feat: execute automation run now#983
Conversation
Astro-Han
added
enhancement
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (2)
🚧 Files skipped from review as they are similar to previous changes (2)
📝 WalkthroughWalkthroughAdds automation run execution with in-memory writer-key concurrency control, a RunExecutor contract, AutomationRunContext threading into session prompts (blocking/step-cap), server wiring to start background execution, and tests validating lifecycle, blockers, concurrency, and step-cap failures. ChangesAutomation Execution Feature
Sequence Diagram(s)sequenceDiagram
participant Client
participant API
participant Automation as Automation.runNowExecuting
participant Executor as sessionPromptExecutor
participant SessionPrompt
participant State as activeWriters
Client->>API: POST /:automationID/run
API->>Automation: runNowExecuting(id, { executor })
Automation->>Automation: runNow() -> scheduled
Automation->>State: try acquire writerKey
alt writerKey busy
Automation->>Automation: revise new run -> stopped(previous_run_awaiting_input)
Automation-->>Client: return scheduled run (immediate)
else acquired
Automation-->>Client: return scheduled run (immediate)
Automation->>Executor: invoke with AbortSignal
Executor->>SessionPrompt: promptWithAutomationContext(...)
SessionPrompt->>Executor: prompt result / abort
Executor-->>Automation: {sessionID, result, cost}
Automation->>Automation: publish RunUpdated -> succeeded/failed/cancelled
Automation->>State: release writerKey (finally)
end
Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes Possibly related PRs
Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
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. Comment |
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?s=60&v=4){kind=small}
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request introduces automation run execution and context management, allowing automations to run asynchronously with support for step caps, attendance tracking, and blocker handling (such as permission requests and user questions). Feedback on the changes highlights a few key issues: first, state-specific fields like blocker or stopReason must be deleted during state transitions in reviseRun to avoid strict Zod validation failures; second, the DefinitionUpdated event should be published when updating the definition's automation session ID; and finally, the block and clear handlers in the runner should be optimized to only publish updates when the run state actually changes, preventing redundant event publications.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 3
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
packages/opencode/src/server/instance/automation.ts (1)
233-234:⚠️ Potential issue | 🟡 Minor | ⚡ Quick winUpdate the now-stale route description.
This handler now actually executes the run in the background, so "Execution lands in a later PR." is no longer accurate for the published OpenAPI spec.
📝 Suggested wording
- description: "Create a scheduled automation run record. Execution lands in a later PR.", + description: "Queue an automation run and start background execution; returns the queued run immediately.",🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@packages/opencode/src/server/instance/automation.ts` around lines 233 - 234, Update the OpenAPI route description string for the handler with operationId "automation.runNow" to reflect that the run is executed in the background rather than deferred; replace the stale text "Create a scheduled automation run record. Execution lands in a later PR." with a concise, accurate message such as "Create a scheduled automation run record and start execution in the background." so the published spec matches the current behavior.
🧹 Nitpick comments (1)
packages/opencode/src/automation/runner.ts (1)
16-23: 💤 Low valuePrefer
Effect.genover.pipe(Effect.flatMap(...))for these handlers.The
block/clearcallbacks can be expressed withEffect.genfor clearer sequencing:♻️ Example
- block: (blocker) => - Effect.sync(() => { - currentRun = Automation.markRunBlocked(currentRun, blocker) - }).pipe(Effect.flatMap(() => Effect.promise(() => Automation.publishRunUpdated(currentRun)))), + block: (blocker) => + Effect.gen(function* () { + currentRun = Automation.markRunBlocked(currentRun, blocker) + yield* Effect.promise(() => Automation.publishRunUpdated(currentRun)) + }),As per coding guidelines: "Use
Effect.gen(function* () { ... })for Effect composition".🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@packages/opencode/src/automation/runner.ts` around lines 16 - 23, The block and clear handlers use Effect.sync(...).pipe(Effect.flatMap(...)) but should be rewritten using Effect.gen for clearer sequencing; replace the current block and clear implementations so they use Effect.gen(function* () { ... }) to update currentRun via Automation.markRunBlocked(currentRun, blocker) / Automation.clearRunBlocker(currentRun) and then yield Automation.publishRunUpdated(currentRun), ensuring you update the outer currentRun variable inside the generator and return the published effect rather than piping with Effect.flatMap.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@packages/opencode/src/automation/index.ts`:
- Line 583: The code uses a possibly stale variable definition (captured before
awaiting executor(...)) when calling setDefinitionAutomationSession(definition,
prepared.sessionID), risking overwriting concurrent updates; before calling
setDefinitionAutomationSession, re-fetch the latest definition (e.g., via the
same repository/service used elsewhere) and use that fresh object (so computing
revision: latest.revision + 1) and then pass the updated definition to
setDefinitionAutomationSession to avoid clobbering concurrent changes.
In `@packages/opencode/src/automation/run-context.ts`:
- Around line 35-40: The attended/unattended factory functions
(AutomationRunContext.attended and .unattended) currently spread input after the
literal attendance field, allowing any runtime attendance present on input to
override the intended mode; change both factories to spread input first and then
set attendance (i.e., return { ...input, attendance: "attended" } and {
...input, attendance: "unattended" }) so the constructor literal always wins and
runtime-attached attendance on the passed-in context cannot override the
intended mode.
In `@packages/opencode/src/automation/runner.ts`:
- Around line 14-25: The issue is that you create context with
AutomationRunContext.attended(...) then wrap it via
AutomationRunContext.unattended(context), which spreads the attended object over
the unattended literal and reverts attendance to "attended"; fix by constructing
the handlers object (with block and clear implementations that update currentRun
and call Automation.publishRunUpdated) once as, e.g., handlers = { stepCap: 50,
block: ..., clear: ... }, then call either
AutomationRunContext.attended(handlers) or
AutomationRunContext.unattended(handlers) based on the attendance variable when
assigning scoped so the attendance tag is set correctly.
---
Outside diff comments:
In `@packages/opencode/src/server/instance/automation.ts`:
- Around line 233-234: Update the OpenAPI route description string for the
handler with operationId "automation.runNow" to reflect that the run is executed
in the background rather than deferred; replace the stale text "Create a
scheduled automation run record. Execution lands in a later PR." with a concise,
accurate message such as "Create a scheduled automation run record and start
execution in the background." so the published spec matches the current
behavior.
---
Nitpick comments:
In `@packages/opencode/src/automation/runner.ts`:
- Around line 16-23: The block and clear handlers use
Effect.sync(...).pipe(Effect.flatMap(...)) but should be rewritten using
Effect.gen for clearer sequencing; replace the current block and clear
implementations so they use Effect.gen(function* () { ... }) to update
currentRun via Automation.markRunBlocked(currentRun, blocker) /
Automation.clearRunBlocker(currentRun) and then yield
Automation.publishRunUpdated(currentRun), ensuring you update the outer
currentRun variable inside the generator and return the published effect rather
than piping with Effect.flatMap.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: deead158-6dea-4c6c-924c-d4fae4b07012
📒 Files selected for processing (8)
packages/opencode/src/automation/index.tspackages/opencode/src/automation/run-context.tspackages/opencode/src/automation/runner.tspackages/opencode/src/permission/index.tspackages/opencode/src/server/instance/automation.tspackages/opencode/src/session/prompt.tspackages/opencode/test/server/automation-routes.test.tspackages/opencode/test/server/automation-runner.test.ts
|
Review follow-up pushed in 1150838. In addition to the inline fixes, I also updated the stale runNow OpenAPI description so it no longer says execution is deferred to a later PR. |
|
Review follow-up pushed in d5d233b. The remaining P1 boundaries are now covered: active writer locking is keyed by project/execution directory instead of automation ID, continue runs only bind |
There was a problem hiding this comment.
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
packages/opencode/test/server/automation-runner.test.ts (1)
15-21:⚠️ Potential issue | 🟡 Minor | ⚡ Quick winFix temp-dir disposal ordering in
withAutomation(return await)
Instance.providereturns the async result offn, butreturn Instance.provide(...)exits theawait usingscope immediately, allowingtmpdisposal to run before that Promise settles. Usereturn awaitso the temp git dir stays alive for the full automation run.🛡️ Proposed fix
async function withAutomation<T>(fn: (projectID: ProjectID) => Promise<T>) { await using tmp = await tmpdir({ git: true }) - return Instance.provide({ + return await Instance.provide({ directory: tmp.path, fn: () => fn(Instance.project.id), }) }🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@packages/opencode/test/server/automation-runner.test.ts` around lines 15 - 21, The withAutomation helper returns Instance.provide(...) while inside an "await using tmp = await tmpdir(...)" scope which causes tmp to be disposed before the async result resolves; change the return to "return await Instance.provide(...)" so the function awaits the provided Promise inside the using scope and the tmp dir (tmp from tmpdir) remains alive for the full automation run; locate the withAutomation function and update its return to await the Instance.provide call (referencing symbols: withAutomation, tmp, tmpdir, Instance.provide, Instance.project.id).
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Outside diff comments:
In `@packages/opencode/test/server/automation-runner.test.ts`:
- Around line 15-21: The withAutomation helper returns Instance.provide(...)
while inside an "await using tmp = await tmpdir(...)" scope which causes tmp to
be disposed before the async result resolves; change the return to "return await
Instance.provide(...)" so the function awaits the provided Promise inside the
using scope and the tmp dir (tmp from tmpdir) remains alive for the full
automation run; locate the withAutomation function and update its return to
await the Instance.provide call (referencing symbols: withAutomation, tmp,
tmpdir, Instance.provide, Instance.project.id).
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: fd0c4984-5284-4224-aa6c-841ae4706b4c
📒 Files selected for processing (5)
packages/opencode/src/automation/index.tspackages/opencode/src/automation/run-context.tspackages/opencode/src/automation/runner.tspackages/opencode/src/server/instance/automation.tspackages/opencode/test/server/automation-runner.test.ts
🚧 Files skipped from review as they are similar to previous changes (3)
- packages/opencode/src/server/instance/automation.ts
- packages/opencode/src/automation/runner.ts
- packages/opencode/src/automation/run-context.ts
|
Review follow-up pushed in b66d426. Deleting an automation now cancels any in-flight run for that automation: the active |
|
Review follow-up pushed in cc73650 for the remaining outside-diff review item: |
|
Review follow-up pushed in eda2ac4. The delete-active-automation path now propagates the automation abort signal into the real session prompt path by calling SessionPrompt.cancel(sessionID, { source: "automation.cancel" }) from sessionPromptExecutor and removing the listener after completion. Added a regression test with the production sessionPromptExecutor and a hanging SSE response: runNow starts the real prompt, Automation.remove aborts it, the run is returned as stopped/cancelled, and the session records MessageAbortedError instead of waiting for the model stream to finish. |
There was a problem hiding this comment.
Actionable comments posted: 1
🧹 Nitpick comments (1)
packages/opencode/test/server/automation-runner.test.ts (1)
322-346: ⚡ Quick winUse the
configoption instead of hand-writingopencode.jsonininit.
server.url.originis already available before thistmpdir(...)call, so the dynamicbaseURLcan be passed straight throughconfig. This also makes thepathimport (Line 1) removable.♻️ Switch from manual write to
config- await using tmp = await tmpdir({ - git: true, - init: async (dir) => { - await Bun.write( - path.join(dir, "opencode.json"), - JSON.stringify({ - $schema: "https://opencode.ai/config.json", - enabled_providers: ["alibaba"], - provider: { - alibaba: { - options: { - apiKey: "test-key", - baseURL: `${server.url.origin}/v1`, - }, - }, - }, - agent: { - build: { - model: "alibaba/qwen-plus", - }, - }, - }), - ) - }, - }) + await using tmp = await tmpdir({ + git: true, + config: { + enabled_providers: ["alibaba"], + provider: { + alibaba: { + options: { + apiKey: "test-key", + baseURL: `${server.url.origin}/v1`, + }, + }, + }, + agent: { + build: { + model: "alibaba/qwen-plus", + }, + }, + }, + })As per coding guidelines: "Use the
configoption intmpdirto write anopencode.jsonconfig file during test setup by passing a partial Config.Info object."Confirm the
tmpdirconfigoption shape accepts these fields:#!/bin/bash # Inspect tmpdir options and how `config` is handled in the test fixture fd -i 'fixture.ts' --exec rg -nP -C4 '\bconfig\b|opencode\.json|function tmpdir|tmpdir\s*[:=]' {}🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@packages/opencode/test/server/automation-runner.test.ts` around lines 322 - 346, Replace the manual opencode.json write inside the tmpdir init callback with the tmpdir "config" option: move the JSON object currently written in the init (the provider/agent settings using server.url.origin for baseURL and model) into the tmpdir call's config parameter (pass a partial Config.Info object), remove the init that uses Bun.write and the path import, and ensure tmpdir is called with git: true and the same config values so server.url.origin is injected directly as baseURL; keep tmpdir variable name and test flow unchanged.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@packages/opencode/test/server/automation-runner.test.ts`:
- Around line 302-303: The assertion against removed.stoppedRun is vacuous;
instead subscribe to the run event bus and assert that a RunUpdated event with
state "succeeded" is NOT emitted after the abort. Replace the current check
using removed.stoppedRun with code that listens to the event stream (the same
bus used elsewhere in tests), start a short listener before calling
executeRun/abort, collect RunUpdated events for the target run id, await the
abort completion, then assert the collected events contain no RunUpdated with
state: "succeeded"; reference the existing symbols removed.stoppedRun,
executeRun, Automation.runs (explain why not used), and RunUpdated to locate the
relevant logic.
---
Nitpick comments:
In `@packages/opencode/test/server/automation-runner.test.ts`:
- Around line 322-346: Replace the manual opencode.json write inside the tmpdir
init callback with the tmpdir "config" option: move the JSON object currently
written in the init (the provider/agent settings using server.url.origin for
baseURL and model) into the tmpdir call's config parameter (pass a partial
Config.Info object), remove the init that uses Bun.write and the path import,
and ensure tmpdir is called with git: true and the same config values so
server.url.origin is injected directly as baseURL; keep tmpdir variable name and
test flow unchanged.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: 745a1962-8be1-4c40-a79a-574de8e7a122
📒 Files selected for processing (4)
packages/opencode/src/automation/index.tspackages/opencode/src/automation/runner.tspackages/opencode/src/server/instance/automation.tspackages/opencode/test/server/automation-runner.test.ts
🚧 Files skipped from review as they are similar to previous changes (1)
- packages/opencode/src/automation/runner.ts
|
Review follow-up pushed in decdbb2. Verified the pre-busy cancel race: SessionRunState.cancel is a no-op when no runner is busy, so automation aborts now also flow into prompt setup via an AbortSignal passed to SessionPrompt.promptWithAutomationContext. The prompt path checks that signal before setup, after user-message setup, immediately before ensureRunning, and inside the runner work before setting busy, so a delete between run-start publication and runner registration cannot continue into LLM/tool side effects. Added a regression test that deletes after the run reaches running but before the prompt runner is busy and asserts the provider is never called and no assistant message is created.\n\nLocal verification:\n- bun test test/server/automation-runner.test.ts\n- bun test test/server/automation-routes.test.ts\n- bun run typecheck\n- git diff --check |
|
Review follow-up pushed in 39f7cee for the two P3 items. The mock-executor delete test now records RunUpdated events and asserts no succeeded event is published after cancellation instead of checking the stopped snapshot. The automation.delete OpenAPI description now documents that an active run is stopped and the stopped run is published before the tombstone; added an OpenAPI assertion for that wording.\n\nLocal verification:\n- bun test test/server/automation-runner.test.ts\n- bun test test/server/automation-routes.test.ts\n- bun run typecheck\n- git diff --check |
There was a problem hiding this comment.
Actionable comments posted: 1
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
packages/opencode/test/server/automation-routes.test.ts (1)
20-29:⚠️ Potential issue | 🟠 Major | ⚡ Quick winAwait
Instance.provide(...)in this helper.Returning the promise directly lets the
await using tmpscope dispose before the async test body fully settles, which can make these route tests race against temp-dir cleanup.💡 Minimal fix
async function withAutomationApp<T>(fn: (input: { app: Hono; projectID: ProjectID }) => Promise<T>) { await using tmp = await tmpdir({ git: true }) - return Instance.provide({ + return await Instance.provide({ directory: tmp.path, fn: async () => { const app = new Hono().route("/automation", AutomationRoutes()) app.onError(ErrorMiddleware) return fn({ app, projectID: Instance.project.id }) }, }) }🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@packages/opencode/test/server/automation-routes.test.ts` around lines 20 - 29, The helper withAutomationApp returns the promise from Instance.provide without awaiting it, allowing the await using tmp scope to dispose the temp dir before the test body finishes; change the function to await Instance.provide(...) so the tmpdir created by tmpdir({ git: true }) remains in scope for the duration of the provided fn. Specifically, in withAutomationApp, await the call to Instance.provide (instead of returning it directly) so that the using tmp lifetime encloses the async work passed into fn; keep the rest of the logic (creating app = new Hono().route("/automation", AutomationRoutes()) and app.onError(ErrorMiddleware)) unchanged.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@packages/opencode/test/server/automation-runner.test.ts`:
- Around line 421-433: The subscription created with Bus.subscribe listening for
Automation.Event.RunUpdated can leak if the timeout path wins because
unsubscribe() is only called after the awaited Promise.race; wrap the await in a
try/finally (or attach a .finally) to ensure unsubscribe() is always invoked
regardless of success or timeout, e.g. call unsubscribe() in the finally block
around the Automation.runNowExecuting(...) + Promise.race call so the listener
is removed even when the timeout rejects.
---
Outside diff comments:
In `@packages/opencode/test/server/automation-routes.test.ts`:
- Around line 20-29: The helper withAutomationApp returns the promise from
Instance.provide without awaiting it, allowing the await using tmp scope to
dispose the temp dir before the test body finishes; change the function to await
Instance.provide(...) so the tmpdir created by tmpdir({ git: true }) remains in
scope for the duration of the provided fn. Specifically, in withAutomationApp,
await the call to Instance.provide (instead of returning it directly) so that
the using tmp lifetime encloses the async work passed into fn; keep the rest of
the logic (creating app = new Hono().route("/automation", AutomationRoutes())
and app.onError(ErrorMiddleware)) unchanged.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro Plus
Run ID: 0127fbb3-c457-473b-9830-cb410b6977a6
📒 Files selected for processing (5)
packages/opencode/src/automation/runner.tspackages/opencode/src/server/instance/automation.tspackages/opencode/src/session/prompt.tspackages/opencode/test/server/automation-routes.test.tspackages/opencode/test/server/automation-runner.test.ts
🚧 Files skipped from review as they are similar to previous changes (1)
- packages/opencode/src/automation/runner.ts
|
Review follow-up pushed in 722a6f6. I fixed the PR2-scoped P3: executeRun now only publishes the running update when it actually transitions the run from scheduled to running, so an executor that already marked/published running does not produce duplicate running events. Added a RunUpdated sequence test that expects [running, succeeded].\n\nI did not change the fire-and-forget supervision shape or introduce a new writer-busy stopReason in this PR: the first is a scheduler/service-layer supervision follow-up, and the second changes the run stopReason contract. Both are better handled in a later execution-runtime/contract slice rather than broadening PR2.\n\nLocal verification:\n- bun test test/server/automation-runner.test.ts\n- bun test test/server/automation-routes.test.ts\n- bun run typecheck\n- git diff --check |
|
Review follow-up pushed in 67dc4f0. The pre-busy delete regression test now wraps the timeout race in try/finally so its RunUpdated subscription is always cleaned up, including the timeout failure path.\n\nLocal verification:\n- bun test test/server/automation-runner.test.ts\n- bun run typecheck\n- git diff --check |
Closes the backend gaps left after PR1-5 so the PR6/PR7 frontend slice can
build on a complete, stable automation contract.
Goal / change boundary:
- model { providerID, modelID } now required on every AutomationDefinition;
runner passes it through so runs no longer depend on the runtime model
fallback chain (which drifts across restarts).
- variant? optional reasoning/effort, validated against the live provider
catalog on create/update; invalid variant returns 422 instead of failing
late in the run.
- stop.kind === "condition" rejected at create/update with structured
{ field: "stop", message: "unsupported_stop_condition" } (scheduler never
schedules condition stops, so accepting them only leaked dead UI surface).
- Derived fields nextFireAt / nextFires / failureStreak populated at
create/update and refreshed after every terminal run; scheduler re-publishes
automation.definition.updated with a bumped revision for global-sync.
- cron validation consolidated into src/automation/cron.ts; scheduler and
derived both consume it.
- automate tool schema picks up model/variant + Provider-backed create-time
validation.
- Migration 20260601100000 drops pre-release rows so model can be NOT NULL.
Verification:
- CI green on 5e177ba (unit-opencode flake on an unrelated VCS-routes
20s timeout cleared on rerun).
- Local: 129 automation tests (routes/runner/scheduler/event-fixtures/tool/
cron) + 35 session processor tests pass.
Review follow-ups addressed:
- codex: scheduler self-loop guard keyed by id:revision (not id alone).
- P2: recordRunOutcome retries on ConflictError instead of silently dropping
the run outcome.
- CodeRabbit: 422 create test uses a guaranteed-invalid model; fixed sleeps
replaced with terminal-state polling.
- Test seam for the ConflictError path moved off the public Automation API
into an internal __test_hooks module.
Residual risk / deferred (tracked, not blocking PR6):
- needs_user_input / loop_gate run-error codes remain reserved; producing
them needs prompt-loop semantics changes.
- Session.automationID reverse lookup deferred (session-contract migration).
- stop=condition kept in the schema layer for SDK shape parity but rejected
at validate time; collapses once a condition evaluator lands.
Linked: issue #950; follows PR #960 #983 #984 #998 #1004; unblocks PR6/PR7.
1 participant
Summary
Adds the PR2 backend execution slice for issue #950: manual
automation.runNownow queues a run and starts background execution through the session prompt runner, with run-ledger updates for running, success, blocker, and hard-stop states.Why
PR1 froze the automation contract and in-memory store but intentionally left execution stubbed. This PR keeps the next slice narrow: it proves manual Run now execution while the user is present, adds the shared blocker seam, and leaves scheduler, persistence, frontend UI, and worktree placement for later PRs.
Related Issue
Closes part of #950. This is PR2 in the agreed 7-PR breakdown.
Human Review Status
Pending
Review Focus
Please focus on the
AutomationRunContextseam throughSessionPrompt, the narrowPermission.ask onPendinghook, and the run ledger transitions inpackages/opencode/src/automation/index.ts.Risk Notes
No scheduler, persistence, frontend, generated SDK, or worktree behavior is included. The run store remains in-memory from PR1. Visible UI/copy, platform/packaging, and docs/release-note surfaces were not touched.
How To Verify
Screenshots or Recordings
Not applicable; no visible UI changes.
Checklist
bug,enhancement,task,documentation. Type labels are author-added; the labeler bot does NOT assign them. Add the label in the GitHub UI, then tick this.app,ui,platform,harness,ci. The labeler bot assigns these on PR open based on changed paths. Confirm the bot's choice (or override if wrong), then tick this.P0,P1,P2,P3. The priority-triage bot suggests one on PR open. Confirm or override, then tick this.Pending,Approved by @<reviewer>, orNot required: <reason>(default isPending; "not required" is restricted to bot-authored low-risk PRs).dev, and my PR title and commit messages use Conventional Commits in English.Summary by CodeRabbit
New Features
Improvements
Tests