fix(server): reconcile active sessions after restart

[ayubun]

What Changed

Reconciles provider sessions that a previous server process left running or starting before the replacement process accepts commands. Persisted runtime bindings are marked stopped, stale active turns are cleared, and provider resume cursors are preserved. Pending approval and Input cards tied to stopped sessions are marked stale on submit, so they clear instead of repeating No active provider session errors.

I have been running this reconciliation locally as my t3-code backend. It cleared my stale Working threads and preserved their resume cursors.

Why

Restarting the server while work is active can leave threads stuck on Working (or Input) forever because the replacement process inherits session state for provider processes that no longer exist. Running this reconciliation during startup repairs those threads without calling provider stop or abort, keeps the conversation resumable, and gives stale provider callbacks an actionable restart-turn path.

Fixes#4584

UI Changes

For the Input state, this fix has a minor UI change.

Before: A dead provider leaves the Input card open, and every Submit adds another generic No active provider session failure.

After:Submit marks the callback stale, removes the dead Input card, and returns the thread to the normal composer with guidance to restart the turn.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes
• I included a video for animation/interaction changes

Models used

• openai/gpt-5.6-sol: ~65%
• anthropic/claude-fable-5: ~35%

---

Note

Medium Risk
Startup now mutates persisted session bindings and orchestration session state before accepting commands; a failed session.set can fail the whole boot. Binding write errors are logged and skipped.

Overview
Stops threads that a previous process left running or starting before the new server accepts commands, so they no longer sit on Working/Input forever.

reconcileProviderSessions runs at the prepared startup boundary. It marks matching directory bindings stopped (resume cursor kept, activeTurnId cleared) and dispatches thread.session.set the same way. Binding I/O failures are warnings; a failed session settle still fails startup.

Submitting leftover approval or user-input cards on a stopped session now records the existing stale-request failure instead of “No active provider session,” so the card can clear and the user can restart the turn.

^{Reviewed by Cursor Bugbot for commit 16271a3. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Reconcile active provider sessions on server restart and improve stale response errors

• Adds reconcileProviderSessions to serverRuntimeStartup.ts, which reads threads from ProjectionSnapshotQuery, transitions any running or starting sessions to stopped, clears activeTurnId in provider bindings, and dispatches thread.session.set commands before the launcher trial begins.
• Updates the thread.approval.respond and thread.user-input.respond handlers in ProviderCommandReactor.ts to use stalePendingRequestDetail with the request id instead of a generic 'no active session' message in failure details.
• Risk: reconcileProviderSessions is called before launcher.prepareTrial; if OrchestrationEngine.dispatch fails for a running session, the startup error propagates and may block server startup.

^{Macroscope summarized 16271a3.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 6a92ee44-e5e6-49c9-9981-0189a4dacbf1

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This PR adds new server startup logic that forcibly reconciles (stops) any sessions that were in 'running' or 'starting' state when the server restarts. This is a significant behavioral change to the startup sequence that affects session lifecycle management and warrants review by someone familiar with the orchestration system.

^{You can add or adjust custom eligibility rules. Learn more.}

[macroscopeapp]

Effect service conventions review: one convention issue found in the new startup reconciliation tests (standalone FooShape type used instead of the inferred Foo["Service"]). The rest of the changed code (namespace imports, dependency acquisition via yield* Service, Effect.catch on a fully handled error channel, no runtime escapes) matches the conventions.

^{Posted via Macroscope — Effect Service Conventions}

[ayubun]

What Changed

Reconciles provider sessions that a previous server process left running or starting before the replacement process accepts commands. Persisted runtime bindings are marked stopped, stale active turns are cleared, and provider resume cursors are preserved. Pending approval and Input cards tied to stopped sessions are marked stale on submit, so they clear instead of repeating No active provider session errors.

I have been running this reconciliation locally as my t3-code backend. It cleared my stale Working threads and preserved their resume cursors.

Why

Restarting the server while work is active can leave threads stuck on Working (or Input) forever because the replacement process inherits session state for provider processes that no longer exist. Running this reconciliation during startup repairs those threads without calling provider stop or abort, keeps the conversation resumable, and gives stale provider callbacks an actionable restart-turn path.

Fixes#4584

UI Changes

For the Input state, this fix has a minor UI change.

Before: A dead provider leaves the Input card open, and every Submit adds another generic No active provider session failure.

After:Submit marks the callback stale, removes the dead Input card, and returns the thread to the normal composer with guidance to restart the turn.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes
• I included a video for animation/interaction changes

Models used

• openai/gpt-5.6-sol: ~65%
• anthropic/claude-fable-5: ~35%

---

Note

Medium Risk
Startup now mutates persisted session bindings and orchestration session state before accepting commands; a failed session.set can fail the whole boot. Binding write errors are logged and skipped.

Overview
Stops threads that a previous process left running or starting before the new server accepts commands, so they no longer sit on Working/Input forever.

reconcileProviderSessions runs at the prepared startup boundary. It marks matching directory bindings stopped (resume cursor kept, activeTurnId cleared) and dispatches thread.session.set the same way. Binding I/O failures are warnings; a failed session settle still fails startup.

Submitting leftover approval or user-input cards on a stopped session now records the existing stale-request failure instead of “No active provider session,” so the card can clear and the user can restart the turn.

^{Reviewed by Cursor Bugbot for commit 16271a3. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Reconcile active provider sessions on server restart and improve stale response errors

• Adds reconcileProviderSessions to serverRuntimeStartup.ts, which reads threads from ProjectionSnapshotQuery, transitions any running or starting sessions to stopped, clears activeTurnId in provider bindings, and dispatches thread.session.set commands before the launcher trial begins.
• Updates the thread.approval.respond and thread.user-input.respond handlers in ProviderCommandReactor.ts to use stalePendingRequestDetail with the request id instead of a generic 'no active session' message in failure details.
• Risk: reconcileProviderSessions is called before launcher.prepareTrial; if OrchestrationEngine.dispatch fails for a running session, the startup error propagates and may block server startup.

^{Macroscope summarized 16271a3.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 6a92ee44-e5e6-49c9-9981-0189a4dacbf1

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This PR adds new server startup logic that forcibly reconciles (stops) any sessions that were in 'running' or 'starting' state when the server restarts. This is a significant behavioral change to the startup sequence that affects session lifecycle management and warrants review by someone familiar with the orchestration system.

^{You can add or adjust custom eligibility rules. Learn more.}

[macroscopeapp]

Effect service conventions review: one convention issue found in the new startup reconciliation tests (standalone FooShape type used instead of the inferred Foo["Service"]). The rest of the changed code (namespace imports, dependency acquisition via yield* Service, Effect.catch on a fully handled error channel, no runtime escapes) matches the conventions.

^{Posted via Macroscope — Effect Service Conventions}

[ayubun]

What Changed

Reconciles provider sessions that a previous server process left running or starting before the replacement process accepts commands. Persisted runtime bindings are marked stopped, stale active turns are cleared, and provider resume cursors are preserved. Pending approval and Input cards tied to stopped sessions are marked stale on submit, so they clear instead of repeating No active provider session errors.

I have been running this reconciliation locally as my t3-code backend. It cleared my stale Working threads and preserved their resume cursors.

Why

Restarting the server while work is active can leave threads stuck on Working (or Input) forever because the replacement process inherits session state for provider processes that no longer exist. Running this reconciliation during startup repairs those threads without calling provider stop or abort, keeps the conversation resumable, and gives stale provider callbacks an actionable restart-turn path.

Fixes#4584

UI Changes

For the Input state, this fix has a minor UI change.

Before: A dead provider leaves the Input card open, and every Submit adds another generic No active provider session failure.

After:Submit marks the callback stale, removes the dead Input card, and returns the thread to the normal composer with guidance to restart the turn.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes
• I included a video for animation/interaction changes

Models used

• openai/gpt-5.6-sol: ~65%
• anthropic/claude-fable-5: ~35%

---

Note

Medium Risk
Startup now mutates persisted session bindings and orchestration session state before accepting commands; a failed session.set can fail the whole boot. Binding write errors are logged and skipped.

Overview
Stops threads that a previous process left running or starting before the new server accepts commands, so they no longer sit on Working/Input forever.

reconcileProviderSessions runs at the prepared startup boundary. It marks matching directory bindings stopped (resume cursor kept, activeTurnId cleared) and dispatches thread.session.set the same way. Binding I/O failures are warnings; a failed session settle still fails startup.

Submitting leftover approval or user-input cards on a stopped session now records the existing stale-request failure instead of “No active provider session,” so the card can clear and the user can restart the turn.

^{Reviewed by Cursor Bugbot for commit 16271a3. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Reconcile active provider sessions on server restart and improve stale response errors

• Adds reconcileProviderSessions to serverRuntimeStartup.ts, which reads threads from ProjectionSnapshotQuery, transitions any running or starting sessions to stopped, clears activeTurnId in provider bindings, and dispatches thread.session.set commands before the launcher trial begins.
• Updates the thread.approval.respond and thread.user-input.respond handlers in ProviderCommandReactor.ts to use stalePendingRequestDetail with the request id instead of a generic 'no active session' message in failure details.
• Risk: reconcileProviderSessions is called before launcher.prepareTrial; if OrchestrationEngine.dispatch fails for a running session, the startup error propagates and may block server startup.

^{Macroscope summarized 16271a3.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 6a92ee44-e5e6-49c9-9981-0189a4dacbf1

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This PR adds new server startup logic that forcibly reconciles (stops) any sessions that were in 'running' or 'starting' state when the server restarts. This is a significant behavioral change to the startup sequence that affects session lifecycle management and warrants review by someone familiar with the orchestration system.

^{You can add or adjust custom eligibility rules. Learn more.}

[macroscopeapp]

Effect service conventions review: one convention issue found in the new startup reconciliation tests (standalone FooShape type used instead of the inferred Foo["Service"]). The rest of the changed code (namespace imports, dependency acquisition via yield* Service, Effect.catch on a fully handled error channel, no runtime escapes) matches the conventions.

^{Posted via Macroscope — Effect Service Conventions}

[ayubun]

What Changed

Reconciles provider sessions that a previous server process left running or starting before the replacement process accepts commands. Persisted runtime bindings are marked stopped, stale active turns are cleared, and provider resume cursors are preserved. Pending approval and Input cards tied to stopped sessions are marked stale on submit, so they clear instead of repeating No active provider session errors.

I have been running this reconciliation locally as my t3-code backend. It cleared my stale Working threads and preserved their resume cursors.

Why

Restarting the server while work is active can leave threads stuck on Working (or Input) forever because the replacement process inherits session state for provider processes that no longer exist. Running this reconciliation during startup repairs those threads without calling provider stop or abort, keeps the conversation resumable, and gives stale provider callbacks an actionable restart-turn path.

Fixes#4584

UI Changes

For the Input state, this fix has a minor UI change.

Before: A dead provider leaves the Input card open, and every Submit adds another generic No active provider session failure.

After:Submit marks the callback stale, removes the dead Input card, and returns the thread to the normal composer with guidance to restart the turn.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes
• I included a video for animation/interaction changes

Models used

• openai/gpt-5.6-sol: ~65%
• anthropic/claude-fable-5: ~35%

---

Note

Medium Risk
Startup now mutates persisted session bindings and orchestration session state before accepting commands; a failed session.set can fail the whole boot. Binding write errors are logged and skipped.

Overview
Stops threads that a previous process left running or starting before the new server accepts commands, so they no longer sit on Working/Input forever.

reconcileProviderSessions runs at the prepared startup boundary. It marks matching directory bindings stopped (resume cursor kept, activeTurnId cleared) and dispatches thread.session.set the same way. Binding I/O failures are warnings; a failed session settle still fails startup.

Submitting leftover approval or user-input cards on a stopped session now records the existing stale-request failure instead of “No active provider session,” so the card can clear and the user can restart the turn.

^{Reviewed by Cursor Bugbot for commit 16271a3. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Reconcile active provider sessions on server restart and improve stale response errors

• Adds reconcileProviderSessions to serverRuntimeStartup.ts, which reads threads from ProjectionSnapshotQuery, transitions any running or starting sessions to stopped, clears activeTurnId in provider bindings, and dispatches thread.session.set commands before the launcher trial begins.
• Updates the thread.approval.respond and thread.user-input.respond handlers in ProviderCommandReactor.ts to use stalePendingRequestDetail with the request id instead of a generic 'no active session' message in failure details.
• Risk: reconcileProviderSessions is called before launcher.prepareTrial; if OrchestrationEngine.dispatch fails for a running session, the startup error propagates and may block server startup.

^{Macroscope summarized 16271a3.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 6a92ee44-e5e6-49c9-9981-0189a4dacbf1

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This PR adds new server startup logic that forcibly reconciles (stops) any sessions that were in 'running' or 'starting' state when the server restarts. This is a significant behavioral change to the startup sequence that affects session lifecycle management and warrants review by someone familiar with the orchestration system.

^{You can add or adjust custom eligibility rules. Learn more.}

[macroscopeapp]

Effect service conventions review: one convention issue found in the new startup reconciliation tests (standalone FooShape type used instead of the inferred Foo["Service"]). The rest of the changed code (namespace imports, dependency acquisition via yield* Service, Effect.catch on a fully handled error channel, no runtime escapes) matches the conventions.

^{Posted via Macroscope — Effect Service Conventions}

[ayubun]

What Changed

Reconciles provider sessions that a previous server process left running or starting before the replacement process accepts commands. Persisted runtime bindings are marked stopped, stale active turns are cleared, and provider resume cursors are preserved. Pending approval and Input cards tied to stopped sessions are marked stale on submit, so they clear instead of repeating No active provider session errors.

I have been running this reconciliation locally as my t3-code backend. It cleared my stale Working threads and preserved their resume cursors.

Why

Restarting the server while work is active can leave threads stuck on Working (or Input) forever because the replacement process inherits session state for provider processes that no longer exist. Running this reconciliation during startup repairs those threads without calling provider stop or abort, keeps the conversation resumable, and gives stale provider callbacks an actionable restart-turn path.

Fixes#4584

UI Changes

For the Input state, this fix has a minor UI change.

Before: A dead provider leaves the Input card open, and every Submit adds another generic No active provider session failure.

After:Submit marks the callback stale, removes the dead Input card, and returns the thread to the normal composer with guidance to restart the turn.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes
• I included a video for animation/interaction changes

Models used

• openai/gpt-5.6-sol: ~65%
• anthropic/claude-fable-5: ~35%

---

Note

Medium Risk
Startup now mutates persisted session bindings and orchestration session state before accepting commands; a failed session.set can fail the whole boot. Binding write errors are logged and skipped.

Overview
Stops threads that a previous process left running or starting before the new server accepts commands, so they no longer sit on Working/Input forever.

reconcileProviderSessions runs at the prepared startup boundary. It marks matching directory bindings stopped (resume cursor kept, activeTurnId cleared) and dispatches thread.session.set the same way. Binding I/O failures are warnings; a failed session settle still fails startup.

Submitting leftover approval or user-input cards on a stopped session now records the existing stale-request failure instead of “No active provider session,” so the card can clear and the user can restart the turn.

^{Reviewed by Cursor Bugbot for commit 16271a3. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Reconcile active provider sessions on server restart and improve stale response errors

• Adds reconcileProviderSessions to serverRuntimeStartup.ts, which reads threads from ProjectionSnapshotQuery, transitions any running or starting sessions to stopped, clears activeTurnId in provider bindings, and dispatches thread.session.set commands before the launcher trial begins.
• Updates the thread.approval.respond and thread.user-input.respond handlers in ProviderCommandReactor.ts to use stalePendingRequestDetail with the request id instead of a generic 'no active session' message in failure details.
• Risk: reconcileProviderSessions is called before launcher.prepareTrial; if OrchestrationEngine.dispatch fails for a running session, the startup error propagates and may block server startup.

^{Macroscope summarized 16271a3.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 6a92ee44-e5e6-49c9-9981-0189a4dacbf1

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This PR adds new server startup logic that forcibly reconciles (stops) any sessions that were in 'running' or 'starting' state when the server restarts. This is a significant behavioral change to the startup sequence that affects session lifecycle management and warrants review by someone familiar with the orchestration system.

^{You can add or adjust custom eligibility rules. Learn more.}

[macroscopeapp]

Effect service conventions review: one convention issue found in the new startup reconciliation tests (standalone FooShape type used instead of the inferred Foo["Service"]). The rest of the changed code (namespace imports, dependency acquisition via yield* Service, Effect.catch on a fully handled error channel, no runtime escapes) matches the conventions.

^{Posted via Macroscope — Effect Service Conventions}

[ayubun]

What Changed

Reconciles provider sessions that a previous server process left running or starting before the replacement process accepts commands. Persisted runtime bindings are marked stopped, stale active turns are cleared, and provider resume cursors are preserved. Pending approval and Input cards tied to stopped sessions are marked stale on submit, so they clear instead of repeating No active provider session errors.

I have been running this reconciliation locally as my t3-code backend. It cleared my stale Working threads and preserved their resume cursors.

Why

Restarting the server while work is active can leave threads stuck on Working (or Input) forever because the replacement process inherits session state for provider processes that no longer exist. Running this reconciliation during startup repairs those threads without calling provider stop or abort, keeps the conversation resumable, and gives stale provider callbacks an actionable restart-turn path.

Fixes#4584

UI Changes

For the Input state, this fix has a minor UI change.

Before: A dead provider leaves the Input card open, and every Submit adds another generic No active provider session failure.

After:Submit marks the callback stale, removes the dead Input card, and returns the thread to the normal composer with guidance to restart the turn.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes
• I included a video for animation/interaction changes

Models used

• openai/gpt-5.6-sol: ~65%
• anthropic/claude-fable-5: ~35%

---

Note

Medium Risk
Startup now mutates persisted session bindings and orchestration session state before accepting commands; a failed session.set can fail the whole boot. Binding write errors are logged and skipped.

Overview
Stops threads that a previous process left running or starting before the new server accepts commands, so they no longer sit on Working/Input forever.

reconcileProviderSessions runs at the prepared startup boundary. It marks matching directory bindings stopped (resume cursor kept, activeTurnId cleared) and dispatches thread.session.set the same way. Binding I/O failures are warnings; a failed session settle still fails startup.

Submitting leftover approval or user-input cards on a stopped session now records the existing stale-request failure instead of “No active provider session,” so the card can clear and the user can restart the turn.

^{Reviewed by Cursor Bugbot for commit 16271a3. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Reconcile active provider sessions on server restart and improve stale response errors

• Adds reconcileProviderSessions to serverRuntimeStartup.ts, which reads threads from ProjectionSnapshotQuery, transitions any running or starting sessions to stopped, clears activeTurnId in provider bindings, and dispatches thread.session.set commands before the launcher trial begins.
• Updates the thread.approval.respond and thread.user-input.respond handlers in ProviderCommandReactor.ts to use stalePendingRequestDetail with the request id instead of a generic 'no active session' message in failure details.
• Risk: reconcileProviderSessions is called before launcher.prepareTrial; if OrchestrationEngine.dispatch fails for a running session, the startup error propagates and may block server startup.

^{Macroscope summarized 16271a3.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 6a92ee44-e5e6-49c9-9981-0189a4dacbf1

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This PR adds new server startup logic that forcibly reconciles (stops) any sessions that were in 'running' or 'starting' state when the server restarts. This is a significant behavioral change to the startup sequence that affects session lifecycle management and warrants review by someone familiar with the orchestration system.

^{You can add or adjust custom eligibility rules. Learn more.}

[macroscopeapp]

Effect service conventions review: one convention issue found in the new startup reconciliation tests (standalone FooShape type used instead of the inferred Foo["Service"]). The rest of the changed code (namespace imports, dependency acquisition via yield* Service, Effect.catch on a fully handled error channel, no runtime escapes) matches the conventions.

^{Posted via Macroscope — Effect Service Conventions}

[ayubun]

What Changed

Reconciles provider sessions that a previous server process left running or starting before the replacement process accepts commands. Persisted runtime bindings are marked stopped, stale active turns are cleared, and provider resume cursors are preserved. Pending approval and Input cards tied to stopped sessions are marked stale on submit, so they clear instead of repeating No active provider session errors.

I have been running this reconciliation locally as my t3-code backend. It cleared my stale Working threads and preserved their resume cursors.

Why

Restarting the server while work is active can leave threads stuck on Working (or Input) forever because the replacement process inherits session state for provider processes that no longer exist. Running this reconciliation during startup repairs those threads without calling provider stop or abort, keeps the conversation resumable, and gives stale provider callbacks an actionable restart-turn path.

Fixes#4584

UI Changes

For the Input state, this fix has a minor UI change.

Before: A dead provider leaves the Input card open, and every Submit adds another generic No active provider session failure.

After:Submit marks the callback stale, removes the dead Input card, and returns the thread to the normal composer with guidance to restart the turn.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes
• I included a video for animation/interaction changes

Models used

• openai/gpt-5.6-sol: ~65%
• anthropic/claude-fable-5: ~35%

---

Note

Medium Risk
Startup now mutates persisted session bindings and orchestration session state before accepting commands; a failed session.set can fail the whole boot. Binding write errors are logged and skipped.

Overview
Stops threads that a previous process left running or starting before the new server accepts commands, so they no longer sit on Working/Input forever.

reconcileProviderSessions runs at the prepared startup boundary. It marks matching directory bindings stopped (resume cursor kept, activeTurnId cleared) and dispatches thread.session.set the same way. Binding I/O failures are warnings; a failed session settle still fails startup.

Submitting leftover approval or user-input cards on a stopped session now records the existing stale-request failure instead of “No active provider session,” so the card can clear and the user can restart the turn.

^{Reviewed by Cursor Bugbot for commit 16271a3. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Reconcile active provider sessions on server restart and improve stale response errors

• Adds reconcileProviderSessions to serverRuntimeStartup.ts, which reads threads from ProjectionSnapshotQuery, transitions any running or starting sessions to stopped, clears activeTurnId in provider bindings, and dispatches thread.session.set commands before the launcher trial begins.
• Updates the thread.approval.respond and thread.user-input.respond handlers in ProviderCommandReactor.ts to use stalePendingRequestDetail with the request id instead of a generic 'no active session' message in failure details.
• Risk: reconcileProviderSessions is called before launcher.prepareTrial; if OrchestrationEngine.dispatch fails for a running session, the startup error propagates and may block server startup.

^{Macroscope summarized 16271a3.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 6a92ee44-e5e6-49c9-9981-0189a4dacbf1

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This PR adds new server startup logic that forcibly reconciles (stops) any sessions that were in 'running' or 'starting' state when the server restarts. This is a significant behavioral change to the startup sequence that affects session lifecycle management and warrants review by someone familiar with the orchestration system.

^{You can add or adjust custom eligibility rules. Learn more.}

[macroscopeapp]

Effect service conventions review: one convention issue found in the new startup reconciliation tests (standalone FooShape type used instead of the inferred Foo["Service"]). The rest of the changed code (namespace imports, dependency acquisition via yield* Service, Effect.catch on a fully handled error channel, no runtime escapes) matches the conventions.

^{Posted via Macroscope — Effect Service Conventions}

[ayubun]

What Changed

Reconciles provider sessions that a previous server process left running or starting before the replacement process accepts commands. Persisted runtime bindings are marked stopped, stale active turns are cleared, and provider resume cursors are preserved. Pending approval and Input cards tied to stopped sessions are marked stale on submit, so they clear instead of repeating No active provider session errors.

I have been running this reconciliation locally as my t3-code backend. It cleared my stale Working threads and preserved their resume cursors.

Why

Restarting the server while work is active can leave threads stuck on Working (or Input) forever because the replacement process inherits session state for provider processes that no longer exist. Running this reconciliation during startup repairs those threads without calling provider stop or abort, keeps the conversation resumable, and gives stale provider callbacks an actionable restart-turn path.

Fixes#4584

UI Changes

For the Input state, this fix has a minor UI change.

Before: A dead provider leaves the Input card open, and every Submit adds another generic No active provider session failure.

After:Submit marks the callback stale, removes the dead Input card, and returns the thread to the normal composer with guidance to restart the turn.

Checklist

• This PR is small and focused
• I explained what changed and why
• I included before/after screenshots for any UI changes
• I included a video for animation/interaction changes

Models used

• openai/gpt-5.6-sol: ~65%
• anthropic/claude-fable-5: ~35%

---

Note

Medium Risk
Startup now mutates persisted session bindings and orchestration session state before accepting commands; a failed session.set can fail the whole boot. Binding write errors are logged and skipped.

Overview
Stops threads that a previous process left running or starting before the new server accepts commands, so they no longer sit on Working/Input forever.

reconcileProviderSessions runs at the prepared startup boundary. It marks matching directory bindings stopped (resume cursor kept, activeTurnId cleared) and dispatches thread.session.set the same way. Binding I/O failures are warnings; a failed session settle still fails startup.

Submitting leftover approval or user-input cards on a stopped session now records the existing stale-request failure instead of “No active provider session,” so the card can clear and the user can restart the turn.

^{Reviewed by Cursor Bugbot for commit 16271a3. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Reconcile active provider sessions on server restart and improve stale response errors

• Adds reconcileProviderSessions to serverRuntimeStartup.ts, which reads threads from ProjectionSnapshotQuery, transitions any running or starting sessions to stopped, clears activeTurnId in provider bindings, and dispatches thread.session.set commands before the launcher trial begins.
• Updates the thread.approval.respond and thread.user-input.respond handlers in ProviderCommandReactor.ts to use stalePendingRequestDetail with the request id instead of a generic 'no active session' message in failure details.
• Risk: reconcileProviderSessions is called before launcher.prepareTrial; if OrchestrationEngine.dispatch fails for a running session, the startup error propagates and may block server startup.

^{Macroscope summarized 16271a3.}

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 6a92ee44-e5e6-49c9-9981-0189a4dacbf1

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — This PR adds new server startup logic that forcibly reconciles (stops) any sessions that were in 'running' or 'starting' state when the server restarts. This is a significant behavioral change to the startup sequence that affects session lifecycle management and warrants review by someone familiar with the orchestration system.

^{You can add or adjust custom eligibility rules. Learn more.}

[macroscopeapp]

Effect service conventions review: one convention issue found in the new startup reconciliation tests (standalone FooShape type used instead of the inferred Foo["Service"]). The rest of the changed code (namespace imports, dependency acquisition via yield* Service, Effect.catch on a fully handled error channel, no runtime escapes) matches the conventions.

^{Posted via Macroscope — Effect Service Conventions}