feat: add sandbox async methods

[luke-lombardi]

Summary by cubic

Add an async API for the sandbox via .aio interfaces, enabling non-blocking process, filesystem, Docker, and lifecycle operations with async streaming of stdout/stderr/logs. Adds retries for transient connection errors and caches stdout/stderr/log streams for consistent reads, while maintaining full backward compatibility with the sync API.

• New Features

  • .aio properties for SandboxInstance, ProcessManager, Process, FileSystem, and DockerManager.
  • Async classes: AsyncSandboxInstance/Process/ProcessManager/FileSystem/DockerManager/DockerResult/DockerComposeStack, plus async stdout/stderr/log streams.
  • Async methods mirror the sync API (run_code, exec, list_processes, upload/download, expose_port, update_ttl, update_network_permissions, list_urls, Docker run/build/pull/push/compose).
  • Uses asyncio.to_thread under the hood; no backend changes required.

• Refactors

  • Added retry_on_transient_error for exec/status/stream reads to handle transient gRPC/connection errors.
  • Cache stdout/stderr/logs streams to preserve buffer state across accesses.
  • Exclude _aio and cached stream attributes from getstate to keep objects picklable; minor import/order cleanups.

^{Written for commit 75db4cc. Summary will update on new commits.}

[cubic-dev-ai]

1 issue found across 1 file

Prompt for AI agents (all issues)

Check if these issues are valid — if so, understand the root cause of each and fix them.


<file name="sdk/src/beta9/abstractions/sandbox.py">

<violation number="1" location="sdk/src/beta9/abstractions/sandbox.py:3430">
P2: The `logs()` async generator buffers all output before yielding, defeating the streaming purpose. It waits for `list(self._sync.logs())` to complete (collecting ALL logs) before yielding any lines. For long-running operations like Docker builds, users won't see any output until the entire operation completes.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}

[cubic-dev-ai]

2 issues found across 3 files (changes from recent commits).

Prompt for AI agents (all issues)

Check if these issues are valid — if so, understand the root cause of each and fix them.


<file name="sdk/src/beta9/utils.py">

<violation number="1" location="sdk/src/beta9/utils.py:40">
P3: The comment says "Exponential backoff" but the formula `delay * (attempt + 1)` produces linear backoff. Either update the comment to "linear backoff" or change the formula to `delay * (2 ** attempt)` for true exponential backoff.</violation>

<violation number="2" location="sdk/src/beta9/utils.py:44">
P2: If `max_retries=0` is passed, the loop never executes and `raise last_exception` will raise `None`, causing a `TypeError`. Consider adding a guard or raising a more descriptive error.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}

[cubic-dev-ai]

P2: If max_retries=0 is passed, the loop never executes and raise last_exception will raise None, causing a TypeError. Consider adding a guard or raising a more descriptive error.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At sdk/src/beta9/utils.py, line 44:

<comment>If `max_retries=0` is passed, the loop never executes and `raise last_exception` will raise `None`, causing a `TypeError`. Consider adding a guard or raising a more descriptive error.</comment>

<file context>
@@ -2,11 +2,48 @@
+                    continue
+            # Non-transient error, raise immediately
+            raise
+    raise last_exception
+
+
</file context>

Suggested change

	raise last_exception
	if last_exception is None:
	raise ValueError("max_retries must be at least 1")
	raise last_exception

[cubic-dev-ai]

P3: The comment says "Exponential backoff" but the formula delay * (attempt + 1) produces linear backoff. Either update the comment to "linear backoff" or change the formula to delay * (2 ** attempt) for true exponential backoff.

Prompt for AI agents

Check if this issue is valid — if so, understand the root cause and fix it. At sdk/src/beta9/utils.py, line 40:

<comment>The comment says "Exponential backoff" but the formula `delay * (attempt + 1)` produces linear backoff. Either update the comment to "linear backoff" or change the formula to `delay * (2 ** attempt)` for true exponential backoff.</comment>

<file context>
@@ -2,11 +2,48 @@
+            ):
+                last_exception = e
+                if attempt < max_retries - 1:
+                    time.sleep(delay * (attempt + 1))  # Exponential backoff
+                    continue
+            # Non-transient error, raise immediately
</file context>

Suggested change

	time.sleep(delay * (attempt + 1)) # Exponential backoff
	time.sleep(delay * (2 ** attempt)) # Exponential backoff