Repository navigation
Conversation
Limrun ends an instance after its inactivity timeout, which idle sessions hit while a model thinks between steps. The session's own client must send the ping, because a separate client leaves the ADB tunnel to go offline. The option is off by default so instance lifetime is unchanged unless requested.
Member
|
The PR is ready at 77b62f4. I found no problems in the diff. Not blocking: The Coverage check fails only in the eager-closure budget test. The extra module comes from the The live Limrun runs in the description (the session timing out without the flag, staying ready with it, and |
thymikee
added
the
ready-for-human
thymikee
marked this pull request as ready for review
October 4, 2026 05:45
thymikee
merged commit Oct 4, 2026
d1694fe
into
callstack:claude/limrun-scoped-token-attach-c41aff
14 of 15 checks passed
thymikee
added a commit
that referenced
this pull request
Oct 4, 2026
…3173) * feat(limrun): drive an existing instance with its own URL and token Read the lim CLI instance variables (LIM_IOS_INSTANCE_URL/TOKEN, LIM_ANDROID_INSTANCE_URL/TOKEN/ADB_URL). A platform with instance access attaches to that instance without LIMRUN_API_KEY and never creates or deletes it. Operations that need the organization API (apps, install, expired-lease recovery) fail with UNSUPPORTED_OPERATION when no key is set. * fix(limrun): keep the owner's state on attached instances Address review on #3173: attached Android teardown removes only the port reverse mappings this session created; app-log reconnect routes by the descriptor's instance id, so a created instance still reattaches through the organization API while another instance is attached; the attached reconnect honors its abort signal; the Android verification hint names the ADB URL variable; connect tests unstub env between tests. * refactor(limrun): simplify instance attach wiring Build attached sessions in the runtime from the existing session constructors instead of two near-identical allocation helpers, fold the attached URLs into the runtime fingerprint directly, merge the attached verification connector, and derive the connect error from the one list of instance variables. * fix(limrun): keep daemon boot independent of Limrun attach config A partial LIM_* instance set no longer stops the daemon: composition skips the Limrun runtime, reports it, and the daemon logs provider_runtime_skipped while every other provider loads. CLI connect still fails fast. Sessions and app-log descriptors carry a typed ownership ('created' or 'attached') instead of an attached- id prefix check; descriptors persisted before the field decode as created. The attached id helper moves into device.ts, so the provider entry evaluates no extra module. Expired-lease recovery is guarded on the API key, the connect message states its daemon condition, and the runtime test resets runCmd between tests. * fix(daemon): log skipped provider runtimes after publishing daemon info publishDaemonInfo truncates daemon.log, and before publication the daemon does not own the state directory, so the provider_runtime_skipped diagnostic now follows the startup diagnostics flush. * test(daemon): shut the skipped-provider startup runtime down in finally * feat(limrun): keep idle sessions alive with LIMRUN_KEEP_ALIVE (#3185) Limrun ends an instance after its inactivity timeout, which idle sessions hit while a model thinks between steps. The session's own client must send the ping, because a separate client leaves the ADB tunnel to go offline. The option is off by default so instance lifetime is unchanged unless requested. --------- Co-authored-by: Jakub Romańczyk <lorczyslav@gmail.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Limrun ends an instance after its inactivity timeout. An agent-device session that sits idle, for example while a model thinks between steps, loses its instance mid-run. With
LIMRUN_KEEP_ALIVE=1, the Limrun runtime pings each leased instance every 30 seconds from the session's own client until the lease is released. It is off by default, so instance lifetime is unchanged unless requested.export LIMRUN_KEEP_ALIVE=1 agent-device connect limrun --platform iosStacked on #3173: the flag is read in
src/provider-limrun-credentials.ts, which that PR adds. 6 files, including help and the Limrun docs page.Validation
At
77b62f4:pnpm check:affected --runpasses (139 files, 1043 tests).startKeepAlivefails the on/off and throwing-ping tests.terminated (InactivityTimeout).LIMRUN_KEEP_ALIVE=1, both stayedreadyandsnapshot -isucceeded after the idle. Release,daemon stopand a daemon SIGKILL still left the attached instances running.