Ai

Daytona 0.182.0 Adds Sandbox Timeout, GPU Fixes for AI Workflows

Daytona 0.182.0 Adds Sandbox Timeout, GPU Fixes for AI Workflows

Daytona just dropped version 0.182.0, and it's a small but mighty update for anyone running AI workloads on the platform. The headline is a new 30-second timeout for sandbox creation, aimed at preventing runaway provisioning. But there's also some important housekeeping around GPU capacity and filtering that keeps your compute costs in check.

What Changed

Sandbox creation gets a hard timeout

Developers can now set a timeout of 30 seconds for sandbox creation via the API. That's new in this release. Previously, there was no default cap – a stuck provisioning call could hang indefinitely. Now the API will automatically fail after half a minute, forcing a retry or a fallback. It's a simple change, but it saves hours of debugging silent failures in CI/CD pipelines.

GPU accounting gets sharper

Two fixes target GPU resource management. The first (#4809) excludes sandboxes in DESTROYED or ARCHIVED states from the GPU capacity count. That means your available GPUs won't be artificially inflated by zombie resources. The second (#4815) enforces the GPU filter during snapshot-runner fallback assignment. If you've specified a GPU type, the system won't silently drop it when falling back to a different runner. Both are subtle but crucial for predictable GPU allocation in training jobs.

Docs expanded for sandbox operations

The documentation now covers sandbox get, labels, and last activity details (#4732). Nothing groundbreaking, but it fills a gap for users who rely on programmatic sandbox management. You can finally query label-based metadata and check the last activity timestamp without guessing.

Why It Matters

This release is about hardening Daytona for production AI workloads. The timeout prevents one hung request from blocking an entire pipeline. It's a developer experience improvement that seasoned engineers will appreciate. The GPU fixes ensure your expensive compute resources aren't wasted or misallocated.

Is it a game-changer? No. But it's the kind of polish that separates a hobby tool from a serious platform. If you're running LLM fine-tuning or batch inference on Daytona, upgrading is a no-brainer. The docs update just makes life easier for everyone else.

Official Source: https://github.com/daytonaio/daytona/releases/tag/v0.182.0