Skip to content
Source
Sign inSign up

pingdotgg

t3codePublic Synced with GitHub
Popular #1 Fork 0

[Bug]: Antigravity sent nothing for 15 minutes, so T3 Code stopped the turn and restarted the agent. Send a message to continue. #34160

Send to my agent
Openyiyun-li(GitHub profile, opens in a new tab) opened this issue 1 comment
yiyun-li(GitHub profile, opens in a new tab)commented

Before submitting

  • I searched existing issues and did not find a duplicate.
  • I included enough detail to reproduce or investigate the problem.

Area

apps/server

Steps to reproduce

In new version, Antigravity sent nothing for 15 minutes, so T3 Code stopped the turn and restarted the agent. Send a message to continue.

Expected behavior

Solve it plz

Actual behavior

Antigravity sent nothing for 15 minutes, so T3 Code stopped the turn and restarted the agent. Send a message to continue.

Impact

Blocks work completely

Version or commit

No response

Environment

No response

Logs or stack traces

Screenshots, recordings, or supporting files

No response

Workaround

No response

  1. yiyun-li(GitHub profile, opens in a new tab) added labelsbug

  2. yiyun-li(GitHub profile, opens in a new tab) added labelsneeds-triage

  3. mr-karan(GitHub profile, opens in a new tab)commented

    Another occurrence on Linux x86_64, T3 Code 0.1.9-nightly.20261111.4515, Antigravity model ID gemini-4-carbon (Carbon 4 High).

    The thread activity and provider request log show this timeline on 2026-11-12, all times UTC:

    • 06:10:49.997: session/prompt dispatched.
    • 06:22:13.840: last visible completed file-read tool result.
    • 06:37:13.843: the outstanding session/prompt request terminates in the provider request log.
    • 06:37:13.848: the thread shows the exact 15-minute inactivity error quoted in this issue.
    • 06:57:37: I manually sent a message to resume.
    • 06:58:56 onward: the resumed turn produced more file-read results.

    This confirms the watchdog fired 15 minutes after the last visible tool result and that manual resume allowed further activity. It does not establish whether the original prompt was still computing or had silently failed. No upstream 502 has been confirmed for this occurrence, and there is no deterministic reproduction yet.

    Related investigation: #33583. The proposed fix in #33596 is still open and improves reporting of errors recorded in Antigravity's session log when the watchdog fires; its description explicitly retains the existing timeout. That may help diagnose this occurrence, but it should not be assumed to address legitimate long silent computation without checking the native turn outcome.

    The logs and review content are private, so this report includes only request metadata and timestamps.

Sign in to join this conversation.