mirror of
https://github.com/prosolis/gogobee.git
synced 2026-07-15 08:32:41 +00:00
The emergence-seam roll added in 3ed2e1d covered extract, abandon,
forced extraction, and death-respawn — but not a natural run-complete
(boss down / dead-end node), which is the most common successful
emergence. Players who cleared a run cleanly never got the arrival roll
despite meeting every condition.
Wire maybeRollPetArrivalOnEmerge into the two real run-complete callers
(expeditionCmdRun foreground + the autorun background ticker), gated on
reason == stopComplete. Kept out of runAutopilotWalk itself so the sim
path (which calls the walk directly) never fires arrival DMs.
203 lines
6.4 KiB
Go
203 lines
6.4 KiB
Go
package plugin
|
|
|
|
import (
|
|
"log/slog"
|
|
"strings"
|
|
"time"
|
|
|
|
"gogobee/internal/db"
|
|
"maunium.net/go/mautrix/id"
|
|
)
|
|
|
|
// Phase 4 of expedition autopilot — background auto-run ticker. Walks
|
|
// active expeditions on real time so the player only engages when a real
|
|
// decision is needed (fork, elite, boss, low HP/SU, rare node, death).
|
|
//
|
|
// Design (settled 2026-05-16):
|
|
// • Tick every autoRunTickInterval; per-expedition rate-limited by
|
|
// autoRunCooldown via a last_autorun_at CAS.
|
|
// • Skip when the player is in active combat, inside the briefing/recap
|
|
// quiet window, or on a very-fresh expedition (give them a beat).
|
|
// • Walk up to autoRunRoomCap rooms per tick. Smaller than foreground's
|
|
// cap so background DMs stay digestible.
|
|
// • DM-suppression rules:
|
|
// - 0 rooms walked → silent (still stuck at a fork / blocked / etc.;
|
|
// the player already knows, no point spamming).
|
|
// - rooms > 0 with stopOK (hit room cap) → silent; we'll keep walking
|
|
// on the next tick. Cadence handles pacing; no "stretch complete"
|
|
// footer churn.
|
|
// - Any other reason with rooms > 0 → DM the bundled walk.
|
|
// • Idempotency: CAS-claim last_autorun_at before doing any work. A
|
|
// double-fire on the same expedition is a no-op.
|
|
|
|
const (
|
|
// autoRunTickInterval — how often the ticker wakes. The per-expedition
|
|
// cooldown is what actually paces; the tick just has to be frequent
|
|
// enough that the cooldown is enforced cleanly.
|
|
autoRunTickInterval = 5 * time.Minute
|
|
|
|
// autoRunCooldown — minimum gap between background walks for one
|
|
// expedition. 15 min keeps the dungeon moving while leaving room for
|
|
// the player to step in and steer if they want.
|
|
autoRunCooldown = 15 * time.Minute
|
|
|
|
// autoRunMinExpeditionAge — don't auto-walk a brand-new expedition;
|
|
// let the player walk the first beat manually so the opening reads
|
|
// as deliberate.
|
|
autoRunMinExpeditionAge = 30 * time.Minute
|
|
|
|
// autoRunRoomCap — rooms per background tick. Smaller than the
|
|
// foreground cap so a single background DM doesn't dump an
|
|
// overwhelming wall of narration.
|
|
autoRunRoomCap = 3
|
|
)
|
|
|
|
// expeditionAutoRunTicker — fires the auto-run pass every tick interval.
|
|
func (p *AdventurePlugin) expeditionAutoRunTicker() {
|
|
ticker := time.NewTicker(autoRunTickInterval)
|
|
defer ticker.Stop()
|
|
for {
|
|
select {
|
|
case <-p.stopCh:
|
|
return
|
|
case <-ticker.C:
|
|
p.fireExpeditionAutoRuns(time.Now().UTC())
|
|
}
|
|
}
|
|
}
|
|
|
|
// fireExpeditionAutoRuns walks active expeditions due for a background
|
|
// step and dispatches one walk per. Cheap per-expedition gating happens
|
|
// in tryAutoRun; this layer just selects candidates.
|
|
func (p *AdventurePlugin) fireExpeditionAutoRuns(now time.Time) {
|
|
// Reuse the ambient quiet window — same intent: don't talk over the
|
|
// scheduled 06:00 briefing or 21:00 recap.
|
|
if inAmbientQuietWindow(now) {
|
|
return
|
|
}
|
|
due, err := loadExpeditionsForAutoRun(
|
|
now.Add(-autoRunCooldown), now.Add(-autoRunMinExpeditionAge))
|
|
if err != nil {
|
|
slog.Error("expedition: load autorun candidates", "err", err)
|
|
return
|
|
}
|
|
for _, expID := range due {
|
|
e, err := getExpedition(expID)
|
|
if err != nil || e == nil {
|
|
continue
|
|
}
|
|
if e.Status != ExpeditionStatusActive {
|
|
continue
|
|
}
|
|
uid := id.UserID(e.UserID)
|
|
if hasActiveCombatSession(uid) {
|
|
continue
|
|
}
|
|
if err := p.tryAutoRun(e, now); err != nil {
|
|
slog.Warn("expedition: autorun step", "expedition", e.ID, "err", err)
|
|
}
|
|
}
|
|
}
|
|
|
|
// loadExpeditionsForAutoRun returns active expeditions whose
|
|
// last_autorun_at is NULL or older than runCutoff, AND whose start_date
|
|
// is older than ageCutoff so a freshly-started expedition isn't yanked
|
|
// out from under the player on tick 1.
|
|
func loadExpeditionsForAutoRun(runCutoff, ageCutoff time.Time) ([]string, error) {
|
|
rows, err := db.Get().Query(`
|
|
SELECT expedition_id
|
|
FROM dnd_expedition
|
|
WHERE status = 'active'
|
|
AND start_date < ?
|
|
AND (last_autorun_at IS NULL OR last_autorun_at < ?)`,
|
|
ageCutoff, runCutoff)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
defer rows.Close()
|
|
var ids []string
|
|
for rows.Next() {
|
|
var s string
|
|
if err := rows.Scan(&s); err != nil {
|
|
return nil, err
|
|
}
|
|
ids = append(ids, s)
|
|
}
|
|
return ids, rows.Err()
|
|
}
|
|
|
|
// tryAutoRun claims the slot for this expedition, runs one background
|
|
// walk, and DMs the result if the suppression rules say to. The CAS-
|
|
// update is the only persistent side effect on the autorun column —
|
|
// supplies/threat/run-graph state are mutated by the walk itself, just
|
|
// as they would be in a foreground !expedition run.
|
|
func (p *AdventurePlugin) tryAutoRun(e *Expedition, now time.Time) error {
|
|
cutoff := now.Add(-autoRunCooldown)
|
|
res, err := db.Get().Exec(`
|
|
UPDATE dnd_expedition
|
|
SET last_autorun_at = ?,
|
|
last_activity = ?
|
|
WHERE expedition_id = ?
|
|
AND status = 'active'
|
|
AND (last_autorun_at IS NULL OR last_autorun_at < ?)`,
|
|
now, now, e.ID, cutoff)
|
|
if err != nil {
|
|
return err
|
|
}
|
|
if n, _ := res.RowsAffected(); n == 0 {
|
|
return nil // someone else got there first
|
|
}
|
|
|
|
uid := id.UserID(e.UserID)
|
|
r := p.runAutopilotWalk(MessageContext{Sender: uid}, autoRunRoomCap, true)
|
|
if r.initErr != "" {
|
|
// "no expedition" / "no run" — race with abandon/extract. Silent.
|
|
return nil
|
|
}
|
|
// Emergence seam: a run-complete reached by the background ticker is
|
|
// still a live emergence — roll pet arrival. See maybeRollPetArrivalOnEmerge.
|
|
if r.reason == stopComplete {
|
|
p.maybeRollPetArrivalOnEmerge(uid)
|
|
}
|
|
if !shouldDMAutoRun(r) {
|
|
return nil
|
|
}
|
|
|
|
body := renderAutoRunDM(r)
|
|
if err := p.SendDM(uid, body); err != nil {
|
|
slog.Warn("expedition: autorun DM", "user", uid, "err", err)
|
|
}
|
|
return nil
|
|
}
|
|
|
|
// shouldDMAutoRun applies the background suppression rules. See the
|
|
// file-top design block for the rationale.
|
|
func shouldDMAutoRun(r autopilotWalkResult) bool {
|
|
if r.rooms == 0 {
|
|
return false
|
|
}
|
|
if r.reason == stopOK {
|
|
// Hit the per-tick room cap. Next tick will continue; no need to
|
|
// post a "stretch complete" filler DM.
|
|
return false
|
|
}
|
|
return true
|
|
}
|
|
|
|
// renderAutoRunDM bundles the staged walk narration into a single DM.
|
|
// Background can't pace via streamFlow, so we concatenate phases with a
|
|
// blank line between each beat and tack on the final message.
|
|
func renderAutoRunDM(r autopilotWalkResult) string {
|
|
var b strings.Builder
|
|
b.WriteString("🚶 *Auto-walk*\n\n")
|
|
for _, s := range r.stream {
|
|
if s == "" {
|
|
continue
|
|
}
|
|
b.WriteString(s)
|
|
b.WriteString("\n\n")
|
|
}
|
|
b.WriteString(r.finalMsg)
|
|
return b.String()
|
|
}
|