From 7a3f5bfa51a487eb58c18cdc2b7362ac358e9fc9 Mon Sep 17 00:00:00 2001 From: Natalie Fisher <53450897+nataliesea@users.noreply.github.com> Date: Thu, 14 May 2026 15:37:02 -0700 Subject: [PATCH 1/9] Added Project scope and intent for AI workflow interoperability Signed-off-by: Natalie Fisher <53450897+nataliesea@users.noreply.github.com> --- .../README.md | 95 ++++++++++++++++++- 1 file changed, 94 insertions(+), 1 deletion(-) diff --git a/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md b/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md index bb5d838b4..07c60c85a 100644 --- a/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md +++ b/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md @@ -19,4 +19,97 @@ Focus on the developer inner loop, everything an AI engineer does on a laptop/de * An technical POC showing <10 min “idea-to-inference” path for cloud-native agent development on a developer laptop. * Clearly documented standards for OCI artefact standardization across runtimes and registries. * Specification / procedure to achieve MOF Class III compliant model distributions via any OCI registry. -* Standardised process for leveraging model signing with artefacts-level provenance to support a verified end-to-end CI/CD reference pipeline including outer loop for AI engineering. \ No newline at end of file +* Standardised process for leveraging model signing with artefacts-level provenance to support a verified end-to-end CI/CD reference pipeline including outer loop for AI engineering. + +## Project Scope & Intent - Cloud-Native AI Developer Workflow Interoperability + +### Overview and Intent +AI developers today frequently work in fragmented local environments that are disconnected from cloud-native operational workflows. While emerging standards like ModelPack and OCI-aligned AI artifact initiatives provide the “packaging” foundations, there is no unified interoperability specification that defines how these artifacts must be structured, secured, and described to move seamlessly from a developers laptop into a Kubernetes-based production system. + +The goal of this initiative is to define a minimal Interoperability Specification (a “Compliance Profile”) for AI Artifacts. Rather than rebuilding the OCI layer structure, this initiative defines the **Standardized Metadata Contract** that must exist on top of packaging formats like **ModelPack**. This ensures that any “Cloud-Native Ready” AI artifact contains the mandatory identity, security, and runtime information required that enables a cohesive developer inner loop and GitOps-driven delivery. + +This initiative intentionally builds on existing OCI-aligned packaging efforts rather than redefining artifact storage or layer mechanics. + +### Scope Overview +This initiative defines the **Interoperability Layer** for AI artifacts, bridging the gap between raw packaging and operational deployment. + +Within this scope, the initiative will explore and document: +* **An Interoperability Profile Spec:** A set of mandatory annotation conventions and metadata requirements (the “Manifest Contract”). +* **Compliance & Trust Requirements:** Standards for signing, SBOMs, and openness classification. +* **Workflow Reference Patterns:** Validating the spec through “Laptop-to-Cluster” GitOps and runtime integration. + +The initiative is intended to encourage ecosystem alignment and workflow interoperability rather than define new standalone packaging specifications or runtime standards. + +### In-Scope Areas +#### 1. The Interoperability Specification +Define a structured, minimal specification for AI artifacts to be considered “Cloud-Native Interoperable”. This does not define OCI layering but specifies the mandatory metadata: +* **Annotation Conventions:** Standardize keys for runtime frameworks (e.g., vllm), hardware accelerators (e.g., nvidia-gpu), and lifecycle status. +* **Agentic Assets:** Standardizing the packaging of “skills”, prompt templates and workflow definitions. + * To ensure interoperability, the internal format for skills will align with the agentskills.io community standard. + * The spec defines how these standardized skills are encapsulated into the OCI layers for consistent distribution and discovery. + * The initiative may leverage Skill DLC as the primary reference for demonstrating how these assets are dynamically loaded and managed. +* Interface Definitions: Define the "Ingredient List” for the different classes of artifacts (Models, RAG contexts, and Agentic Assets). + +This includes defining how artifacts relate to and compose with one another. + +#### 2. Metadata, Relationships, & MOF Mapping +Define how artifacts describe themselves and their dependencies to enable cross-tool discovery: +* **MOF-to-OCI Mapping:** Standardize how the LF AI & Data Model Openness Framework (MOF) classifications (e.g., Class I, II, III) are represented as machine-readable OCI metadata. +* **Lineage & Authorship:** Standardizing metadata for provenance, versioning, and authorship to ensure clear ownership as artifacts move from local environments to registries. +* **Relationship Mapping:** Defining minimal relationships to metadata conventions between related artifacts (e.g., model → skill → pipeline) within the OCI manifest. +* **Dependencies & Compatibility:** A schema for declaring software dependencies and infrastructure requirements (e.g., specific CUDA versions or vRAM minimums) to ensure cross-runtime interoperability. +* **Large Binary Asset Optimization:** Establishing best practices and metadata conventions for the efficient handling of large-scale binary artifacts (multi-GB model weights) within OCI registries, focusing on registry compatibility and layer deduplication. +* **Alignment with CNCF:** Build on existing efforts (e.g., ModelPack, ModelKit, ModelCar) + +#### 3. Supply Chain Security and Transparency +Define the mandatory “Trust Profile” for AI artifacts to ensure they are verifiable before entering production: +* **Cryptographic Identity:** Standardize artifact signing and verification using Sigstore and Notary v2 at the point of creation on a developer's machine. +* **Transparency Manifests:** Mandatory requirements for SBOM (Software Bill of Materials) generation and attachment for all artifact layers. +* **Provenance Metadata:** Defining the "Hardened Provenance" requirements to ensure the journey from local experimentation to a secure registry is immutable and documented. + +The goal is to ensure artifacts are trusted and verifiable before entering CI/CD pipelines. + +#### 4. Developer Inner-Loop & Workflow Interoperability +Define the operational patterns that allow the specification to be utilized in a portable "laptop-to-cluster" journey. +* **Workflow Consistency:** Documenting how existing OCI-aligned tools (ModelPack, ModelKit, ModelCar) can produce artifacts that adhere to this initiative's compliance spec. +* **Local Execution Patterns:** Reference patterns for running specified artifacts in local, container-based environments to ensure high-fidelity parity with remote clusters. +* **Rapid Iteration Flow:** Validation of the spec through a reference implementation targeting a sub-10-minute "idea-to-inference" experience. + +#### 5. GitOps and Kubernetes Integration Patterns +Define the "Handoff" patterns for how artifacts transition into production cloud-native systems. +* **GitOps Delivery Patterns:** Reference architectures for pulling compliant artifacts into Flux or Argo CD workflows. +* **Runtime Integration:** Standardized patterns for the seamless deployment of artifacts into serving platforms (e.g., KServe, vLLM) and registration into model registries (e.g., Kubeflow Model Registry). +* **Enterprise Requirements:** Ensuring the promotion spec accounts for air-gapped, regulated, and hybrid-cloud infrastructure constraints. + +#### 6. Real-World Deployment Considerations +Ensure the approach accounts for: +* Air-gapped and regulated environments +* Enterprise security and compliance requirements +* Regulated environments +* Hybrid and multi-platform infrastructure +* Resource-constrained developer environments + +This ensures the solution is practical and broadly applicable. + +#### 7. Ecosystem Collaboration +This initiative will be developed in collaboration with: +* ModelPack and related OCI-aligned initiatives +* CNCF projects +* LF AI & Data communities +* OpenSSF and supply chain security initiatives +* Kubernetes AI and platform engineering communities + +The intent is to align efforts across communities rather than define a solution in isolation. + +### Out of Scope Areas +* Define new low-level binary compression or OCI layering techniques (deferring to OCI/ModelPack) +* Define model architectures or ML training frameworks +* Mandate specific vendor-locked developer tools +* Standardize outer-loop pipelines beyond reference integration patterns + +### Definition of Success +* **A Published Interoperability Spec:** A validated specification that existing tools can adopt to ensure cloud-native readiness. +* **Cross-Tool Portability:** Demonstrated ability for an artifact built by one tool to be verified and deployed by a different runtime. +* **The "10-Minute Flow":** A successful reference implementation demonstrating the journey from a local idea to a running inference service on Kubernetes. +* **Ecosystem Alignment:** Broad adoption of the "Compliance Profile" metadata across CNCF and LF AI & Data communities. + From 9355fcbb9b0b1e4b49d250d38f89aa0c9582118c Mon Sep 17 00:00:00 2001 From: Natalie Fisher <53450897+nataliesea@users.noreply.github.com> Date: Mon, 18 May 2026 14:39:35 -0700 Subject: [PATCH 2/9] Updated typo and changed "cloud-native" to "cloud native" Signed-off-by: Natalie Fisher <53450897+nataliesea@users.noreply.github.com> --- .../README.md | 18 +++++++++--------- 1 file changed, 9 insertions(+), 9 deletions(-) diff --git a/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md b/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md index 07c60c85a..acc2cbc30 100644 --- a/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md +++ b/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md @@ -2,11 +2,11 @@ https://github.com/cncf/toc/issues/1740 -Integrating the AI developer inner loop into an end-to-end CI/CD process leveraging cloud-native technologies and tooling +Integrating the AI developer inner loop into an end-to-end CI/CD process leveraging cloud native technologies and tooling ## Initiative description -Focus on the developer inner loop, everything an AI engineer does on a laptop/desktop before code or models ever reach CI/CD in a cloud-native environment: +Focus on the developer inner loop, everything an AI engineer does on a laptop/desktop before code or models ever reach CI/CD in a cloud native environment: * Local container workspaces: Reference inner loop workflow using desktop tooling such as Podman Desktop / Podman AI Lab for root-less, GPU-aware experimentation, including template images for PyTorch/LLM stacks and volume-mounted datasets. ​ * Unified model build & run CLI: Hardening inference on developer machine and agentic frameworks to leverage container-based tooling so engineers can easily spin-up inference, RAG and multi-agent services locally with one command. @@ -16,17 +16,17 @@ Focus on the developer inner loop, everything an AI engineer does on a laptop/de ## Deliverable(s) or exit criteria -* An technical POC showing <10 min “idea-to-inference” path for cloud-native agent development on a developer laptop. +* An technical POC showing <10 min “idea-to-inference” path for cloud native agent development on a developer laptop. * Clearly documented standards for OCI artefact standardization across runtimes and registries. * Specification / procedure to achieve MOF Class III compliant model distributions via any OCI registry. * Standardised process for leveraging model signing with artefacts-level provenance to support a verified end-to-end CI/CD reference pipeline including outer loop for AI engineering. -## Project Scope & Intent - Cloud-Native AI Developer Workflow Interoperability +## Project Scope & Intent - Cloud Native AI Developer Workflow Interoperability ### Overview and Intent -AI developers today frequently work in fragmented local environments that are disconnected from cloud-native operational workflows. While emerging standards like ModelPack and OCI-aligned AI artifact initiatives provide the “packaging” foundations, there is no unified interoperability specification that defines how these artifacts must be structured, secured, and described to move seamlessly from a developers laptop into a Kubernetes-based production system. +AI developers today frequently work in fragmented local environments that are disconnected from cloud native operational workflows. While emerging standards like ModelPack and OCI-aligned AI artifact initiatives provide the “packaging” foundations, there is no unified interoperability specification that defines how these artifacts must be structured, secured, and described to move seamlessly from a developers laptop into a Kubernetes-based production system. -The goal of this initiative is to define a minimal Interoperability Specification (a “Compliance Profile”) for AI Artifacts. Rather than rebuilding the OCI layer structure, this initiative defines the **Standardized Metadata Contract** that must exist on top of packaging formats like **ModelPack**. This ensures that any “Cloud-Native Ready” AI artifact contains the mandatory identity, security, and runtime information required that enables a cohesive developer inner loop and GitOps-driven delivery. +The goal of this initiative is to define a minimal Interoperability Specification (a “Compliance Profile”) for AI Artifacts. Rather than rebuilding the OCI layer structure, this initiative defines the **Standardized Metadata Contract** that must exist on top of packaging formats like **ModelPack**. This ensures that any “Cloud Native Ready” AI artifact contains the mandatory identity, security, and runtime information required that enables a cohesive developer inner loop and GitOps-driven delivery. This initiative intentionally builds on existing OCI-aligned packaging efforts rather than redefining artifact storage or layer mechanics. @@ -42,7 +42,7 @@ The initiative is intended to encourage ecosystem alignment and workflow interop ### In-Scope Areas #### 1. The Interoperability Specification -Define a structured, minimal specification for AI artifacts to be considered “Cloud-Native Interoperable”. This does not define OCI layering but specifies the mandatory metadata: +Define a structured, minimal specification for AI artifacts to be considered “Cloud Native Interoperable”. This does not define OCI layering but specifies the mandatory metadata: * **Annotation Conventions:** Standardize keys for runtime frameworks (e.g., vllm), hardware accelerators (e.g., nvidia-gpu), and lifecycle status. * **Agentic Assets:** Standardizing the packaging of “skills”, prompt templates and workflow definitions. * To ensure interoperability, the internal format for skills will align with the agentskills.io community standard. @@ -76,7 +76,7 @@ Define the operational patterns that allow the specification to be utilized in a * **Rapid Iteration Flow:** Validation of the spec through a reference implementation targeting a sub-10-minute "idea-to-inference" experience. #### 5. GitOps and Kubernetes Integration Patterns -Define the "Handoff" patterns for how artifacts transition into production cloud-native systems. +Define the "Handoff" patterns for how artifacts transition into production cloud native systems. * **GitOps Delivery Patterns:** Reference architectures for pulling compliant artifacts into Flux or Argo CD workflows. * **Runtime Integration:** Standardized patterns for the seamless deployment of artifacts into serving platforms (e.g., KServe, vLLM) and registration into model registries (e.g., Kubeflow Model Registry). * **Enterprise Requirements:** Ensuring the promotion spec accounts for air-gapped, regulated, and hybrid-cloud infrastructure constraints. @@ -108,7 +108,7 @@ The intent is to align efforts across communities rather than define a solution * Standardize outer-loop pipelines beyond reference integration patterns ### Definition of Success -* **A Published Interoperability Spec:** A validated specification that existing tools can adopt to ensure cloud-native readiness. +* **A Published Interoperability Spec:** A validated specification that existing tools can adopt to ensure cloud native readiness. * **Cross-Tool Portability:** Demonstrated ability for an artifact built by one tool to be verified and deployed by a different runtime. * **The "10-Minute Flow":** A successful reference implementation demonstrating the journey from a local idea to a running inference service on Kubernetes. * **Ecosystem Alignment:** Broad adoption of the "Compliance Profile" metadata across CNCF and LF AI & Data communities. From 94f46ed331c03408e79820333b05de668c9169c5 Mon Sep 17 00:00:00 2001 From: Natalie Fisher <53450897+nataliesea@users.noreply.github.com> Date: Thu, 18 Jun 2026 09:58:31 -0700 Subject: [PATCH 3/9] Apply suggestion from @sabre1041 Co-authored-by: Andrew Block Signed-off-by: Natalie Fisher <53450897+nataliesea@users.noreply.github.com> --- .../initiatives/cloud-native-oci-compliant-inner-loop/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md b/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md index acc2cbc30..cc94635fe 100644 --- a/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md +++ b/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md @@ -6,7 +6,7 @@ Integrating the AI developer inner loop into an end-to-end CI/CD process leverag ## Initiative description -Focus on the developer inner loop, everything an AI engineer does on a laptop/desktop before code or models ever reach CI/CD in a cloud native environment: +Focus on inner loop development which incorporates everything an AI engineer does on a local environment before code or models ever reach CI/CD in a cloud native environment: * Local container workspaces: Reference inner loop workflow using desktop tooling such as Podman Desktop / Podman AI Lab for root-less, GPU-aware experimentation, including template images for PyTorch/LLM stacks and volume-mounted datasets. ​ * Unified model build & run CLI: Hardening inference on developer machine and agentic frameworks to leverage container-based tooling so engineers can easily spin-up inference, RAG and multi-agent services locally with one command. From 6e804797ba64709c6211ac62757584a978a4030f Mon Sep 17 00:00:00 2001 From: Natalie Fisher <53450897+nataliesea@users.noreply.github.com> Date: Thu, 18 Jun 2026 10:00:49 -0700 Subject: [PATCH 4/9] Update README.md Signed-off-by: Natalie Fisher <53450897+nataliesea@users.noreply.github.com> --- .../initiatives/cloud-native-oci-compliant-inner-loop/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md b/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md index cc94635fe..48de9eede 100644 --- a/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md +++ b/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md @@ -9,7 +9,7 @@ Integrating the AI developer inner loop into an end-to-end CI/CD process leverag Focus on inner loop development which incorporates everything an AI engineer does on a local environment before code or models ever reach CI/CD in a cloud native environment: * Local container workspaces: Reference inner loop workflow using desktop tooling such as Podman Desktop / Podman AI Lab for root-less, GPU-aware experimentation, including template images for PyTorch/LLM stacks and volume-mounted datasets. ​ -* Unified model build & run CLI: Hardening inference on developer machine and agentic frameworks to leverage container-based tooling so engineers can easily spin-up inference, RAG and multi-agent services locally with one command. +* Unified Inner-Loop CLI: Hardening local inference and agentic frameworks via container-based tooling, allowing engineers to spin up inference, RAG, and multi-agent services locally with a single command. * Standard packaging of artefacts: Drive convergence between various implementations such as ModelKit, ModelCar towards the emerging ModelPack spec to create a single OCI-manifest that can hold model weights, metadata and SBOM. * Inner-loop supply-chain security: Integrate Notary v2 / model authenticity and transparency via Sigstore, LF AI & Data Model Openness Framework-generated model & data cards, plus SBOM annotations directly into the OCI artefact so that security & openness are “baked in” before CI. ​ * Fast hand-off to outer loop: Provide reference GitOps flows (Flux/Argo) that pull the signed artefact into KServe with ModelPack image-mount optimisation, and register versions in Kubeflow Model Registry. From 630f7ce9161aca2ad4fe116ae1def3af1d8be99c Mon Sep 17 00:00:00 2001 From: Natalie Fisher <53450897+nataliesea@users.noreply.github.com> Date: Thu, 18 Jun 2026 10:06:18 -0700 Subject: [PATCH 5/9] Apply suggestions from code review Co-authored-by: Andrew Block Signed-off-by: Natalie Fisher <53450897+nataliesea@users.noreply.github.com> --- .../cloud-native-oci-compliant-inner-loop/README.md | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md b/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md index 48de9eede..8095bbfd0 100644 --- a/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md +++ b/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md @@ -24,7 +24,7 @@ Focus on inner loop development which incorporates everything an AI engineer doe ## Project Scope & Intent - Cloud Native AI Developer Workflow Interoperability ### Overview and Intent -AI developers today frequently work in fragmented local environments that are disconnected from cloud native operational workflows. While emerging standards like ModelPack and OCI-aligned AI artifact initiatives provide the “packaging” foundations, there is no unified interoperability specification that defines how these artifacts must be structured, secured, and described to move seamlessly from a developers laptop into a Kubernetes-based production system. +AI developers today frequently work in fragmented local environments that are disconnected from cloud native operational workflows. While emerging standards like ModelPack and OCI-aligned AI artifact initiatives provide the “packaging” foundations, there is no unified interoperability specification that defines how these artifacts must be structured, secured, and described to move seamlessly from a developers environments into a Kubernetes-based production system. The goal of this initiative is to define a minimal Interoperability Specification (a “Compliance Profile”) for AI Artifacts. Rather than rebuilding the OCI layer structure, this initiative defines the **Standardized Metadata Contract** that must exist on top of packaging formats like **ModelPack**. This ensures that any “Cloud Native Ready” AI artifact contains the mandatory identity, security, and runtime information required that enables a cohesive developer inner loop and GitOps-driven delivery. @@ -36,7 +36,7 @@ This initiative defines the **Interoperability Layer** for AI artifacts, bridgin Within this scope, the initiative will explore and document: * **An Interoperability Profile Spec:** A set of mandatory annotation conventions and metadata requirements (the “Manifest Contract”). * **Compliance & Trust Requirements:** Standards for signing, SBOMs, and openness classification. -* **Workflow Reference Patterns:** Validating the spec through “Laptop-to-Cluster” GitOps and runtime integration. +* **Workflow Reference Patterns:** Validating the spec through “Local Environment-to-Cluster” GitOps and runtime integration. The initiative is intended to encourage ecosystem alignment and workflow interoperability rather than define new standalone packaging specifications or runtime standards. @@ -70,7 +70,7 @@ Define the mandatory “Trust Profile” for AI artifacts to ensure they are ver The goal is to ensure artifacts are trusted and verifiable before entering CI/CD pipelines. #### 4. Developer Inner-Loop & Workflow Interoperability -Define the operational patterns that allow the specification to be utilized in a portable "laptop-to-cluster" journey. +Define the operational patterns that allow the specification to be utilized in a portable "local environment-to-cluster" journey. * **Workflow Consistency:** Documenting how existing OCI-aligned tools (ModelPack, ModelKit, ModelCar) can produce artifacts that adhere to this initiative's compliance spec. * **Local Execution Patterns:** Reference patterns for running specified artifacts in local, container-based environments to ensure high-fidelity parity with remote clusters. * **Rapid Iteration Flow:** Validation of the spec through a reference implementation targeting a sub-10-minute "idea-to-inference" experience. From 1e60f7fbad66ac7a3188f96918af0b24b88fbf45 Mon Sep 17 00:00:00 2001 From: Natalie Fisher <53450897+nataliesea@users.noreply.github.com> Date: Thu, 18 Jun 2026 10:18:53 -0700 Subject: [PATCH 6/9] Update README.md Signed-off-by: Natalie Fisher <53450897+nataliesea@users.noreply.github.com> --- .../cloud-native-oci-compliant-inner-loop/README.md | 8 ++++---- 1 file changed, 4 insertions(+), 4 deletions(-) diff --git a/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md b/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md index 8095bbfd0..c993c996f 100644 --- a/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md +++ b/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md @@ -26,7 +26,7 @@ Focus on inner loop development which incorporates everything an AI engineer doe ### Overview and Intent AI developers today frequently work in fragmented local environments that are disconnected from cloud native operational workflows. While emerging standards like ModelPack and OCI-aligned AI artifact initiatives provide the “packaging” foundations, there is no unified interoperability specification that defines how these artifacts must be structured, secured, and described to move seamlessly from a developers environments into a Kubernetes-based production system. -The goal of this initiative is to define a minimal Interoperability Specification (a “Compliance Profile”) for AI Artifacts. Rather than rebuilding the OCI layer structure, this initiative defines the **Standardized Metadata Contract** that must exist on top of packaging formats like **ModelPack**. This ensures that any “Cloud Native Ready” AI artifact contains the mandatory identity, security, and runtime information required that enables a cohesive developer inner loop and GitOps-driven delivery. +The goal of this initiative is to define a minimal Interoperability Specification (a “Compliance Profile”) for AI Artifacts. Rather than rebuilding the OCI layer structure, this initiative defines the **Standardized Metadata Contract** that must exist on top of packaging formats like **ModelPack**. This ensures that any ["Cloud-Native Ready"](https://github.com/cncf/foundation/blob/main/style-guide.md#1-cloud-native-and-open-source) AI artifact contains the mandatory identity, security, and runtime information required that enables a cohesive developer inner loop and GitOps-driven delivery. This initiative intentionally builds on existing OCI-aligned packaging efforts rather than redefining artifact storage or layer mechanics. @@ -45,7 +45,7 @@ The initiative is intended to encourage ecosystem alignment and workflow interop Define a structured, minimal specification for AI artifacts to be considered “Cloud Native Interoperable”. This does not define OCI layering but specifies the mandatory metadata: * **Annotation Conventions:** Standardize keys for runtime frameworks (e.g., vllm), hardware accelerators (e.g., nvidia-gpu), and lifecycle status. * **Agentic Assets:** Standardizing the packaging of “skills”, prompt templates and workflow definitions. - * To ensure interoperability, the internal format for skills will align with the agentskills.io community standard. + * To ensure interoperability, the internal format for skills will align with the ["agentskills.io"] (https://agentskills.io/home) community standard. * The spec defines how these standardized skills are encapsulated into the OCI layers for consistent distribution and discovery. * The initiative may leverage Skill DLC as the primary reference for demonstrating how these assets are dynamically loaded and managed. * Interface Definitions: Define the "Ingredient List” for the different classes of artifacts (Models, RAG contexts, and Agentic Assets). @@ -59,7 +59,7 @@ Define how artifacts describe themselves and their dependencies to enable cross- * **Relationship Mapping:** Defining minimal relationships to metadata conventions between related artifacts (e.g., model → skill → pipeline) within the OCI manifest. * **Dependencies & Compatibility:** A schema for declaring software dependencies and infrastructure requirements (e.g., specific CUDA versions or vRAM minimums) to ensure cross-runtime interoperability. * **Large Binary Asset Optimization:** Establishing best practices and metadata conventions for the efficient handling of large-scale binary artifacts (multi-GB model weights) within OCI registries, focusing on registry compatibility and layer deduplication. -* **Alignment with CNCF:** Build on existing efforts (e.g., ModelPack, ModelKit, ModelCar) +* **Alignment with CNCF:** Build on existing efforts (e.g., ["ModelPack"] (https://modelpack.org/), ["ModelKit"] (https://kitops.org/docs/modelkit/intro/), ["ModelCar"] (https://github.com/redhat-ai-services/modelcar-catalog) #### 3. Supply Chain Security and Transparency Define the mandatory “Trust Profile” for AI artifacts to ensure they are verifiable before entering production: @@ -78,7 +78,7 @@ Define the operational patterns that allow the specification to be utilized in a #### 5. GitOps and Kubernetes Integration Patterns Define the "Handoff" patterns for how artifacts transition into production cloud native systems. * **GitOps Delivery Patterns:** Reference architectures for pulling compliant artifacts into Flux or Argo CD workflows. -* **Runtime Integration:** Standardized patterns for the seamless deployment of artifacts into serving platforms (e.g., KServe, vLLM) and registration into model registries (e.g., Kubeflow Model Registry). +* **Runtime Integration:** Standardized patterns for the seamless deployment of artifacts into serving platforms (e.g., ["KServe"] (https://kserve.github.io/website), [vLLM](https://github.com/vllm-project/vllm), and registration into model registries (e.g., Kubeflow Model Registry). * **Enterprise Requirements:** Ensuring the promotion spec accounts for air-gapped, regulated, and hybrid-cloud infrastructure constraints. #### 6. Real-World Deployment Considerations From 8b61cafd5c5e257e3ae6c20ee65d441687350022 Mon Sep 17 00:00:00 2001 From: Natalie Fisher <53450897+nataliesea@users.noreply.github.com> Date: Thu, 18 Jun 2026 14:06:14 -0700 Subject: [PATCH 7/9] Update tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md Co-authored-by: Terry Howe Signed-off-by: Natalie Fisher <53450897+nataliesea@users.noreply.github.com> --- .../initiatives/cloud-native-oci-compliant-inner-loop/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md b/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md index c993c996f..2386e4a34 100644 --- a/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md +++ b/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md @@ -83,7 +83,7 @@ Define the "Handoff" patterns for how artifacts transition into production cloud #### 6. Real-World Deployment Considerations Ensure the approach accounts for: -* Air-gapped and regulated environments +* Air-gapped * Enterprise security and compliance requirements * Regulated environments * Hybrid and multi-platform infrastructure From 353ef7157b70bff90df288ff3cc4d252526dcc28 Mon Sep 17 00:00:00 2001 From: Natalie Fisher <53450897+nataliesea@users.noreply.github.com> Date: Thu, 18 Jun 2026 14:19:53 -0700 Subject: [PATCH 8/9] Apply suggestions from code review Co-authored-by: Andrew Block Signed-off-by: Natalie Fisher <53450897+nataliesea@users.noreply.github.com> --- .../initiatives/cloud-native-oci-compliant-inner-loop/README.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md b/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md index 2386e4a34..a14eca432 100644 --- a/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md +++ b/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md @@ -16,7 +16,7 @@ Focus on inner loop development which incorporates everything an AI engineer doe ## Deliverable(s) or exit criteria -* An technical POC showing <10 min “idea-to-inference” path for cloud native agent development on a developer laptop. +* An technical POC showing <10 min “idea-to-inference” path for cloud native agent development on a developer environment. * Clearly documented standards for OCI artefact standardization across runtimes and registries. * Specification / procedure to achieve MOF Class III compliant model distributions via any OCI registry. * Standardised process for leveraging model signing with artefacts-level provenance to support a verified end-to-end CI/CD reference pipeline including outer loop for AI engineering. From 89dc284b95577ed36b01ccbcf8c221b983a7e6f5 Mon Sep 17 00:00:00 2001 From: Natalie Fisher <53450897+nataliesea@users.noreply.github.com> Date: Thu, 18 Jun 2026 14:24:24 -0700 Subject: [PATCH 9/9] Update README.md Updated from feedback provided Signed-off-by: Natalie Fisher <53450897+nataliesea@users.noreply.github.com> --- .../README.md | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md b/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md index a14eca432..787c2db1b 100644 --- a/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md +++ b/tags/tag-developer-experience/initiatives/cloud-native-oci-compliant-inner-loop/README.md @@ -16,7 +16,7 @@ Focus on inner loop development which incorporates everything an AI engineer doe ## Deliverable(s) or exit criteria -* An technical POC showing <10 min “idea-to-inference” path for cloud native agent development on a developer environment. +* A technical POC showing <10 min “idea-to-inference” path for cloud native agent development on a developer environment. * Clearly documented standards for OCI artefact standardization across runtimes and registries. * Specification / procedure to achieve MOF Class III compliant model distributions via any OCI registry. * Standardised process for leveraging model signing with artefacts-level provenance to support a verified end-to-end CI/CD reference pipeline including outer loop for AI engineering. @@ -47,7 +47,7 @@ Define a structured, minimal specification for AI artifacts to be considered “ * **Agentic Assets:** Standardizing the packaging of “skills”, prompt templates and workflow definitions. * To ensure interoperability, the internal format for skills will align with the ["agentskills.io"] (https://agentskills.io/home) community standard. * The spec defines how these standardized skills are encapsulated into the OCI layers for consistent distribution and discovery. - * The initiative may leverage Skill DLC as the primary reference for demonstrating how these assets are dynamically loaded and managed. + * The initiative may leverage ["Skill DLC"] (https://agentskills.io/home) as the primary reference for demonstrating how these assets are dynamically loaded and managed. * Interface Definitions: Define the "Ingredient List” for the different classes of artifacts (Models, RAG contexts, and Agentic Assets). This includes defining how artifacts relate to and compose with one another. @@ -63,9 +63,9 @@ Define how artifacts describe themselves and their dependencies to enable cross- #### 3. Supply Chain Security and Transparency Define the mandatory “Trust Profile” for AI artifacts to ensure they are verifiable before entering production: -* **Cryptographic Identity:** Standardize artifact signing and verification using Sigstore and Notary v2 at the point of creation on a developer's machine. +* **Cryptographic Identity:** Standardize artifact signing and verification at the point of creation on a developer's machine, leveraging established frameworks like Sigstore and Notary v2 (Notation), while exploring emerging zero-trust identity protocols such as OpenPubkey to enable seamless OIDC-bound signing. * **Transparency Manifests:** Mandatory requirements for SBOM (Software Bill of Materials) generation and attachment for all artifact layers. -* **Provenance Metadata:** Defining the "Hardened Provenance" requirements to ensure the journey from local experimentation to a secure registry is immutable and documented. +* **Provenance Metadata:** Defining "Hardened Provenance" requirements to ensure the journey from local experimentation to an enterprise registry is immutable and documented. This includes reference patterns for utilizing local, ephemeral OCI registries during the developer inner loop, allowing artifact manifests to be generated, signed, and verified via Sigstore or Notary v2 entirely on the local machine before entering external CI/CD pipelines. The goal is to ensure artifacts are trusted and verifiable before entering CI/CD pipelines. @@ -77,7 +77,7 @@ Define the operational patterns that allow the specification to be utilized in a #### 5. GitOps and Kubernetes Integration Patterns Define the "Handoff" patterns for how artifacts transition into production cloud native systems. -* **GitOps Delivery Patterns:** Reference architectures for pulling compliant artifacts into Flux or Argo CD workflows. +* **GitOps Delivery Patterns:** Reference architectures for pulling compliant artifacts into Flux or Argo CD workflows. This includes standardizing K8s Init Container patterns (leveraging tooling like the ["KitOps init container"] (https://github.com/kitops-ml/kitops)) to pull, cryptographically verify signatures, and unpack specified artifact layers into a shared volume before the main inference or serving runtime boots. * **Runtime Integration:** Standardized patterns for the seamless deployment of artifacts into serving platforms (e.g., ["KServe"] (https://kserve.github.io/website), [vLLM](https://github.com/vllm-project/vllm), and registration into model registries (e.g., Kubeflow Model Registry). * **Enterprise Requirements:** Ensuring the promotion spec accounts for air-gapped, regulated, and hybrid-cloud infrastructure constraints. @@ -93,8 +93,8 @@ This ensures the solution is practical and broadly applicable. #### 7. Ecosystem Collaboration This initiative will be developed in collaboration with: -* ModelPack and related OCI-aligned initiatives -* CNCF projects +* Related OCI-aligned initiatives +* CNCF projects, like ModelPack * LF AI & Data communities * OpenSSF and supply chain security initiatives * Kubernetes AI and platform engineering communities @@ -111,5 +111,5 @@ The intent is to align efforts across communities rather than define a solution * **A Published Interoperability Spec:** A validated specification that existing tools can adopt to ensure cloud native readiness. * **Cross-Tool Portability:** Demonstrated ability for an artifact built by one tool to be verified and deployed by a different runtime. * **The "10-Minute Flow":** A successful reference implementation demonstrating the journey from a local idea to a running inference service on Kubernetes. -* **Ecosystem Alignment:** Broad adoption of the "Compliance Profile" metadata across CNCF and LF AI & Data communities. +* ** (Stetch Goal) Ecosystem Alignment:** Broad adoption of the "Compliance Profile" metadata across CNCF and LF AI & Data communities.