Changelogs

​

Added

• Accelerator reporting reaches self-hosted installations. The backend moves from v0.2.0 to v0.3.1 and the dashboard from v0.2.0 to v0.3.0, which together bring per-card utilization, memory, power and temperature, partitioned cards reported per partition, idle card detection, and model serving and gateway reporting with cost per million tokens
• Accelerator charts on virtual machine pages, not only Kubernetes nodes

​

Fixed

• Accelerator charts were empty on every virtual machine page. The charts rendered as though the card were idle
• Accelerator charts also work for virtual machines that are not Kubernetes nodes

​

Changed

• Chart 0.5.1 changes no other image, value or object

​

Added

• Cloud and SaaS spend that CostGraph manages but does not monitor as a compute resource (object storage, managed databases, egress, ingested provider bills) now appears on your invoice as CostGraph Cloud, itemized per provider
• Invoice lines for spend billed in a currency other than USD now show the amount in that currency alongside the charge
• Self-hosted installations now report their own spend, so usage is billed from what the instance measures rather than from resource counts alone
• The cost view now defaults to the last 30 days
• Network edge properties are recorded on aggregated traffic

​

Changed

• The $500 free threshold is measured across everything under management, hosted and self-hosted together, rather than each separately

​

Fixed

• Security fixes in bundled dependencies

​

Fixed

• The backend failed to start against the bundled database

​

Changed

• Each component is configured under its own key with a deployment block, or statefulSet for the bundled datastores. See the values reference