test(agent): cover background wake-up forwarding between turns#3200
Merged
charlesvien merged 1 commit intoJul 6, 2026
Merged
Conversation
Member
Author
|
Warning This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
This stack of pull requests is managed by Graphite. Learn more about stacking. |
2 tasks
|
React Doctor found no issues in the changed files. 🎉 Reviewed by React Doctor for commit |
charlesvien
changed the title
Contributor
|
Reviews (1): Last reviewed commit: "add wake-up forwarding regression tests" | Re-trigger Greptile |
![greptile-apps[bot]](https://avatars.githubusercontent.com/in/867647?s=60&v=4){kind=small}
Comment on lines
+28
to
+200
|
|
||
| interface ClientMocks { | ||
| sessionUpdate: ReturnType<typeof vi.fn>; | ||
| extNotification: ReturnType<typeof vi.fn>; | ||
| } | ||
|
|
||
| function makeAgent(): { agent: Agent; client: ClientMocks } { | ||
| const client: ClientMocks = { | ||
| sessionUpdate: vi.fn().mockResolvedValue(undefined), | ||
| extNotification: vi.fn().mockResolvedValue(undefined), | ||
| }; | ||
| const agent = new ClaudeAcpAgent(client as unknown as AgentSideConnection); | ||
| return { agent, client }; | ||
| } | ||
|
|
||
| function installFakeSession( | ||
| agent: Agent, | ||
| sessionId: string, | ||
| ): { query: MockQuery; input: Pushable<SDKUserMessage> } { | ||
| const query = createMockQuery(); | ||
| const input = new Pushable<SDKUserMessage>(); | ||
| const abortController = new AbortController(); | ||
|
|
||
| const session = { | ||
| query, | ||
| queryOptions: { sessionId, cwd: "/tmp/repo", abortController }, | ||
| buildInProcessMcpServers: () => ({}), | ||
| localToolsServerNames: [] as string[], | ||
| input, | ||
| cancelled: false, | ||
| interruptReason: undefined, | ||
| settingsManager: { dispose: vi.fn(), getRepoRoot: () => "/tmp/repo" }, | ||
| permissionMode: "default" as const, | ||
| abortController, | ||
| accumulatedUsage: { | ||
| inputTokens: 0, | ||
| outputTokens: 0, | ||
| cachedReadTokens: 0, | ||
| cachedWriteTokens: 0, | ||
| }, | ||
| sessionResources: new Set(), | ||
| configOptions: [], | ||
| turnQueue: [], | ||
| activeTurn: null, | ||
| pendingOrphanResults: 0, | ||
| queryGeneration: 0, | ||
| cwd: "/tmp/repo", | ||
| notificationHistory: [] as unknown[], | ||
| taskRunId: "run-1", | ||
| lastContextWindowSize: 200_000, | ||
| modelId: "claude-sonnet-4-6", | ||
| taskState: new Map(), | ||
| }; | ||
|
|
||
| (agent as unknown as { session: typeof session }).session = session; | ||
| (agent as unknown as { sessionId: string }).sessionId = sessionId; | ||
|
|
||
| return { query, input }; | ||
| } | ||
|
|
||
| function tick(): Promise<void> { | ||
| return new Promise((resolve) => setImmediate(resolve)); | ||
| } | ||
|
|
||
| async function send(query: MockQuery, message: unknown): Promise<void> { | ||
| query._mockHelpers.sendMessage(message as SDKMessage); | ||
| await tick(); | ||
| } | ||
|
|
||
| // Replays the prompt's own user message back through the query so the consumer | ||
| // activates its Turn before the terminal result arrives. | ||
| async function echoUserMessage( | ||
| query: MockQuery, | ||
| input: Pushable<SDKUserMessage>, | ||
| ): Promise<void> { | ||
| const { value: pushed } = await input[Symbol.asyncIterator]().next(); | ||
| await send(query, pushed); | ||
| } | ||
|
|
||
| function messageStart(sessionId: string, apiId: string) { | ||
| return { | ||
| type: "stream_event", | ||
| parent_tool_use_id: null, | ||
| session_id: sessionId, | ||
| uuid: `start-${apiId}`, | ||
| event: { type: "message_start", message: { id: apiId, usage: {} } }, | ||
| }; | ||
| } | ||
|
|
||
| function textDelta(sessionId: string, text: string) { | ||
| return { | ||
| type: "stream_event", | ||
| parent_tool_use_id: null, | ||
| session_id: sessionId, | ||
| uuid: `delta-${text}`, | ||
| event: { | ||
| type: "content_block_delta", | ||
| index: 0, | ||
| delta: { type: "text_delta", text }, | ||
| }, | ||
| }; | ||
| } | ||
|
|
||
| function assistantMessage(sessionId: string, apiId: string, text: string) { | ||
| return { | ||
| type: "assistant", | ||
| parent_tool_use_id: null, | ||
| session_id: sessionId, | ||
| uuid: `assistant-${apiId}`, | ||
| message: { | ||
| id: apiId, | ||
| role: "assistant", | ||
| content: [{ type: "text", text }], | ||
| }, | ||
| }; | ||
| } | ||
|
|
||
| function resultSuccess(sessionId: string, uuid: string) { | ||
| return { | ||
| type: "result", | ||
| subtype: "success", | ||
| session_id: sessionId, | ||
| uuid, | ||
| result: "", | ||
| is_error: false, | ||
| usage: {}, | ||
| modelUsage: {}, | ||
| }; | ||
| } | ||
|
|
||
| function taskNotificationResult(sessionId: string, uuid: string) { | ||
| return { | ||
| ...resultSuccess(sessionId, uuid), | ||
| origin: { kind: "task-notification" }, | ||
| }; | ||
| } | ||
|
|
||
| function taskNotification(sessionId: string) { | ||
| return { | ||
| type: "system", | ||
| subtype: "task_notification", | ||
| session_id: sessionId, | ||
| uuid: "task-note-1", | ||
| task_id: "task-1", | ||
| status: "completed", | ||
| output_file: "/tmp/task-1.out", | ||
| summary: "Background command completed", | ||
| }; | ||
| } | ||
|
|
||
| function messageChunkTexts( | ||
| calls: ClientMocks["sessionUpdate"]["mock"]["calls"], | ||
| ): string[] { | ||
| return calls | ||
| .map( | ||
| ([call]) => | ||
| ( | ||
| call as { | ||
| update?: { sessionUpdate?: string; content?: { text?: string } }; | ||
| } | ||
| ).update, | ||
| ) | ||
| .filter((update) => update?.sessionUpdate === "agent_message_chunk") | ||
| .map((update) => update?.content?.text ?? ""); | ||
| } | ||
|
|
||
| async function runTurn( | ||
| agent: Agent, | ||
| client: ClientMocks, | ||
| query: MockQuery, | ||
| input: Pushable<SDKUserMessage>, | ||
| sessionId: string, | ||
| text: string, |
Contributor
There was a problem hiding this comment.
ClientMocks, makeAgent, installFakeSession, tick, send, echoUserMessage, messageStart, textDelta, assistantMessage, and messageChunkTexts are copied character-for-character from claude-agent.streamed-text.test.ts. A third test file for this adapter will copy them again. Extracting these into a shared module (e.g., claude-agent.test-helpers.ts) would satisfy the OnceAndOnlyOnce rule and keep all adapter test files in sync automatically.
Context Used: Do not attempt to comment on incorrect alphabetica... (source)
Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!
charlesvien
merged commit 8f12c5e
into
chore/sync-claude-adapter-v0.54.1
16 of 17 checks passed
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
1 participant
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.
Problem
Agent wake-ups (Bash
run_in_backgroundtask notifications, Monitor andScheduleWakeuptimers) fire after the turn has settled. Onmainthe SDK stream is only read insideprompt(), so wake-up output sits unread until the next user message and then appears all at once. The session looks hung for anyone waiting on a background task or a scheduled wake-up.Closes #1919
Closes #2414
Changes
The fix itself lands in #3076 below this PR: its upstream #780 port replaces the per-prompt read loop with a session-lifetime consumer that forwards between-turn output live. This PR stacks regression tests on top so that behavior can't silently regress:
How did you test this?
pnpm --filter @posthog/agent test: 931 tests pass, including the 4 new onessleep 10 && echo finishedin the background, turn ended withend_turn, "the test succeeded" streamed in live 9s later with no follow-up promptScheduleWakeUptool call properly #2414: agent calledScheduleWakeupfor 60s out, turn ended withend_turn, "the wakeup succeeded" streamed in live when the timer fired with no follow-up promptAutomatic notifications