Support Matrix
Document Information
Version: v1.0 Updated: 2026-07-13 Functional baseline: User Manual updated on 2026-07-10
Status Definitions
| Status | Meaning |
|---|---|
| Supported | The current user manual or installation guide documents the capability. Account permission and prerequisites still apply. |
| Conditionally supported | Availability depends on version, platform configuration, delivery package, infrastructure combination, or project validation. |
| Planned | The capability is on the roadmap but is not currently available for production delivery. |
| Temporarily unsupported | The current version must not be planned around this capability. |
This matrix is the product-level status entry. Release notes, the delivered package, the live environment, and project validation take precedence for a specific delivery.
Subsystem Capability Matrix
| Subsystem | Capability | Status | Main Conditions | Manual Entry |
|---|---|---|---|---|
| AI Infra On-Prem | Regions, availability zones, clusters, nodes, and accelerator management | Supported | Compatible infrastructure is prepared and authorized | Regions & Zones |
| AI Infra On-Prem | Specifications, images, storage, and templates | Supported | Required backend resources and templates are configured | Resource Specs |
| AI Infra On-Prem | Development environments, training jobs, and online inference | Supported | Quota, image, storage, accelerator, and template requirements are met | User Overview |
| AI Infra On-Prem | Quotas, credits, metering, usage, and monitoring | Supported | Data visibility depends on role, tenant, and collection status | Usage |
| AI Infra On-Cloud | Cloud platform, account, resource-pool, and authorization management | Supported | Only configured and validated platform types are usable | Access Overview |
| AI Infra On-Cloud | Frameworks, models, runtime images, and scheduling policies | Supported | Assets and policies must match the target cloud platform and region | Models |
| AI Infra On-Cloud | Quick deployment and deployment tracking | Supported | Account, platform, region, assets, compute plan, and authorization are available | Quick Deployment |
| Model Services | Meta-models, sources, templates, tags, and currency settings | Supported | Operator permission and valid base configuration are required | Meta Models |
| Model Services | Single-model and BYOK publishing | Supported | Valid endpoint or deployment information, pricing, visibility, and review requirements | My Models |
| Model Services | Aggregate-model publishing and routing | Supported | Performed by a model provider using eligible member models and available strategies | My Models |
| Model Services | Model and app review | Supported | Performed by an operator; publication state affects user visibility | Model Reviews |
| Model Services | Model discovery and Playground | Supported | A visible, reviewed, and usable model is required; modality depends on the model | Models |
| Model Services | Model-specific API access and call records | Supported | Valid access, endpoint, credentials, quota, and model availability are required | My Calls Overview |
| Model Services | Usage, customer calls, and model revenue | Supported | Scope and freshness depend on role, metering, and synchronization | Model Usage |
| Billing | User billing, transactions, top-up orders, monthly bills, and quota governance | Supported | Scope depends on account, tenant, billing cycle, and synchronization status | Account Overview |
| Billing | Customer finance, finance operations, settlement, reconciliation, and account adjustment | Supported | Operator permission, billing-cycle consistency, and source-record synchronization are required | Reconciliation Center |
| Billing | License quota, validity, activation status, and module authorization | Supported | License status depends on the delivered package, activation state, and configured quota | License |
| Settings | Personal Keys, profile, projects, members, roles, and tenants | Supported | Visibility depends on role, tenant scope, and menu authorization | Dashboard |
| Settings | Platform settings, login properties, operation logs, and audit records | Supported | Platform or security administration permission is required for high-risk changes | Platform Settings |
| Settings | API rate-control rules, observability audit, node cache, and publish center | Supported | Rule effects depend on published state, node synchronization, and current traffic scope | API Rate Control Overview |
Cloud and Accelerator Status
| Item | Status | Product Position |
|---|---|---|
| Cloud platform type and account management | Supported | AI Infra On-Cloud can manage configured platform types, accounts, resource pools, and authorization. The exact provider and access method require environment validation. |
| Huawei Cloud access | Temporarily unsupported | Do not use Huawei Cloud as a current On-Cloud delivery target. |
| Listed NVIDIA, Huawei Ascend, Enflame, Biren, and Hygon accelerator models | Conditionally supported | The listed models can enter On-Prem onboarding; the full software and model combination still requires validation. See Supported Accelerators. |
| Unlisted accelerators | Conditionally supported | An adaptation assessment and PoC are required before a delivery commitment. |
Model Capability Status
| Capability | Status | Product Position |
|---|---|---|
| Text model discovery, experience, publishing, and calling | Supported | Actual results depend on the selected model, endpoint, quota, and inference resources. |
| Image, audio, and video Playground entries | Supported | A compatible and visible model must be configured for the corresponding modality. |
| Single-model, BYOK, and aggregate-model publication | Supported | Provider publishing and operator review responsibilities must be separated. |
| Call logs, analytics, usage, metering, billing, settlement, and revenue views | Supported | Data scope and synchronization depend on role, module, billing cycle, and source records. |
| RAG | Planned | Not currently available as a production delivery capability. |
| Function Calling | Planned | Not currently available as a production delivery capability. |
The matrix does not claim support for a specific API protocol, model family, context length, quantization method, or performance target unless that item is separately confirmed in the delivered version.
Deployment Status
| Deployment Area | Status | Source and Boundary |
|---|---|---|
| AGIOne application single-node quick installation | Supported | Intended for the scope described in the Quick Installation Guide; suitability for production must be assessed separately. |
| AGIOne application host-mode multi-node installation | Supported | Node roles, SSH, private IPs, offline assets, ports, and data safety requirements apply. See Multi-node Installation. |
| Compute-node Kubernetes onboarding | Supported | OS, CPU architecture, Kubernetes, driver, device plugin, network, and storage requirements apply. See Compute Node Requirements. |
| Offline delivery | Conditionally supported | Delivery bundle completeness and image, runtime, and checksum availability must be verified. |
Required Confirmation Before Commitment
- Confirm the deployed AGIOne version and delivery package.
- Confirm the target role, tenant, and authorization scope.
- Confirm cloud platform or accelerator compatibility.
- Confirm the model, endpoint, image, driver, runtime, quota, storage, and network combination.
- Confirm expected capacity, performance, availability, recovery, metering, and acceptance evidence.