July 16, 2026
Pod web apps that use URL-encoded file paths — like AI Toolkit dashboards — now work correctly, plus edge security updates and a landing-page cost calculator.
A focused fix release for web apps running inside Pods, along with security updates to the platform’s edge.
URL-encoded paths in Pod web apps now work
Section titled “URL-encoded paths in Pod web apps now work”Some web applications served from Pods encode full file paths into their URLs — a common pattern in dataset browsers and training dashboards such as Ostris AI Toolkit. Requests like these were previously rejected at the platform edge with 400 Bad Request, so dataset image previews would not load and trained model files (e.g. .safetensors) could not be downloaded from the app’s own UI.
The platform’s routing layer now accepts URL-encoded paths (%2F and similar) end to end, so these tools work out of the box. The fix applies to every Pod endpoint — no changes are needed on your side.
Improvements
Section titled “Improvements”- The routing layer that fronts all Pod endpoints and APIs received a security-focused upgrade, picking up fixes for multiple published vulnerabilities. Rollouts were performed node by node, so no downtime was involved.
- The meshive.ai landing page now shows the live GPU catalog with an interactive cost calculator, so you can estimate the price of a configuration before signing in.