Fix frozen command_id default causing concurrent execution stdout corruption (#757)

Author: dines-rl

src/runloop_api_client/resources/devboxes/devboxes.py

@@ -867,7 +867,7 @@ def execute(
         id: str,
         *,
         command: str,
-        command_id: str = str(uuid7()),
+        command_id: str | None = None,
         last_n: str | Omit = omit,
         optimistic_timeout: Optional[int] | Omit = omit,
         shell_name: Optional[str] | Omit = omit,
@@ -892,7 +892,8 @@ def execute(
               specified the command is run from the directory based on the recent state of the
               persistent shell.
 
-          command_id: The command ID in UUIDv7 string format for idempotency and tracking
+          command_id: The command ID in UUIDv7 string format for idempotency and tracking.
+              A fresh UUID is generated per call if not provided.
 
           last_n: Last n lines of standard error / standard out to return (default: 100)
 
@@ -915,6 +916,8 @@ def execute(
         """
         if not id:
             raise ValueError(f"Expected a non-empty value for `id` but received {id!r}")
+        if command_id is None:
+            command_id = str(uuid7())
         if not is_given(timeout) and self._client.timeout == DEFAULT_TIMEOUT:
             timeout = 600
         return self._post(
@@ -944,7 +947,7 @@ def execute_and_await_completion(
         devbox_id: str,
         *,
         command: str,
-        command_id: str = str(uuid7()),
+        command_id: str | None = None,
         last_n: str | Omit = omit,
         optimistic_timeout: Optional[int] | Omit = omit,
         shell_name: Optional[str] | Omit = omit,
@@ -963,9 +966,11 @@ def execute_and_await_completion(
         return the result within the initial request's timeout. If the execution is not yet
         complete, it switches to using wait_for_command to minimize latency while waiting.
 
-        A command_id (UUIDv7) is automatically generated for idempotency and tracking.
+        A command_id (UUIDv7) is automatically generated per call for idempotency and tracking.
         You can provide your own command_id to enable custom retry logic or external tracking.
         """
+        if command_id is None:
+            command_id = str(uuid7())
         execution = self.execute(
             devbox_id,
             command=command,
@@ -2543,7 +2548,7 @@ async def execute(
         id: str,
         *,
         command: str,
-        command_id: str = str(uuid7()),
+        command_id: str | None = None,
         last_n: str | Omit = omit,
         optimistic_timeout: Optional[int] | Omit = omit,
         shell_name: Optional[str] | Omit = omit,
@@ -2568,7 +2573,8 @@ async def execute(
               specified the command is run from the directory based on the recent state of the
               persistent shell.
 
-          command_id: The command ID in UUIDv7 string format for idempotency and tracking
+          command_id: The command ID in UUIDv7 string format for idempotency and tracking.
+              A fresh UUID is generated per call if not provided.
 
           last_n: Last n lines of standard error / standard out to return (default: 100)
 
@@ -2591,6 +2597,8 @@ async def execute(
         """
         if not id:
             raise ValueError(f"Expected a non-empty value for `id` but received {id!r}")
+        if command_id is None:
+            command_id = str(uuid7())
         if not is_given(timeout) and self._client.timeout == DEFAULT_TIMEOUT:
             timeout = 600
         return await self._post(
@@ -2620,7 +2628,7 @@ async def execute_and_await_completion(
         devbox_id: str,
         *,
         command: str,
-        command_id: str = str(uuid7()),
+        command_id: str | None = None,
         last_n: str | Omit = omit,
         optimistic_timeout: Optional[int] | Omit = omit,
         shell_name: Optional[str] | Omit = omit,
@@ -2639,7 +2647,7 @@ async def execute_and_await_completion(
         return the result within the initial request's timeout. If the execution is not yet
         complete, it switches to using wait_for_command to minimize latency while waiting.
 
-        A command_id (UUIDv7) is automatically generated for idempotency and tracking.
+        A command_id (UUIDv7) is automatically generated per call for idempotency and tracking.
         You can provide your own command_id to enable custom retry logic or external tracking.
         """
 

tests/test_command_id.py

@@ -0,0 +1,94 @@
+"""Tests for command_id default generation in execute / execute_and_await_completion.
+
+Verifies that each call generates a fresh UUIDv7 rather than reusing a frozen
+default (the bug fixed in this change).
+"""
+
+from __future__ import annotations
+
+import json
+from typing import cast
+
+import httpx
+import pytest
+from respx import Route, MockRouter
+
+from runloop_api_client import Runloop, AsyncRunloop
+
+base_url = "http://127.0.0.1:4010"
+EXECUTE_PATH = "/v1/devboxes/dbx_test/execute"
+
+STUB_RESPONSE = {
+    "execution_id": "exec_1",
+    "command_id": "ignored",
+    "devbox_id": "dbx_test",
+    "status": "completed",
+    "exit_status": 0,
+    "stdout": "",
+    "stderr": "",
+}
+
+
+def _get_command_ids(route: Route) -> list[str]:
+    return [
+        json.loads(cast(bytes, call.request.content))["command_id"]  # type: ignore[union-attr]
+        for call in route.calls  # type: ignore[union-attr]
+    ]
+
+
+def _get_request_body(route: Route, index: int = 0) -> dict[str, object]:
+    return json.loads(cast(bytes, route.calls[index].request.content))  # type: ignore[union-attr]
+
+
+class TestCommandIdGeneration:
+    """command_id must be a fresh UUIDv7 per call when not explicitly provided."""
+
+    @pytest.mark.respx(base_url=base_url)
+    def test_execute_generates_unique_command_ids(self, respx_mock: MockRouter) -> None:
+        route = respx_mock.post(EXECUTE_PATH).mock(return_value=httpx.Response(200, json=STUB_RESPONSE))
+        client = Runloop(base_url=base_url, bearer_token="test")
+
+        for _ in range(5):
+            client.devboxes.execute(id="dbx_test", command="echo hi")
+
+        assert route.call_count == 5
+        ids = _get_command_ids(route)
+        assert len(set(ids)) == 5, f"command_ids should all be unique, got: {ids}"
+
+    @pytest.mark.respx(base_url=base_url)
+    def test_execute_preserves_explicit_command_id(self, respx_mock: MockRouter) -> None:
+        route = respx_mock.post(EXECUTE_PATH).mock(return_value=httpx.Response(200, json=STUB_RESPONSE))
+        client = Runloop(base_url=base_url, bearer_token="test")
+
+        client.devboxes.execute(id="dbx_test", command="echo hi", command_id="my-custom-id")
+
+        body = _get_request_body(route)
+        assert body["command_id"] == "my-custom-id"
+
+
+class TestAsyncCommandIdGeneration:
+    """Async variant: command_id must be a fresh UUIDv7 per call when not explicitly provided."""
+
+    @pytest.mark.respx(base_url=base_url)
+    @pytest.mark.asyncio
+    async def test_execute_generates_unique_command_ids(self, respx_mock: MockRouter) -> None:
+        route = respx_mock.post(EXECUTE_PATH).mock(return_value=httpx.Response(200, json=STUB_RESPONSE))
+        client = AsyncRunloop(base_url=base_url, bearer_token="test")
+
+        for _ in range(5):
+            await client.devboxes.execute(id="dbx_test", command="echo hi")
+
+        assert route.call_count == 5
+        ids = _get_command_ids(route)
+        assert len(set(ids)) == 5, f"command_ids should all be unique, got: {ids}"
+
+    @pytest.mark.respx(base_url=base_url)
+    @pytest.mark.asyncio
+    async def test_execute_preserves_explicit_command_id(self, respx_mock: MockRouter) -> None:
+        route = respx_mock.post(EXECUTE_PATH).mock(return_value=httpx.Response(200, json=STUB_RESPONSE))
+        client = AsyncRunloop(base_url=base_url, bearer_token="test")
+
+        await client.devboxes.execute(id="dbx_test", command="echo hi", command_id="my-custom-id")
+
+        body = _get_request_body(route)
+        assert body["command_id"] == "my-custom-id"