Skip to main content

Release notes

Qry ships in numbered releases; the version an installation runs is shown in the app. This page summarises what changed, month by month, and links to the detailed notes where they exist.

November 2026​

Installation and security

  • Protected by default. A new installation turns on network policies, so only Qry's own pods reach its PostgreSQL and Redis. Both get random passwords, and pods that run people's code get no Kubernetes token. Upgrading an existing installation turns the policies on and keeps its passwords; Before installing says how to change them.
  • helm test checks a running installation: the backend is ready, the API and the web interface answer, and Redis refuses a pod without its password.
  • Changing a password or another secret with helm upgrade now restarts the pods that use it.
  • The qry CLI is signed. install.sh, install.ps1 and qry update install only releases signed by IXEN, and qry update never moves to an older version on its own.
  • Signed license files that work without a cloud account. Online ones can be withdrawn; offline ones are issued for a quarter.
  • First administrator. A new installation has no users and no default password: the first administrator is created with a one-time setup code.

Google Cloud Marketplace

  • Qry is packaged for Google Cloud Marketplace, with a public repository for command-line installs. The listing is being prepared.

February – October 2026​

The detailed monthly pages for this period have not been written up yet. The main additions:

New areas

  • Forge migration manager — Oracle, Teradata, SQL Server and SAS estates migrated to BigQuery, PostgreSQL or AlloyDB, code included: assessment report, waves built from the dependency graph, translation with self-repair, data reconciliation, function and procedure parity checks, scheduler jobs moved to Lakeflow, and dbt export.
  • Lakeflow — declarative pipelines and DAG jobs, including SQL tasks that run inside the warehouse, quality gates, run reconciliation and freshness alerts.
  • Skills — one model for business guidance (instructions, documents, procedures, scripts) by scope, with an IDE, Git versioning and Skill Health.
  • Nexus — data products with an API for other systems.

Conversations and documents

  • Workspace search is hybrid (meaning plus exact terms) and reranked; answers cite the passage they used.
  • A warning next to any figure in an answer that does not appear in the data read.
  • ⌘K / Ctrl+K to search everything or ask a new question.

Platform

  • A new design, with the interface in English and Spanish and built to WCAG 2.2 AA.
  • Azure AI Foundry as a model provider (Claude and GPT, with Entra ID), alongside Anthropic, OpenAI, Google Vertex AI and Model Garden.
  • Oracle databases older than 12.1 through the customer's own Spark cluster.
  • Issue reports go to IXEN's support desk (Zammad).
  • A Telegram bot over the same permissions as the web.
  • The qry operator CLI for checking installations before and after upgrades.

January 2026​

New features

  • Notebook IDE — command palette (Cmd+P), multi-tab editing, split view, a cell dependency graph and keyboard shortcuts.

Improvements

  • Find and replace across notebook cells.
  • Tab reordering by drag and drop, with previews.
  • Variables panel for inspecting SQL results and Python state.
  • Collapsible cells.

Full notes


December 2025​

New features

  • ML Hub — in-database machine learning from a conversation: training, experiment tracking, feature store and monitoring.

Improvements

  • Forge dependency graph view.
  • Faster model inference.
  • Experiment comparison in ML Hub.

Full notes


November 2025​

New features

  • Dashboards — custom dashboards with drag-and-drop layout, multiple chart types and public sharing.
  • DataFlow — data movement between sources from natural-language instructions.

Improvements

  • Chart rendering with ECharts 5.
  • Faster dashboard tile queries.

Full notes


October 2025​

New features

  • Scheduled tasks — recurring analyses and reports scheduled in plain language.
  • Python execution — Python for advanced analysis, with matplotlib charts.
  • File upload — CSV and Excel files analysed in a conversation.
  • Model permissions — administrators decide which models each group may use.
  • Workspaces — team collaboration with shared conversations and memory.

Improvements

  • A better mobile view.
  • Longer model output, so Python code is not cut off.
  • Email delivery for automated reports.

Infrastructure

  • License validation.
  • Fine-grained access control (RBAC).
  • Memory with vector embeddings.

Full notes