#
Highlights - November 2026
Added AI models used to the Data Pools documentation: which AI model reads, describes and indexes each kind of file, and when. How documents are read now covers document conversion with Docling and picture descriptions.
Added guidance on preparing your workloads for worker node replacement to the Hosted Cluster documentation, covering what to avoid (node-name pinning, acting on nodes during maintenance, restrictive PodDisruptionBudgets, data on the node), the fact that node labels/taints/cordons are not kept, and recommended practices for high availability during updates.
Added an Applications Catalog to the IaaS Platform Architecture documentation, listing the open-source applications and operators available on the shared clusters, what is platform-internal, what is planned, and how to consume each item (self-serve vs. Hosted Cluster).
Added Auto Mode to PHOENIQS Chat: how Auto Mode picks a model for each message, the status line and turn summary shown while it works, the models it uses, and how its model calls are billed.
Added Memory to PHOENIQS Chat: creating, editing and deleting memories in the Memories panel, switching memory off, and the memory permissions administrators set per role.
Added Admin Panel to PHOENIQS Chat: configuration, roles and groups, permissions, grants and branding for the administrators of a Chat Dedicated instance, including the settings managed by PHOENIQS that show greyed out.
Added Usage Dashboard to PHOENIQS Chat: usage per user and model, filters, Excel export and the monthly credit budget of the instance.
Updated the PHOENIQS Chat Quick Start with a new interface screenshot and overview table, filtering, sorting and pinning chats, the Memories, Usage Dashboard and Admin Panel entries, and Auto Mode in the model selector.
Updated the PHOENIQS Chat Feature Overview, Subscriptions and Provision PHOENIQS Chat pages for Auto Mode, Memory, the Admin Panel and the Usage Dashboard. Custom branding of a Chat Dedicated instance is now set by its administrators in the Admin Panel.
inference-deepseek-v32(DeepSeek V3.2) remains on Active Models with a Decommissioning date of 31.12.2026.Updated
inference-miner-u25(MinerU 2.5 Pro) pricing on Active Models: 0.5385 / 0.0769 Credits/M Tokens (input / output).