Execution failed: Error: Failed to get response from the AI model; retried 5 times ... Last error: Unknown error

[njdldkl666699]

Describe the bug

When a session is very very long, at some point the copilot raises an error. The error log in events.jsonl says:

{
    "type": "session.error",
    "data": {
        "errorType": "query",
        "message": "Execution failed: Error: Failed to get response from the AI model; retried 5 times (total retry wait time: 6.05 seconds) Last error: Unknown error",
        "stack": "Error: Failed to get response from the AI model; retried 5 times (total retry wait time: 6.05 seconds) Last error: Unknown error\n    at zht.getCompletionWithTools (file:///home/cola/.cache/copilot/pkg/linux-x64/1.0.36-0/app.js:1992:18607)\n    at async xMe.getCompletionWithTools (file:///home/cola/.cache/copilot/pkg/linux-x64/1.0.36-0/app.js:2010:3387)\n    at async t.runAgenticLoop (file:///home/cola/.cache/copilot/pkg/linux-x64/1.0.36-0/app.js:2625:3830)\n    at async t.processQueuedItems (file:///home/cola/.cache/copilot/pkg/linux-x64/1.0.36-0/app.js:2623:9584)\n    at async t.processQueue (file:///home/cola/.cache/copilot/pkg/linux-x64/1.0.36-0/app.js:2623:3027)\n    at async t.send (file:///home/cola/.cache/copilot/pkg/linux-x64/1.0.36-0/app.js:2621:3954)"
    },
    "id": "8d72fccd-ee18-4444-8a78-f7d7f336982b",
    "timestamp": "2026-05-30T14:02:04.783Z",
    "parentId": "678ae21c-40d6-4d71-a30f-b7a8dc873755"
}

Stack:

Error: Failed to get response from the AI model; retried 5 times (total retry wait time: 6.05 seconds) Last error: Unknown error
    at zht.getCompletionWithTools (file:///home/cola/.cache/copilot/pkg/linux-x64/1.0.36-0/app.js:1992:18607)
    at async xMe.getCompletionWithTools (file:///home/cola/.cache/copilot/pkg/linux-x64/1.0.36-0/app.js:2010:3387)
    at async t.runAgenticLoop (file:///home/cola/.cache/copilot/pkg/linux-x64/1.0.36-0/app.js:2625:3830)
    at async t.processQueuedItems (file:///home/cola/.cache/copilot/pkg/linux-x64/1.0.36-0/app.js:2623:9584)
    at async t.processQueue (file:///home/cola/.cache/copilot/pkg/linux-x64/1.0.36-0/app.js:2623:3027)
    at async t.send (file:///home/cola/.cache/copilot/pkg/linux-x64/1.0.36-0/app.js:2621:3954)

And as far as I know, there is some mechanism to compact the session automatically, and this is enabled by default. But it doesn't help in this situation.

Affected version

1.0.56

Steps to reproduce the behavior

1. Open a new session
2. Send may may much messages (text, images), interact with copilot in a session may times
3. After enough message history, the copilot raises an error.

Expected behavior

No matter how long the session is, the copilot should never fails to send request.

Additional context

OS: Ubuntu 22.04
CPU: x86_64