Ai-OPs Koios Kubernetes Deployment Support & Compatibility Policy

Ai-OPs Koios Kubernetes Deployment Support & Compatibility Policy

Version: 1
Effective Date: February 22, 2026
Status: Official Release

1. Purpose

This document defines Ai-OPs’ official support position, compatibility guarantees, and upgrade practices for Koios deployments on Kubernetes, including clarifications intended to avoid unintended or open‑ended commitments. It is written to support enterprise customers who require clear operational assurances while preserving Ai-OPs’ ability to innovate and evolve the Koios platform.

This policy applies specifically to the Koios Kubernetes deployment artifacts maintained by Ai‑OPs, including Helm charts, YAML manifests, container images, and associated deployment documentation

2. Definitions

For the purposes of this document, the following definitions apply:

• Koios: Ai-OPs’ self-hosted model inferencing and model management software platform.

• Ai-OPs: Ai-OPs, Inc., the developer and maintainer of the Koios platform.

• Kubernetes: The open-source container orchestration platform used to deploy, manage, and scale containerized applications.

• Supported Kubernetes Versions: Kubernetes releases that are actively maintained upstream and explicitly validated and documented by Ai-OPs for a given Koios release.

• Deployment Artifacts: Kubernetes manifests, Helm charts, YAML files, container images, and related configuration maintained by Ai-OPs for Koios deployments.

• Upgrade: An in-place transition from one Koios version to another using supported deployment artifacts and migration tooling, without requiring a full reinstallation.

• Semantic Versioning (SemVer): The MAJOR.MINOR.PATCH versioning scheme used to communicate compatibility and change scope.

• Model Runtime Dependencies: Third-party libraries (e.g., Python packages) required to execute models within Koios.

• Customer: A legal entity that has licensed Koios from Ai-OPs and is contractually entitled to deploy and operate the software.

• Authorized Users: Individuals explicitly permitted by the Customer to access Koios, in accordance with licensing terms, access controls, and security policies.

3. Governing Agreement and Support Boundaries

3.1. Governing Customer Agreement

All Koios deployments, licensed use, access rights, and support obligations are governed by Ai-OPs’ Customer Agreement (including any applicable Order Forms, Statements of Work, or amendments).

In the event of any inconsistency between this document and the Koios Customer Agreement, the Koios Customer Agreement shall control.

This document is intended to describe technical support posture and deployment practices only and does not modify, expand, or override contractual terms.

3.2. Support Boundaries and Responsibilities

Ai‑OPs provides support for:

• Koios application behavior as delivered

• Official Kubernetes deployment artifacts (helm chart) published by Ai-OPs

• Supported Kubernetes versions and configurations as documented

Ai‑OPs does not assume responsibility for:

• Customer specific Kubernetes cluster architecture or administration, including external reverse proxies specific Kubernetes cluster architecture or administration‑specific Kubernetes cluster architecture or administration

• Infrastructure provider outages or failures

• Unsupported Kubernetes versions

• Any forked or modified software components, deployment manifests or helm charts are not covered under Ai-OPs’ official Kubernetes deployment support obligations.

  Official Kubernetes Support Position

  Kubernetes as a Primary Deployment Platform

4. Official Kubernetes Support Position

4.1. Kubernetes as a Primary Deployment Platform

Ai-OPs confirms that Kubernetes is and will remain a primary, officially supported deployment platform for Koios.

This means:

• Koios is designed, tested, and released with Kubernetes as a first-class deployment target.

• Ai-OPs maintains Kubernetes deployment assets (semantically versioned Helm Chart) as part of its standard product lifecycle. Customers are not allowed to modify the Koios helm chart itself, but are explicitly allowed to customize and configure Koios based on the helm chart values for their specific customer environment.

• Koios is architected as a deployment-agnostic, containerized software platform. Kubernetes support does not introduce coupling to Kubernetes-specific APIs or internal architectural dependencies. Ai-OPs maintains a unified Koios codebase and release structure across supported deployment models.  Kubernetes support is implemented through deployment configuration and orchestration artifacts and does not introduce coupling or dependency changes within the Koios application stack itself.  

AI-OPs will start to officially support Kubernetes according to the definitions lined out in this document end of March 2026

4.2. Interpretation of “All Future” Kubernetes Compatibility

Any reference to “all future” Kubernetes compatibility is intended to mean that Kubernetes will remain a primary and officially supported deployment platform for Koios.

This statement explicitly excludes the following interpretations:

• An obligation to support every future Kubernetes version indefinitely. However, Koios will support all future officially CNCF certified Kubernetes distributions (like RKE2 and k3s) at latest three months after official GA release and until at latest three months after official end of life. Kubernetes releases and end of life announcements are published by the CNCF under this link: https://kubernetes.io/releases/

• A commitment to backward compatibility with deprecated or end-of-life Kubernetes releases  

• A restriction or freeze on Koios feature development or architectural evolution  

Ai-OPs maintains compatibility with Kubernetes through versioned deployment artifacts and validation against a defined set of supported Kubernetes versions. Compatibility is managed as part of the normal product lifecycle and does not impose open-ended or impractical support guarantees.

5. Kubernetes Deployment Artifacts (GitHub)

Ai-OPs maintains official Kubernetes deployment artifacts through its authoritative public GitHub repository:

https://github.com/Ai-Ops-Inc/koios-kubernetes-deploy

This repository contains the supported Helm charts, Kubernetes YAML manifests, and version-aligned configuration required for Koios deployments:

• Helm charts / YAML manifests

• Versioned configuration aligned to Koios releases.  Additionally, the Koios helm chart will be published as an OCI artifact to an AI-OPs operated OCI registry.

These artifacts are:

• Version-controlled and maintained by Ai-OPs

• Updated in coordination with each Koios release

• The sole authoritative reference for supported Kubernetes deployment and upgrade practices for self-hosted Koios

6. Kubernetes Version Support Policy

Ai‑OPs will:

• Publish and maintain a supported Kubernetes version matrix per Koios release.

• This supported Kubernetes version matrix will be published in Koios release documentation and/or within the Koios-Kubernetes-deploy repository to ensure customers have a clear, version-specific source of truth.

• Publish potential bug fixes to the Koios helm chart in a release cycle independent from the Koios release, i.e. the helm chart may get a minor/major update while the Koios release version remains static. To control these release cycles, structured semantic versioning will be used according to the mechanisms foreseen by helm charts (i.e. The helm chart appVersion and version)

Ai-OPs may discontinue support for Kubernetes versions that have reached upstream end-of-life status. Notice of such changes will be provided through Koios release notes and deployment documentation.

7. Semantic Versioning (SemVer)

Ai-OPs follows Semantic Versioning (MAJOR.MINOR.PATCH) for Koios releases.

Versioning commitments are as follows:

• PATCH releases: Bug fixes and security updates only. No breaking changes.

• MINOR releases: Backward-compatible feature additions and improvements.

• MAJOR releases: May introduce breaking changes. Any such changes will be intentional, documented clearly, and accompanied by upgrade guidance.

8. Upgrade Path vs “Fresh Installation”

8.1. Supported Upgrade Model

Koios is architected to support in‑place upgrades on Kubernetes deployments.

Ai‑OPs provides:

• Tested upgrade paths between supported versions, including major version upgrades (e.g., 1.x à 2.x, 2.x à 3.x).

• Automated database migration tooling as part of the upgrade process.

8.2. Database and Configuration Preservation

As part of the standard upgrade mechanism:

• Database migrations are executed automatically.

• Application data, configuration, and model metadata are preserved.

• Upgrade processes are designed to prevent data loss or configuration corruption.AI-OPs will provide a tested and officially support backup and restore mechanism that allows to generate and store consistent application backups of the complete Koios application state.

8.3. Clarification on “Operational Overhead”

Ai‑OPs does not interpret “fresh installation” avoidance as a requirement for zero downtime or zero operational interruption.

Rather:

• The intent is to avoid unnecessary teardown/rebuild cycles that add administrative complexity.

• Planned maintenance windows remain acceptable and expected in industrial and production environments.

Koios upgrades align with standard enterprise software and industrial automation maintenance practices.

9. Runtime Library Compatibility

9.1. Supported Library Versions

Ai‑OPs cannot and does not commit to unlimited compatibility with all future versions of upstream Python libraries.

Instead, Ai‑OPs commits to:

• Supporting a defined set of tested library versions per Koios release.

• Clearly documenting supported versions of key libraries:  NumPy, Pandas, SciPy, scikitlearn, polars.

This approach reflects the reality that upstream data science libraries evolve continuously and independently of Ai‑OPs.

9.2. Common Data Science Libraries

Libraries such as NumPy and Pandas are widely adopted, stable, and well‑supported. Ai‑OPs actively validates Koios against these libraries within defined version ranges.

For other libraries with faster or less predictable evolution, compatibility is evaluated on a case‑by‑case basis.

10. ONNX Portability Strategy

10.1. Reducing Third‑Party Dependency Risk

Ai‑OPs actively works to minimize reliance on runtime‑bound third‑party Python libraries in production inference environments.

10.2. ONNX as a Preferred Interface

Koios natively supports ONNX (Open Neural Network Exchange) as a production‑grade model interface.

Benefits include:

• Decoupling model execution from Python runtime dependencies

• Improved long‑term compatibility and portability

• Stable inference behavior across platform and library changes

Ai-OPs supports defined and validated ONNX versions per Koios release. Supported ONNX versions are documented as part of the respective release documentation. Adoption of newer ONNX versions occurs following stability assessment, regression testing, and compatibility validation within the normal Koios product lifecycle. Ai-OPs does not commit to indefinite support for all historical or future ONNX versions.

To promote long-term portability and reduce dependency risk, Ai-OPs recommends migrating trained models to ONNX where practical.

11. Summary Statement

In summary:

• Kubernetes remains a primary and supported deployment platform for Koios.

• Ai‑OPs avoids open‑ended or impractical guarantees while providing enterprise‑grade assurances.

• Semantic versioning, upgrade paths, and data preservation are standard and supported.

• Python library compatibility is defined, tested, and documented—not unlimited.

• ONNX is the preferred long‑term strategy for model portability and stability.

This approach balances operational reliability, customer confidence, and continued innovation.

Ai‑OPs, Inc.
Koios Platform Engineering & Support