%+v, %v), want attempt 2", current, err)
}
if _, _, err := steps.StartStep(ctx, task.StepStart{ExecutionID: current.ID, Attempt: 2, Kind: task.StepKindStage, Name: "implement"}); err != nil {
t.Fatalf("StartStep attempt 2 implement: %v", err)
}
// No hand-inserted events: StartStep/FinishStep already write
// stage_start/stage_finish events in the same transaction as the step
// row (steps.go), which is what attemptBounds and event_count read --
// attempt 1 gets exactly the 4 those two StartStep+FinishStep pairs
// produce, attempt 2 the 1 its lone StartStep produces. Their
// timestamps are the store's own clock, so bounds are checked for
// presence and order below rather than against a literal railBase time.
w := getTaskAPI(t, srv, taskRoute(current.ID, "/attempts"))
if w.Code != http.StatusOK {
t.Fatalf("status = %d, body %s", w.Code, w.Body)
@@ -345,14 +393,13 @@ func TestHandleTaskAttemptsRendersBothAttemptsOfARetry(t *testing.T) {
if first.EventCount != 4 {
t.Errorf("attempt 1 event_count = %d, want 4", first.EventCount)
}
if first.StartedAt == nil || !first.StartedAt.Equal(at(0)) {
t.Errorf("attempt 1 started_at = %v, want %v", first.StartedAt, at(0))
// StartStep/FinishStep stamp their events with the store's own clock, so
// only presence, order and non-negativity are checked here.
if first.StartedAt == nil || first.FinishedAt == nil || first.FinishedAt.Before(*first.StartedAt) {
t.Errorf("attempt 1 bounds = started %v finished %v, want finished not before started", first.StartedAt, first.FinishedAt)
}
if first.FinishedAt == nil || !first.FinishedAt.Equal(at(231323*time.Millisecond)) {
t.Errorf("attempt 1 finished_at = %v, want %v", first.FinishedAt, at(231323*time.Millisecond))
}
if first.DurationMS == nil || *first.DurationMS != 231323 {
t.Errorf("attempt 1 duration_ms = %v, want 231323", first.DurationMS)
if first.DurationMS == nil || *first.DurationMS < 0 {
t.Errorf("attempt 1 duration_ms = %v, want a non-negative measured span", first.DurationMS)
}
if len(first.Stages) != 2 {
t.Fatalf("attempt 1 stages = %d, want 2 (%s)", len(first.Stages), w.Body)
@@ -361,11 +408,12 @@ func TestHandleTaskAttemptsRendersBothAttemptsOfARetry(t *testing.T) {
if prepare.Name != "prepare" || prepare.Seq != 0 || prepare.Status != "ok" || prepare.Error != "" {
t.Errorf("stage 0 = %+v, want prepare/0/ok with no error", prepare)
}
if prepare.DurationMS == nil || *prepare.DurationMS != 1200 {
t.Errorf("stage 0 duration_ms = %v, want 1200", prepare.DurationMS)
}
if prepare.StartedAt == nil || !prepare.StartedAt.Equal(at(0)) {
t.Errorf("stage 0 started_at = %v, want %v", prepare.StartedAt, at(0))
// StartStep/FinishStep stamp started_at/finished_at with the store's own
// clock (now()), not a caller-supplied time, so only their presence and
// non-negativity are checked here -- railBase no longer applies to a step
// row the way it did to a hand-inserted event.
if prepare.StartedAt == nil || prepare.DurationMS == nil || *prepare.DurationMS < 0 {
t.Errorf("stage 0 = %+v, want a recorded start and a non-negative duration", prepare)
}
implement := first.Stages[1]
if implement.Name != "implement" || implement.Seq != 1 || implement.Status != "failed" {
@@ -374,8 +422,8 @@ func TestHandleTaskAttemptsRendersBothAttemptsOfARetry(t *testing.T) {
if implement.Error != "stage implement: builder exited 1" {
t.Errorf("stage 1 error = %q, want the recorded error text", implement.Error)
}
if implement.DurationMS == nil || *implement.DurationMS != 230000 {
t.Errorf("stage 1 duration_ms = %v, want 230000", implement.DurationMS)
if implement.DurationMS == nil || *implement.DurationMS < 0 {
t.Errorf("stage 1 duration_ms = %v, want a non-negative measured duration", implement.DurationMS)
}
second := body.Attempts[1]
@@ -431,9 +479,8 @@ func TestHandleTaskAttemptsSeparatesUnattributedEventsFromAFirstRun(t *testing.T
if err != nil {
t.Fatal(err)
}
insertTaskEvent(t, srv, stageStartEvent(task.ID, 0, "prepare", railBase))
insertTaskEvent(t, srv, stageFinishEvent(task.ID, 0, "prepare", railBase.Add(time.Second),
map[string]any{"duration_ms": int64(1000)}))
insertTaskEvent(t, srv, attemptEvent(task.ID, 0, railBase))
insertTaskEvent(t, srv, attemptEvent(task.ID, 0, railBase.Add(time.Second)))
w := getTaskAPI(t, srv, taskRoute(task.ID, "/attempts"))
if w.Code != http.StatusOK {
+8
View File
@@ -26,6 +26,14 @@ type Server struct {
Store storecontract.TaskStore
Log *slog.Logger
// Steps lists one execution's recorded StepExecutions
// (docs/prds/execution-tree-state-machine.md): the run detail's
// authoritative source, GET /api/tasks/{id}/attempts. Nil degrades the
// route to reporting no steps rather than folding stage_start/
// stage_finish events, which tasks.stage used to back and no longer
// exists to fall back to.
Steps storecontract.StepReader
// ConfigSource supplies the configuration projection GET /api/config
// renders. Composition sets it to RemoteConfigView in a process that
// only displays configuration; unset means this process builds the view
+1 -1
View File
@@ -35,7 +35,7 @@ func newTestServer(t *testing.T) *Server {
t.Helper()
s := pgstore.Open(t)
t.Cleanup(func() { _ = s.Close() })
srv := &Server{Store: s, Log: slog.New(slog.DiscardHandler)}
srv := &Server{Store: s, Steps: s, Log: slog.New(slog.DiscardHandler)}
// Composition always gives the dashboard a Gateway contract; task
// actions travel through it now, so a test server without one could
// only ever answer 503.
+34 -1
View File
@@ -48,6 +48,9 @@ service StateStoreService {
// task-scoped grant): a cancel is a dashboard, API and dispatch action
// decided by the Authorizer, not a worker call.
rpc CancelExecution(CancelExecutionRequest) returns (CancelExecutionResponse);
// ListSteps is a dashboard/API/dispatch action (Authorizer: read on the
// run), not task-scoped-grant callable like StartStep/FinishStep.
rpc ListSteps(ListStepsRequest) returns (ListStepsResponse);
rpc Requeue(RequeueRequest) returns (RequeueResponse);
rpc ParkTask(ParkTaskRequest) returns (ParkTaskResponse);
rpc RecoverStale(RecoverStaleRequest) returns (RecoverStaleResponse);
@@ -262,7 +265,11 @@ message Task {
string labels = 7;
string status = 8;
string workflow = 9;
string stage = 10;
// Retired: the per-task "current stage" column was overwritten as
// workflow.Run advanced and carried no history. step_executions is the
// durable per-step record now (docs/prds/execution-tree-state-machine.md);
// ListSteps is its read.
string stage = 10 [deprecated = true];
string branch = 11;
string plan = 12;
string notes = 13;
@@ -576,6 +583,32 @@ message CancelExecutionResponse {
repeated int64 cancelled_step_ids = 1;
}
// StepExecution mirrors internal/domain/workflow/task.StepExecution.
message StepExecution {
int64 id = 1;
int64 execution_id = 2;
int64 attempt = 3;
// parent_id is 0 for a stage at the tree's root.
int64 parent_id = 4;
int32 depth = 5;
string kind = 6;
string name = 7;
string status = 8;
string detail = 9;
int64 tokens_used = 10;
google.protobuf.Timestamp started_at = 11;
google.protobuf.Timestamp finished_at = 12;
}
message ListStepsRequest {
int64 execution_id = 1;
// attempt selects one attempt's steps; 0 lists every attempt of the
// execution, oldest first -- the dashboard's run-detail rail renders every
// attempt from a single call rather than one per attempt.
int64 attempt = 2;
}
message ListStepsResponse { repeated StepExecution steps = 1; }
message UpdateRequest { Task task = 1; }
message UpdateResponse {}