feat: Add crew session cycling fix and daemon exponential backoff (gt-ws8ol)
- Fix crew next/prev: Pass session name via key binding to avoid run-shell context issue - Add TouchTownActivity() for town-level activity signaling - Implement daemon exponential backoff based on activity.json: - 0-5 min idle → 5 min heartbeat - 5-15 min idle → 10 min heartbeat - 15-45 min idle → 30 min heartbeat - 45+ min idle → 60 min heartbeat (max) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
This commit is contained in:
@@ -273,6 +273,11 @@ func init() {
|
|||||||
crewCmd.AddCommand(crewRenameCmd)
|
crewCmd.AddCommand(crewRenameCmd)
|
||||||
crewCmd.AddCommand(crewPristineCmd)
|
crewCmd.AddCommand(crewPristineCmd)
|
||||||
crewCmd.AddCommand(crewRestartCmd)
|
crewCmd.AddCommand(crewRestartCmd)
|
||||||
|
|
||||||
|
// Add --session flag to next/prev commands for tmux key binding support
|
||||||
|
// When run via run-shell, tmux session context may be wrong, so we pass it explicitly
|
||||||
|
crewNextCmd.Flags().StringVarP(&crewCycleSession, "session", "s", "", "tmux session name (for key bindings)")
|
||||||
|
crewPrevCmd.Flags().StringVarP(&crewCycleSession, "session", "s", "", "tmux session name (for key bindings)")
|
||||||
crewCmd.AddCommand(crewNextCmd)
|
crewCmd.AddCommand(crewNextCmd)
|
||||||
crewCmd.AddCommand(crewPrevCmd)
|
crewCmd.AddCommand(crewPrevCmd)
|
||||||
crewCmd.AddCommand(crewStartCmd)
|
crewCmd.AddCommand(crewStartCmd)
|
||||||
|
|||||||
@@ -8,16 +8,30 @@ import (
|
|||||||
"github.com/spf13/cobra"
|
"github.com/spf13/cobra"
|
||||||
)
|
)
|
||||||
|
|
||||||
|
// crewCycleSession is the --session flag for crew next/prev commands.
|
||||||
|
// When run via tmux key binding (run-shell), the session context may not be
|
||||||
|
// correct, so we pass the session name explicitly via #{session_name} expansion.
|
||||||
|
var crewCycleSession string
|
||||||
|
|
||||||
// cycleCrewSession switches to the next or previous crew session in the same rig.
|
// cycleCrewSession switches to the next or previous crew session in the same rig.
|
||||||
// direction: 1 for next, -1 for previous
|
// direction: 1 for next, -1 for previous
|
||||||
func cycleCrewSession(direction int) error {
|
// sessionOverride: if non-empty, use this instead of detecting current session
|
||||||
// Get current session (uses existing function from handoff.go)
|
func cycleCrewSession(direction int, sessionOverride string) error {
|
||||||
currentSession, err := getCurrentTmuxSession()
|
var currentSession string
|
||||||
if err != nil {
|
var err error
|
||||||
return fmt.Errorf("not in a tmux session: %w", err)
|
|
||||||
}
|
if sessionOverride != "" {
|
||||||
if currentSession == "" {
|
// Use the provided session name (from tmux key binding)
|
||||||
return fmt.Errorf("not in a tmux session")
|
currentSession = sessionOverride
|
||||||
|
} else {
|
||||||
|
// Get current session (uses existing function from handoff.go)
|
||||||
|
currentSession, err = getCurrentTmuxSession()
|
||||||
|
if err != nil {
|
||||||
|
return fmt.Errorf("not in a tmux session: %w", err)
|
||||||
|
}
|
||||||
|
if currentSession == "" {
|
||||||
|
return fmt.Errorf("not in a tmux session")
|
||||||
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
// Parse rig name from current session
|
// Parse rig name from current session
|
||||||
@@ -73,9 +87,9 @@ func cycleCrewSession(direction int) error {
|
|||||||
}
|
}
|
||||||
|
|
||||||
func runCrewNext(cmd *cobra.Command, args []string) error {
|
func runCrewNext(cmd *cobra.Command, args []string) error {
|
||||||
return cycleCrewSession(1)
|
return cycleCrewSession(1, crewCycleSession)
|
||||||
}
|
}
|
||||||
|
|
||||||
func runCrewPrev(cmd *cobra.Command, args []string) error {
|
func runCrewPrev(cmd *cobra.Command, args []string) error {
|
||||||
return cycleCrewSession(-1)
|
return cycleCrewSession(-1, crewCycleSession)
|
||||||
}
|
}
|
||||||
|
|||||||
@@ -22,6 +22,10 @@ across distributed teams of AI agents working on shared codebases.`,
|
|||||||
// Build command path: gt status, gt mail send, etc.
|
// Build command path: gt status, gt mail send, etc.
|
||||||
cmdPath := buildCommandPath(cmd)
|
cmdPath := buildCommandPath(cmd)
|
||||||
keepalive.TouchWithArgs(cmdPath, args)
|
keepalive.TouchWithArgs(cmdPath, args)
|
||||||
|
|
||||||
|
// Also signal town-level activity for daemon exponential backoff
|
||||||
|
// This resets the backoff when any gt command runs
|
||||||
|
keepalive.TouchTownActivity(cmdPath)
|
||||||
},
|
},
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|||||||
@@ -11,6 +11,7 @@ import (
|
|||||||
"syscall"
|
"syscall"
|
||||||
"time"
|
"time"
|
||||||
|
|
||||||
|
"github.com/steveyegge/gastown/internal/keepalive"
|
||||||
"github.com/steveyegge/gastown/internal/polecat"
|
"github.com/steveyegge/gastown/internal/polecat"
|
||||||
"github.com/steveyegge/gastown/internal/tmux"
|
"github.com/steveyegge/gastown/internal/tmux"
|
||||||
)
|
)
|
||||||
@@ -77,11 +78,13 @@ func (d *Daemon) Run() error {
|
|||||||
sigChan := make(chan os.Signal, 1)
|
sigChan := make(chan os.Signal, 1)
|
||||||
signal.Notify(sigChan, syscall.SIGINT, syscall.SIGTERM, syscall.SIGUSR1)
|
signal.Notify(sigChan, syscall.SIGINT, syscall.SIGTERM, syscall.SIGUSR1)
|
||||||
|
|
||||||
// Heartbeat ticker
|
// Dynamic heartbeat timer with exponential backoff based on activity
|
||||||
ticker := time.NewTicker(d.config.HeartbeatInterval)
|
// Start with base interval
|
||||||
defer ticker.Stop()
|
nextInterval := d.config.HeartbeatInterval
|
||||||
|
timer := time.NewTimer(nextInterval)
|
||||||
|
defer timer.Stop()
|
||||||
|
|
||||||
d.logger.Printf("Daemon running, heartbeat every %v", d.config.HeartbeatInterval)
|
d.logger.Printf("Daemon running, initial heartbeat interval %v", nextInterval)
|
||||||
|
|
||||||
// Initial heartbeat
|
// Initial heartbeat
|
||||||
d.heartbeat(state)
|
d.heartbeat(state)
|
||||||
@@ -102,12 +105,66 @@ func (d *Daemon) Run() error {
|
|||||||
return d.shutdown(state)
|
return d.shutdown(state)
|
||||||
}
|
}
|
||||||
|
|
||||||
case <-ticker.C:
|
case <-timer.C:
|
||||||
d.heartbeat(state)
|
d.heartbeat(state)
|
||||||
|
|
||||||
|
// Calculate next interval based on activity
|
||||||
|
nextInterval = d.calculateHeartbeatInterval()
|
||||||
|
timer.Reset(nextInterval)
|
||||||
|
d.logger.Printf("Next heartbeat in %v", nextInterval)
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
|
// Backoff thresholds for exponential slowdown when idle
|
||||||
|
const (
|
||||||
|
// Base interval when there's recent activity
|
||||||
|
baseInterval = 5 * time.Minute
|
||||||
|
|
||||||
|
// Tier thresholds for backoff
|
||||||
|
tier1Threshold = 5 * time.Minute // 0-5 min idle → 5 min interval
|
||||||
|
tier2Threshold = 15 * time.Minute // 5-15 min idle → 10 min interval
|
||||||
|
tier3Threshold = 45 * time.Minute // 15-45 min idle → 30 min interval
|
||||||
|
// 45+ min idle → 60 min interval (max)
|
||||||
|
|
||||||
|
// Corresponding intervals
|
||||||
|
tier1Interval = 5 * time.Minute
|
||||||
|
tier2Interval = 10 * time.Minute
|
||||||
|
tier3Interval = 30 * time.Minute
|
||||||
|
tier4Interval = 60 * time.Minute // max
|
||||||
|
)
|
||||||
|
|
||||||
|
// calculateHeartbeatInterval determines the next heartbeat interval based on activity.
|
||||||
|
// Reads ~/gt/daemon/activity.json to determine how long since the last gt/bd command.
|
||||||
|
// Returns exponentially increasing intervals as idle time grows.
|
||||||
|
//
|
||||||
|
// | Idle Duration | Next Heartbeat |
|
||||||
|
// |---------------|----------------|
|
||||||
|
// | 0-5 min | 5 min (base) |
|
||||||
|
// | 5-15 min | 10 min |
|
||||||
|
// | 15-45 min | 30 min |
|
||||||
|
// | 45+ min | 60 min (max) |
|
||||||
|
func (d *Daemon) calculateHeartbeatInterval() time.Duration {
|
||||||
|
activity := keepalive.ReadTownActivity()
|
||||||
|
if activity == nil {
|
||||||
|
// No activity file - assume recent activity (might be first run)
|
||||||
|
return baseInterval
|
||||||
|
}
|
||||||
|
|
||||||
|
idleDuration := activity.Age()
|
||||||
|
|
||||||
|
switch {
|
||||||
|
case idleDuration < tier1Threshold:
|
||||||
|
return tier1Interval
|
||||||
|
case idleDuration < tier2Threshold:
|
||||||
|
return tier2Interval
|
||||||
|
case idleDuration < tier3Threshold:
|
||||||
|
return tier3Interval
|
||||||
|
default:
|
||||||
|
return tier4Interval
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
// heartbeat performs one heartbeat cycle.
|
// heartbeat performs one heartbeat cycle.
|
||||||
// The daemon's job is minimal: ensure Deacon is running and send heartbeats.
|
// The daemon's job is minimal: ensure Deacon is running and send heartbeats.
|
||||||
// All health checking and decision-making belongs in the Deacon molecule.
|
// All health checking and decision-making belongs in the Deacon molecule.
|
||||||
|
|||||||
@@ -64,6 +64,69 @@ func TouchInWorkspace(workspaceRoot, command string) {
|
|||||||
_ = os.WriteFile(keepalivePath, data, 0644) // non-fatal: status file for debugging
|
_ = os.WriteFile(keepalivePath, data, 0644) // non-fatal: status file for debugging
|
||||||
}
|
}
|
||||||
|
|
||||||
|
// TouchTownActivity writes a town-level activity signal to ~/gt/daemon/activity.json.
|
||||||
|
// This is used by the daemon to implement exponential backoff when the town is idle.
|
||||||
|
// Any gt command activity resets the backoff to the base heartbeat interval.
|
||||||
|
// It silently ignores errors (best-effort signaling).
|
||||||
|
func TouchTownActivity(command string) {
|
||||||
|
// Get town root from GT_TOWN_ROOT or default to ~/gt
|
||||||
|
townRoot := os.Getenv("GT_TOWN_ROOT")
|
||||||
|
if townRoot == "" {
|
||||||
|
home, err := os.UserHomeDir()
|
||||||
|
if err != nil {
|
||||||
|
return
|
||||||
|
}
|
||||||
|
townRoot = filepath.Join(home, "gt")
|
||||||
|
}
|
||||||
|
|
||||||
|
daemonDir := filepath.Join(townRoot, "daemon")
|
||||||
|
|
||||||
|
// Ensure daemon directory exists
|
||||||
|
if err := os.MkdirAll(daemonDir, 0755); err != nil {
|
||||||
|
return
|
||||||
|
}
|
||||||
|
|
||||||
|
state := State{
|
||||||
|
LastCommand: command,
|
||||||
|
Timestamp: time.Now().UTC(),
|
||||||
|
}
|
||||||
|
|
||||||
|
data, err := json.Marshal(state)
|
||||||
|
if err != nil {
|
||||||
|
return
|
||||||
|
}
|
||||||
|
|
||||||
|
activityPath := filepath.Join(daemonDir, "activity.json")
|
||||||
|
_ = os.WriteFile(activityPath, data, 0644) // non-fatal: activity signal for daemon
|
||||||
|
}
|
||||||
|
|
||||||
|
// ReadTownActivity returns the current town-level activity state.
|
||||||
|
// Returns nil if the file doesn't exist or can't be read.
|
||||||
|
func ReadTownActivity() *State {
|
||||||
|
townRoot := os.Getenv("GT_TOWN_ROOT")
|
||||||
|
if townRoot == "" {
|
||||||
|
home, err := os.UserHomeDir()
|
||||||
|
if err != nil {
|
||||||
|
return nil
|
||||||
|
}
|
||||||
|
townRoot = filepath.Join(home, "gt")
|
||||||
|
}
|
||||||
|
|
||||||
|
activityPath := filepath.Join(townRoot, "daemon", "activity.json")
|
||||||
|
|
||||||
|
data, err := os.ReadFile(activityPath)
|
||||||
|
if err != nil {
|
||||||
|
return nil
|
||||||
|
}
|
||||||
|
|
||||||
|
var state State
|
||||||
|
if err := json.Unmarshal(data, &state); err != nil {
|
||||||
|
return nil
|
||||||
|
}
|
||||||
|
|
||||||
|
return &state
|
||||||
|
}
|
||||||
|
|
||||||
// Read returns the current keepalive state for the workspace.
|
// Read returns the current keepalive state for the workspace.
|
||||||
// Returns nil if the file doesn't exist or can't be read.
|
// Returns nil if the file doesn't exist or can't be read.
|
||||||
func Read(workspaceRoot string) *State {
|
func Read(workspaceRoot string) *State {
|
||||||
|
|||||||
Reference in New Issue
Block a user