> ## Documentation Index
> Fetch the complete documentation index at: https://docs.abbyy.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Vantage self-hosted 2026.10.16 release notes

> ABBYY Vantage self-hosted 2026.10.16, built from Vantage Cloud 3.6.1.

| Component | Version |
| - | - |
| Vantage Cloud | [3.6.1](/vantage/release-notes/3-6-1) |
| `vantage-operator` and `vantage-selfhosted` charts | `0.91.1` |

This release is built from Vantage Cloud 3.6.1. For the product features, improvements, and fixes in that version, see the [Vantage 3.6.1 release notes](/vantage/release-notes/3-6-1).

## Breaking changes

### Default skills are no longer tied to minimal installation

`vantage.minimalInstall` now controls only workload sizing. The set of default skills is chosen by the new `defaultSkills.mode` value, which defaults to `all` (every version of every default skill).

If you run a minimal installation, add `defaultSkills.mode: minimal` to your values before you upgrade. Otherwise, the next sync installs the full default skill set, which can take an hour or more. Installed default skills can't be removed.

For more information, see [Default skills](/vantage/self-hosted/v3.0/default-skills).

## Features

### Default skill selection

You can now choose which default skills to install with `defaultSkills.mode`: `all`, `minimal`, or `custom`.

For the modes, how to find the skill IDs for your chart version, and how to add skills later, see [Default skills](/vantage/self-hosted/v3.0/default-skills).

### Per-workload service accounts

You can now give individual workloads their own Kubernetes service accounts with `serviceAccounts.workloads`, instead of sharing one service account across all workloads. The `vantage-selfhosted` chart can create these service accounts or use existing ones. This lets you grant each workload only the access it needs, for example through Azure Workload Identity.

### Pod labels for workloads

You can now add labels to workload pods with `vantage.workloads.pods.labels` (all workloads) or `vantage.workloads.pods.charts` (specific workloads), for example the `azure.workload.identity/use` label that Azure Workload Identity requires.

### OCI migration resources

You can now set CPU and memory requests and limits for the OCI migration job with `vantage.ociMigration.resources`. Use this when your cluster requires limits on every container.

## Improvements

* TechCore workers scale out faster on full installations. Prometheus now scrapes Vantage metrics every 15 seconds instead of 60, and TechCore workers use their production scaling profile. Minimal installations keep 60-second intervals. To change the scrape interval, set `observability.prometheus.serviceMonitor.interval`. For more information, see [Autoscaling](/vantage/self-hosted/v3.0/performance/autoscaling).

## Fixes

* Fixed an issue where notifications in the Vantage UI didn't work with Gateway API ingress, because some implementations refused to route WebSocket connections. Request timeouts on the Gateway API route are now also turned off by default, so long-running requests aren't cut off.

* Fixed an issue on Azure where the operator created and validated Key Vault secret configuration for workloads that weren't installed.