[eersnington]

Pi Durable 1.4.0 runs generation retry backoffs and deferred polls as durable sleeping tasks, and reports when the Harness next needs a process through HarnessOptions.onWake. pi() used to find those waits by reading Pi's private task checkpoints (phase: "retry" with until, phase: "poll" with pollAt). Under 1.4.0 those tasks are sleeping, which that code did not handle.

flowchart LR
  onWake["Pi Durable onWake"] -->|active| awake["c.keepAwake"]
  onWake -->|at| schedule["c.schedule.at(at, 'pi.wake')"]
  schedule --> wake["pi.wake: resume due sleeps,<br/>schedule again if early"]
  onWake -->|neither| sleep["actor may sleep"]

Loading

	Before	After
pi-durable, chord peers	^1.0.0	^1.4.0
pi-telemetry	undeclared	peer ^1.4.0
When the actor sleeps	parsed checkpoints, waits over 60 s only	Pi's onWake, every sleeping task
onReport	reopened Pi on every report	reopens only on session_failed; passes ReportInfo to the app

• chore(pi): require Pi Durable 1.4.0: peers, pi-telemetry peer (the action types reference it), checkpointWhen on every defineDoc (now required), onReport(error, info).
• fix(pi): wake sleeping tasks through Pi Durable's onWake: wakeActor() replaces busyState, waitDeadline, and watchBusy.
  • Pi runs a sleep only once until <= now, and calls onWake only when the value changes. A wake that fires early re-reads harness.wake() and schedules again.
  • If c.schedule.at fails, the actor stays awake and resumes Pi on a timer.
  • The sleep drain waits on harness.wake().active and rechecks when the wake value changes, because Pi can settle that value after the last commit.
• fix(pi): reopen Pi only after a failed session: 1.2.0 and later report much more than storage failures through onReport.

Opening a stored actor migrates its SQLite to Pi Durable schema version 2. 1.3.0 and earlier cannot open it, so an actor cannot roll back past this release.

[railway-app]

This PR was not deployed automatically as @eersnington does not have access to the Railway project.

In order to get automatic PR deploys, please add @eersnington to your workspace on Railway.

[github-actions]

📟 Pip-Boy · f55c2fc

📦 Packages · fresh npm install resolve, no lockfile · this PR vs base

No notable install size, dependency, or entrypoint changes across 10 packages.

All 10 packages · linux-x64-gnu

package (linux-x64-gnu)	@latest	base	this PR	Δ vs base	packages
@rivet-dev/pi	429.8 MB	429.9 MB	429.9 MB	−143 B (−0.0%)	218 (0)
@rivet-dev/sandbox-adapter	76.0 MB	113.8 KB	113.8 KB	0	1 (0)
@sandbox-agent/cli-shared	27.2 KB	27.2 KB	27.2 KB	0	1 (0)
@sandbox-agent/persist-indexeddb	12.4 KB	12.4 KB	12.4 KB	0	1 (0)
@sandbox-agent/persist-postgres	12.3 KB	12.3 KB	12.3 KB	0	1 (0)
@sandbox-agent/persist-rivet	12.2 KB	12.2 KB	12.2 KB	0	1 (0)
@sandbox-agent/persist-sqlite	12.3 KB	12.3 KB	12.3 KB	0	1 (0)
@sandbox-agent/react	41.5 MB	41.9 MB	41.9 MB	0	15 (0)
acp-http-client	6.7 MB	6.7 MB	6.7 MB	0	3 (0)
sandbox-agent	30.9 MB	31.0 MB	31.0 MB	0	8 (0)

🧹 Rust dependencies

Skipped: this PR doesn't change any Rust code or dependencies.

🦀 Binaries

Skipped: this PR doesn't change any Rust code.

⏱ Timing

job	wall time	slowest step
changes	18s	checkout · 13s
zizmor	14s	checkout · 9s
footprint	3m05s	Run pnpm --dir sandbox-agent install --frozen-lockfile · 1m28s
⏭ native · ${{ matrix.platform }}	—	skipped
⏭ self-test	—	skipped
⏭ hygiene	—	skipped
report	34s	setup-node · 16s

🔭 Full CI trace for this commit

One OpenTelemetry trace of CI and Pip-Boy for f55c2fc: every job and step, with the sizes above attached. Run after CI finishes.

pnpm install --frozen-lockfile --filter 'pipboy...'
pnpm --filter pipboy pipboy trace --sha f55c2fc5c1a50e52b93bea9d26545006a988cbe8

Add --otlp <url> to send it to any OTLP/HTTP collector.

Agent instruction: From a rivet-dev/agents checkout, run the commands above, then read the trace file the last command prints. List the slowest jobs and steps and explain where CI time went.