Prerequisites
Ensure that all AI Services prerequisites are fulfilled before proceeding with application creation. Refer to the Configuration and Installation section. AI Services supports two runtime environments: Podman and OpenShift.
Hardware Requirements
Before deploying Digital Assistant, ensure the following hardware components are available and configured:
- Storage: 256 GB
- Memory: 256 GB
- CPU: 24 cores (48 vCPU with SMT2) - SMT2 is recommended.
- Spyre Cards(5):
- 4 cards for Granite Instruct model
- 1 card for Re-ranker model
Ensure that the worker node can allocate a minimum of 5 Spyre cards
- Storage:
The following values specify the minimum required disk capacity. The directories do not need separate disks; if they share a filesystem, ensure the total available capacity meets the combined requirement. Additional capacity is recommended for production use.
/var/lib/containers1 : 128 GB (container images and logs)/var/lib/ai-services1 : 64 GB+ (AI models, caddy assets; scale as needed)
If the required storage is not available, see the Troubleshooting Storage Issues for workarounds.
- Memory: 256 GB
- CPU: 15 cores for high concurrency and AI workloads
- Spyre Cards(5)
- 4 cards for Granite Instruct model
- 1 card for Re-ranker model
- Storage:
The following values specify the minimum required disk capacity. The directories do not need separate disks; if they share a filesystem, ensure the total available capacity meets the combined requirement. Additional capacity is recommended for production use.
/var/lib/containers1 : 128 GB (container images and logs)/var/lib/ai-services1 : 64 GB+ (AI models, caddy assets; scale as needed)
If the required storage is not available, see the Troubleshooting Storage Issues for workarounds.
- Memory: 256 GB
- CPU: 15 cores for high concurrency and AI workloads
- Storage:
The following values specify the minimum required disk capacity. The directories do not need separate disks; if they share a filesystem, ensure the total available capacity meets the combined requirement. Additional capacity is recommended for production use.
/var/lib/containers1 : 128 GB (container images and logs)/var/lib/ai-services1 : 64 GB+ (AI models, caddy assets; scale as needed)
If the required storage is not available, see the Troubleshooting Storage Issues for workarounds.
- Memory: 75 GB
- CPU: 10 cores for high concurrency and AI workloads
/var/lib is the default base directory for Podman runtime. The base directory can be overridden during initial catalog configuration using the --basedir flag.