{"resourceId":"nim-operator-312-release-notes","versions":[{"version":"external-db6857f47a131c6de2802863c5db17a38cd06acee8429d702044257a2e03365f","resource":{"id":"nim-operator-312-release-notes","title":"NIM Operator v3.1.2: NeMo Retriever 2.x+ weight caching; NeMo Microservices CRDs deprecated toward NeMo Platform","organization":"NVIDIA","sector":"Kubernetes AI operator / NIM lifecycle","geography":"Global (vendor docs)","publishedAt":"NIM Operator 3.1.2 release notes (page undated; inspected 2026-09-27)","publicationDate":null,"eventDate":"2026-09-27","sourceName":"NVIDIA NIM Operator Documentation","sourceLabel":"Vendor release notes","sourceUrl":"https://docs.nvidia.com/nim-operator/3.1.2/release-notes.html","evidenceClass":"vendor-claim","outcomeClass":"emerging","topics":["infrastructure","operating-model","governance-procurement"],"finding":"NVIDIA NIM Operator 3.1.2 release notes add caching of model weights for NeMo Retriever NIM 2.x+ containers (detect download workflow, cache without starting service, reuse PVC; profiles unused because 2.0 introduces a purpose-built runtime), plus priorityClassName and service-only labels. Notes deprecate managing NeMo Microservices CRDs/deployments as they transition to NeMo Platform. Earlier 3.1.x/3.0.x entries in the same page record multi-node Ray path, experimental Dynamo CRDs, experimental Kata/confidential path, and government-ready designation from v3.0.1.","sledRelevance":"Campus/agency K8s platforms running NIM Operator must plan NeMo Microservices-to-Platform migration and can use Retriever 2.x+ caching to cut cold-start latency—government-ready remains a vendor designation, not an agency ATO.","evidence":"Opened 3.1.2 release notes page; v3.1.2 features/bugfixes; deprecation callouts; historical government-ready and Dynamo experimental notes on same page.","architectureImplications":"Operator-managed Retriever caches become first-class; NeMo Microservice CRs are sunset path; Dynamo/Kata remain experimental overlays.","governanceImplications":"Migration ownership for NeMo Platform cutover and whether government-ready stacks still need agency authorization packages.","securityPrivacyImplications":"Caching weights on cluster storage expands the secret/model artifact trust boundary; priority classes affect multi-tenant fairness.","caveats":"Docs page undated beyond version labels; government-ready is NVIDIA designation for FedRAMP High-class use cases, not a substitute for agency ATO; experimental features explicitly limited.","streamIds":["nvidia"],"roles":{"sales":"Open with platform hygiene: NIM Operator 3.1.2 improves NeMo Retriever 2.x+ weight caching and signals NeMo Microservices CRD deprecation toward the NeMo Platform. Sell a migration and readiness engagement, not a net-new SLED application use case. Remind buyers that government-ready language elsewhere in the notes is an NVIDIA designation, not an agency ATO. Unsupported claims include asserting that an Operator upgrade alone equals FedRAMP authorization or that 3.1.2 productionizes Dynamo.","engineering":"Fit for clusters already on NIM Operator that need Retriever cache efficiency and a NeMo Platform migration plan. Integrate priorityClassName and service-only labels, and design PVC cache strategy for Retriever 2.x+ containers that no longer rely on classic model profiles. Prerequisites include the 3.1.2 Helm path, suitable storage classes, and NGC pull credentials. Constraints include deprecated NeMo Microservice CRs and still-experimental Dynamo/Kata paths. Proof of value is shorter first-inference after cache reuse plus a completed CR deprecation inventory.","delivery":"Work is to upgrade the Operator, enable Retriever caching, inventory NeMo Microservice custom resources, and draft Platform migration epics. Dependencies include storage capacity, NGC secrets, and application owners. Ownership sits with Kubernetes platform engineering and MLOps. Skills adoption covers CRD migration and cache-volume operations. Governance needs a change window and a dated deprecation milestone. Acceptance is 3.1.2 running in the target clusters, validated Retriever cache behavior, and a filed migration backlog. Risks include applications stuck on deprecated CRs after removal."},"roleDetails":{"sales":{"customerProblem":"Platform teams fear silent breakages as NeMo packaging moves and cold starts hurt UX.","stakeholders":["platform engineering","MLOps","procurement"],"discoveryQuestions":["Which Operator version are you on?","Do you use NeMo Microservice CRs?","Retriever NIMs in scope?"],"valueHypothesis":"Versioned Operator upgrade plus migration planning reduces outage and cold-start risk.","potentialEngagement":"NIM Operator upgrade and NeMo Platform migration assessment.","unsupportedClaims":["Government-ready equals ATO","3.1.2 productionizes Dynamo"]},"engineering":{"fit":"Operator 3.1.2 for Retriever caching and CR deprecation planning.","architectureIntegration":"Helm upgrade; configure Retriever cache PVCs; map CRs to NeMo Platform targets.","prerequisites":["Helm/GitOps","PVC capacity","NGC auth"],"constraints":["NeMo Microservice CR deprecation","experimental Dynamo/Kata not prod-default"],"security":"Protect model caches and pull secrets; RBAC on Operator.","proofOfValue":["Upgrade dry-run","cache reuse demo","CR inventory"]},"delivery":{"work":"Execute upgrade, validate caches, open migration epics.","dependencies":["cluster maintainers","NeMo app owners","storage"],"ownership":"Platform engineering with MLOps co-owner.","skillsAdoption":"Train on Retriever 2.x cache behavior and Platform transition.","governanceCheckpoints":["Upgrade CAB","deprecation milestone review"],"acceptanceCriteria":["3.1.2 in prod","cache SLO met","migration backlog dated"],"risks":["Skipped inventory","breaking deprecated CR removals"]}},"retrievedAt":"2026-09-28T02:59:36Z","enrichedAt":"2026-09-28T03:06:44Z","enrichmentBasis":"retrieved source","accessibilityWorkforceImplications":"Deprecation communications must reach app teams early to avoid last-minute breakages.","procurementImplications":"AI Enterprise / support entitlements should be checked against Operator and NeMo Platform SKUs.","operatingModelImplications":"Add Operator version and NeMo CR inventory to quarterly platform reviews.","sourceVerification":{"openedUrl":"https://docs.nvidia.com/nim-operator/3.1.2/release-notes.html","referenceExcerpt":"Added support for caching model weights for NVIDIA NeMo Retriever NIM 2.x+ containers","promptVersion":"sled-research-v4.0","model":null,"basis":"agent-reported inspection"}}}]}