Handle websocket error gracefully + add retry logic

[jakubno]

Handle connection error between the server and kernel gracefully
Add reconnect logic

---

Note

Adds WebSocket reconnect with retries and ping timeout, conditionally sets auth header in Python clients, and surfaces connection-loss errors to prevent hangs.

• Server runtime (template/server/messaging.py):
  • WebSocket resilience: Add reconnect() with MAX_RECONNECT_RETRIES and PING_TIMEOUT; wrap send in retry loop catching ConnectionClosedError/WebSocketException and reconnecting.
  • Error propagation: On receiver teardown, enqueue Error and UnexpectedEndOfExecution for ongoing executions to avoid hangs.
  • Execution flow: Create per-execution message_id and Execution right before send; serialize request with _get_execute_request.
• Python client (python/e2b_code_interpreter/*):
  • Auth header handling: Set headers with X-Access-Token only if token present for both async and sync run_code and create_code_context.
• Changesets: Patch releases for @e2b/code-interpreter-template (add retry) and @e2b/code-interpreter-python (fix header issue).

^{Written by Cursor Bugbot for commit e51313b. This will update automatically on new commits. Configure here.}

[cursor]

Bug: Reconnect Hangs on Blocked Receive Task

The reconnect method awaits self._receive_task without cancelling it first. If the receive task is blocked waiting for WebSocket messages, this can cause the reconnect process to hang indefinitely, unlike the close() method's approach.

template/server/messaging.py#L70-L78

code-interpreter/template/server/messaging.py

Lines 70 to 78 in 8818ce7

	async def reconnect(self):
	if self._ws is not None:
	await self._ws.close(reason="Reconnecting")
	if self._receive_task is not None:
	await self._receive_task
	await self.connect()

 

---