AsyncSandbox.run_code output handlers are never awaited

I pass an async callback to on_error, and get the following warning:

ERROR /app/autogpt_platform/backend/.venv/lib/python3.13/site-packages/e2b_code_interpreter/models.py:444: RuntimeWarning: coroutine 'AsyncSandbox.kill' was never awaited
ERROR   on_error(execution.error)
ERROR RuntimeWarning: Enable tracemalloc to get the object allocation traceback

This is because on_error, on_result etc are called but never actually awaited in parse_output:

code-interpreter/python/e2b_code_interpreter/models.py

Lines 417 to 446 in b1b4af5

    
           def parse_output( 
        
               execution: Execution, 
        
               output: str, 
        
               on_stdout: Optional[OutputHandler[OutputMessage]] = None, 
        
               on_stderr: Optional[OutputHandler[OutputMessage]] = None, 
        
               on_result: Optional[OutputHandler[Result]] = None, 
        
               on_error: Optional[OutputHandler[ExecutionError]] = None, 
        
           ): 
        
               data = json.loads(output) 
        
               data_type = data.pop("type") 
        
               if data_type == "result": 
        
                   result = Result(**data) 
        
                   execution.results.append(result) 
        
                   if on_result: 
        
                       on_result(result) 
        
               elif data_type == "stdout": 
        
                   execution.logs.stdout.append(data["text"]) 
        
                   if on_stdout: 
        
                       on_stdout(OutputMessage(data["text"], data["timestamp"], False)) 
        
               elif data_type == "stderr": 
        
                   execution.logs.stderr.append(data["text"]) 
        
                   if on_stderr: 
        
                       on_stderr(OutputMessage(data["text"], data["timestamp"], True)) 
        
               elif data_type == "error": 
        
                   execution.error = ExecutionError(data["name"], data["value"], data["traceback"]) 
        
                   if on_error: 
        
                       on_error(execution.error) 
        
               elif data_type == "number_of_executions": 
        
                   execution.execution_count = data["execution_count"]

ENG-3217

Hey, can you provide us a sample reproduction code where the issue surfaces? I need bit more context here.

This is where it occurs for me: https://github.com/Significant-Gravitas/AutoGPT/blob/e06e7ff33f20a8ec6990d44a7889cc9ff0720670/autogpt_platform%2Fbackend%2Fbackend%2Fblocks%2Fcode_executor.py#L116-L120

sandbox.kill() returns a promise, which is awaitable and allowed by the type of on_error, but like I mentioned parse_output doesn't await the result if it's awaitable, causing that warning and keeping the callback from running properly.

Thanks for reporting, fix in #160

This should be fixed in https://github.com/e2b-dev/code-interpreter/releases/tag/%40e2b%2Fcode-interpreter-python%402.2.1

If you still see the error, feel free to reopen the issue.

Navigation Menu

Repository navigation

`AsyncSandbox.run_code` output handlers are never awaited #159

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

	def parse_output(
	execution: Execution,
	output: str,
	on_stdout: Optional[OutputHandler[OutputMessage]] = None,
	on_stderr: Optional[OutputHandler[OutputMessage]] = None,
	on_result: Optional[OutputHandler[Result]] = None,
	on_error: Optional[OutputHandler[ExecutionError]] = None,
	):
	data = json.loads(output)
	data_type = data.pop("type")

	if data_type == "result":
	result = Result(**data)
	execution.results.append(result)
	if on_result:
	on_result(result)
	elif data_type == "stdout":
	execution.logs.stdout.append(data["text"])
	if on_stdout:
	on_stdout(OutputMessage(data["text"], data["timestamp"], False))
	elif data_type == "stderr":
	execution.logs.stderr.append(data["text"])
	if on_stderr:
	on_stderr(OutputMessage(data["text"], data["timestamp"], True))
	elif data_type == "error":
	execution.error = ExecutionError(data["name"], data["value"], data["traceback"])
	if on_error:
	on_error(execution.error)
	elif data_type == "number_of_executions":
	execution.execution_count = data["execution_count"]

AsyncSandbox.run_code output handlers are never awaited #159

Description

Activity

linear commented on Oct 25, 2025

mishushakov commented on Oct 25, 2025

Pwuts commented on Oct 26, 2025

jakubno commented on Oct 27, 2025

jakubno commented on Oct 27, 2025

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

`AsyncSandbox.run_code` output handlers are never awaited #159