Description:
I’m experiencing persistent HTTP 500 errors when using the chat/Cascade feature in Antigravity IDE. The issue is account-specific — switching to a different
account on the same machine resolves the problem immediately.
Environment:
- Antigravity IDE version: 0.2.0 (Windows 11)
- Region: China
Symptoms:
- Every chat request returns 500 Internal Server Error instantly (~86-95ms response time), suggesting the server rejects the request before any model
processing occurs. - The Antigravity Cockpit plugin’s “Wake Up” feature (which calls /v1internal:streamGenerateContent) works normally with the same account — only the
Cascade/Trajectory-based chat is broken. - Switching to a different account on the same IDE instance works fine.
- The issue persists across different workspaces/folders.
Troubleshooting already performed:
- Cleared local session/storage cache (Session Storage, Local Storage, workspaceStorage) — no effect
- Re-logged in / refreshed OAuth token — no effect
- Switched workspaces — no effect
- Confirmed other accounts work fine on the same machine
Error samples:
Request 1:
- Trajectory ID: 5d4258f8-f0a4-49d1-b500-a46a9f5bbe6d
- TraceID: 0xdaf381734fab082
- Time: 2026-05-09 08:21:30 UTC
Request 2:
- Trajectory ID: 3885ab3d-fdfd-4e04-82bd-830c2744
- TraceID: 0x74015d8ef98a2
- Time: 2026-05-09 08:55:02 UTC
Conclusion:
The issue appears to be server-side state corruption tied to this specific account’s Trajectory/Cascade processing pipeline. The account can successfully call
streamGenerateContent (via Cockpit), but all Cascade chat requests fail with 500 before reaching model inference. Please investigate and reset the
account-level Trajectory state if applicable.
