Advance workflow stages from the state-driven coordinator

This commit is contained in:
Ian Gulliver
2026-10-21 13:11:03 -07:00
parent c77a5ba02b
commit be777e469c
6 changed files with 49 additions and 40 deletions
+3
View File
@@ -72,6 +72,9 @@ func (m *Manager) coordinatePass(ctx context.Context, started map[string]bool) {
for _, r := range work.SettleReviews {
once(fmt.Sprintf("settle:%s:%d", r.TaskID, r.N), func() { m.settleReviews(r) })
}
for _, id := range work.AdvanceStages {
once("advance:"+id, func() { m.advanceStage(id) })
}
for _, p := range work.ReadyPrompts {
once("ready:"+p.TaskID, func() { m.promptReady(p) })
}
+17 -12
View File
@@ -367,11 +367,6 @@ func (m *Manager) Recover(ctx context.Context) error {
return err
}
staged, err := m.recoverStages(ctx)
if err != nil {
return err
}
err = m.recoverPreviews(ctx)
if err != nil {
return err
@@ -399,9 +394,6 @@ func (m *Manager) Recover(ctx context.Context) error {
}
for _, sess := range sessions {
if staged[sess.Id] {
continue
}
task, err := m.taskFor(ctx, sess)
if err != nil {
slog.Error("resume session: load task", "session", sess.Id, "task", sess.TaskId, "err", err)
@@ -412,6 +404,23 @@ func (m *Manager) Recover(ctx context.Context) error {
if err != nil {
return err
}
started, err := m.store.SessionStarted(ctx, sess.Id)
if err != nil {
return err
}
finished, err := m.store.StageFinished(ctx, sess.Id)
if err != nil {
return err
}
if finished || !pending && !started {
m.endMergeTurn(sess.Id, true)
err = m.store.SetSessionState(ctx, sess.Id, api.AgentSessionStateIdle, "byre", nil)
if err != nil {
return err
}
slog.Info("session left idle at startup: its stage finished or it never got its first message", "session", sess.Id, "task", task.Id)
continue
}
if pending {
found, err := m.deliverRecorded(ctx, sess, task)
if err != nil {
@@ -1307,9 +1316,6 @@ func (m *Manager) turnEnded(ctx context.Context, l *live, res endResult) error {
if end.Discard != store.DiscardNone {
m.logUndelivered(ctx, l.id, "turn ended in discard")
if end.Discard == store.DiscardHeld && stageFinished {
go m.advanceStage(l)
}
return nil
}
if handoff {
@@ -1322,7 +1328,6 @@ func (m *Manager) turnEnded(ctx context.Context, l *live, res endResult) error {
return nil
}
if stageFinished {
go m.advanceStage(l)
return nil
}
go m.afterTurn(context.WithoutCancel(ctx), l)
+15 -25
View File
@@ -362,34 +362,24 @@ func (m *Manager) recordStage(ctx context.Context, l *live) error {
return m.store.RecordStage(ctx, l.taskID, output, head)
}
func (m *Manager) advanceStage(l *live) {
func (m *Manager) advanceStage(sessionID string) {
ctx := context.Background()
_, err := m.nextStage(ctx, l.id)
if err != nil {
slog.Error("advance workflow stage", "session", l.id, "err", err)
m.recordError(l, fmt.Sprintf("The workflow didn't move on from this stage: %v. byre tries again when it next starts.", err))
}
m.afterTurn(ctx, l)
}
func (m *Manager) recoverStages(ctx context.Context) (map[string]bool, error) {
ids, err := m.store.FinishedStages(ctx)
if err != nil {
return nil, err
}
started := map[string]bool{}
for _, id := range ids {
next, err := m.nextStage(ctx, id)
if err != nil {
slog.Error("advance workflow stage", "session", id, "err", err)
continue
next, err := m.nextStage(ctx, sessionID)
if err == nil {
slog.Info("advanced workflow stage", "session", sessionID, "next", next)
} else {
slog.Error("advance workflow stage", "session", sessionID, "err", err)
sess, serr := m.store.GetSession(ctx, sessionID)
if serr == nil {
m.fileError(ctx, sess.TaskId, fmt.Sprintf("The workflow didn't move on from this stage: %v. byre tries again when it next starts.", err))
}
for _, n := range next {
started[n] = true
}
slog.Info("advanced workflow stage", "session", id, "next", next)
}
return started, nil
m.mu.Lock()
l := m.live[sessionID]
m.mu.Unlock()
if l != nil {
m.afterTurn(ctx, l)
}
}
func (m *Manager) nextStage(ctx context.Context, sessionID string) ([]string, error) {