fix(ralph-loop): guard delayed start snapshots

Prevent delayed loop-start message counts from overwriting active Ralph Loop state after the loop has already advanced, so ULW completion can still enter Oracle verification instead of iterating forever.

Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>
This commit is contained in:
YeonGyu-Kim
2026-05-13 18:59:07 +09:00
parent 1fa97c6efe
commit 61ba4e3b41
3 changed files with 64 additions and 2 deletions
+45
View File
@@ -1291,6 +1291,51 @@ Original task: Build something`
expect(verificationToast!.message).toMatch(/Oracle verification is now required/)
})
test("#given loop-start message count resolves late after progress #when ulw DONE appears #then oracle verification still starts", async () => {
// given - the initial message-count request is delayed past the first continuation
let messageCallCount = 0
let resolveInitialMessages: ((value: { data: typeof mockSessionMessages }) => void) | undefined
const delayedMock = createMockPluginInput()
Object.defineProperty(delayedMock.client.session, "messages", {
value: async (opts: { path: { id: string } }) => {
messagesCalls.push({ sessionID: opts.path.id })
messageCallCount += 1
if (messageCallCount === 1) {
return new Promise<{ data: typeof mockSessionMessages }>((resolve) => {
resolveInitialMessages = resolve
})
}
return { data: mockSessionMessages }
},
})
const hook = createRalphLoopHook(delayedMock, {
getTranscriptPath: () => join(TEST_DIR, "missing-transcript.jsonl"),
idleSettleMs: 0,
})
hook.startLoop("session-123", "Build API", { ultrawork: true })
await hook.event({ event: { type: "session.idle", properties: { sessionID: "session-123" } } })
expect(hook.getState()?.iteration).toBe(2)
mockSessionMessages = [
{
info: { role: "assistant" },
parts: [{ type: "text", text: "All work is complete. <promise>DONE</promise>" }],
},
]
// when - delayed start snapshot resolves after the loop has already advanced
resolveInitialMessages?.({ data: mockSessionMessages })
await new Promise((resolve) => setTimeout(resolve, 0))
await hook.event({ event: { type: "session.idle", properties: { sessionID: "session-123" } } })
// then - the late snapshot must not hide the DONE message from verification gating
expect(hook.getState()?.verification_pending).toBe(true)
expect(hook.getState()?.completion_promise).toBe("VERIFIED")
expect(promptCalls[promptCalls.length - 1]?.text).toContain('task(subagent_type="oracle"')
})
test("should show regular completion toast when ultrawork disabled", async () => {
// given - hook without ultrawork
const transcriptPath = join(TEST_DIR, "transcript.jsonl")
+13 -1
View File
@@ -104,11 +104,23 @@ export function createLoopStateController(options: {
return state
},
setMessageCountAtStart(sessionID: string, messageCountAtStart: number): RalphLoopState | null {
setMessageCountAtStart(
sessionID: string,
messageCountAtStart: number,
expectedStartedAt?: string,
): RalphLoopState | null {
const state = readState(directory, stateDir)
if (!state || state.session_id !== sessionID) {
return null
}
if (
state.iteration !== 1
|| state.verification_pending
|| state.message_count_at_start !== undefined
|| (expectedStartedAt !== undefined && state.started_at !== expectedStartedAt)
) {
return null
}
state.message_count_at_start = messageCountAtStart
if (!writeState(directory, state, stateDir)) {
+6 -1
View File
@@ -73,6 +73,11 @@ export function createRalphLoopHook(
return startSuccess
}
const startedState = loopState.getState()
const expectedStartedAt = startedState?.session_id === sessionID
? startedState.started_at
: undefined
ctx.client.session
.messages({
path: { id: sessionID },
@@ -80,7 +85,7 @@ export function createRalphLoopHook(
})
.then((messagesResponse: unknown) => {
const messageCountAtStart = getMessageCountFromResponse(messagesResponse)
loopState.setMessageCountAtStart(sessionID, messageCountAtStart)
loopState.setMessageCountAtStart(sessionID, messageCountAtStart, expectedStartedAt)
})
.catch(() => {})