Provider Attributes

Provider attributes allow you to advertise your provider’s capabilities, location, hardware, and features to potential tenants. These attributes help the Akash network match deployments with providers that meet specific requirements.

Why Attributes Matter

Accurate provider attributes enable:

Better Deployment Matching: Tenants can find providers with specific capabilities
Geographic Selection: Tenants can deploy to specific regions or countries
Hardware Requirements: Match deployments to specific CPU, GPU, or storage types
Feature Discovery: Advertise support for persistent storage, IP leases, custom domains
Transparency: Build trust by sharing provider details (organization, location, contact)

Attribute Configuration

Attributes are defined in your provider.yaml file under the attributes section:

attributes:
  - key: host
    value: akash
  - key: tier
    value: community
  - key: organization
    value: my-provider-org

After updating attributes, restart your provider:

cd /root/provider
helm upgrade akash-provider akash/provider \
  -n akash-services \
  -f provider.yaml \
  --set bidpricescript="$(cat price_script.sh | openssl base64 -A)"

Standard Attributes

Basic information about your provider.

host

Value: akash
Purpose: Identifies provider as Akash Network member
Required: Yes

- key: host
  value: akash

tier

Value: community
Purpose: Provider tier classification
Required: Yes

- key: tier
  value: community

organization

Value: Your organization name
Purpose: Provider operator identification
Example: my-company

- key: organization
  value: my-provider-org

email

Value: Contact email address
Purpose: Support and communication
Required: Yes

- key: email
  value: [email protected]

discord-username

Value: Your Discord username (without @)
Purpose: Contact for audit and provider support
Required for audit: Yes — audited providers must be in the Akash Discord with the Provider role

- key: discord-username
  value: your_discord_username

See Provider Audit for full audit attribute requirements.

website

Value: Provider website URL
Purpose: Marketing and information
Required for audit: Yes

- key: website
  value: https://example.com

status-page

Value: Status page URL
Purpose: Service health monitoring
Optional: Yes

- key: status-page
  value: https://status.example.com

Location Attributes

Geographic and facility information.

country

Value: ISO 3166 Alpha-2 country code
Purpose: Country identification
Examples: US, GB, DE, JP, SG
Recommended: Yes

- key: country
  value: US

city

Value: Three-letter city code
Purpose: City identification
Examples: NYC, LAX, LON, FRA, SIN

- key: city
  value: LAX

location-region

Value: UN geoscheme region code
Purpose: Regional classification for deployment matching
Required: Yes
Note: Use the key location-region. The legacy region key is not supported.
Examples:
- na-us-west, na-us-northeast, na-us-midwest
- eu-west, eu-central, eu-north
- as-east, as-southeast, oc-aus

- key: location-region
  value: na-us-west

See the provider attributes schema for the full list of accepted region values.

timezone

Value: UTC offset (e.g., utc-8, utc+1)
Purpose: Time zone identification

- key: timezone
  value: utc-8

location-type

Values: datacenter, colo, home, office, mix
Purpose: Hosting environment classification

- key: location-type
  value: datacenter

hosting-provider

Value: Data center or cloud provider name
Purpose: Facility identification
Examples: equinix, aws, azure, ovh, hetzner

- key: hosting-provider
  value: equinix

Hardware Attributes

CPU, GPU, storage, and memory specifications.

capabilities/cpu

Values: intel, amd, arm
Purpose: CPU vendor identification
Required: Yes

- key: capabilities/cpu
  value: amd

capabilities/cpu/arch

Values: x86, x86-64, arm, arm-64
Purpose: CPU architecture

- key: capabilities/cpu/arch
  value: x86-64

capabilities/gpu

Values: nvidia, amd, intel, xilinx
Purpose: GPU vendor (if GPUs available)

- key: capabilities/gpu
  value: nvidia

GPU models (capabilities/gpu/vendor/…)

Purpose: Advertise available GPU models
Format: One attribute per model, synced from provider-configs
Examples: rtx4090, a100, h100, gtx1050 (lowercase, no spaces)

- key: capabilities/gpu/vendor/nvidia/model/rtx4090
  value: "true"
- key: capabilities/gpu/vendor/nvidia/model/a100
  value: "true"

See GPU Capabilities below for RAM and interface sub-keys.

cuda

Value: CUDA version string (for example 12.7, 13.0)
Purpose: Advertise the CUDA version available on GPU nodes

- key: cuda
  value: "12.7"

capabilities/memory

Values: ddr2, ddr3, ddr3ecc, ddr4, ddr4ecc, ddr5, ddr5ecc
Purpose: RAM type
Required: Yes

- key: capabilities/memory
  value: ddr5

hardware-energy-source

Values: solar, wind, nuclear, coal, mix, unknown
Purpose: Energy source transparency

- key: hardware-energy-source
  value: solar

hardware-cooling

Values: free, air, liquid
Purpose: Cooling system type
Note: free = free cooling (outside air)

- key: hardware-cooling
  value: liquid

Network Attributes

Internet connectivity specifications.

network-provider

Value: ISP name
Purpose: Network provider identification

- key: network-provider
  value: level3

network-speed-up

Value: Upload speed in Mbps
Purpose: Advertise upload bandwidth

- key: network-speed-up
  value: 10000

network-speed-down

Value: Download speed in Mbps
Purpose: Advertise download bandwidth

- key: network-speed-down
  value: 10000

Feature Attributes

Advertise support for advanced features.

feat-persistent-storage

Values: true, false
Purpose: Advertise persistent storage availability
Required: Yes
Note: Requires Rook-Ceph or similar. Also set the storage class capability keys below when true.

- key: feat-persistent-storage
  value: true

capabilities/storage/1/class

Values: beta1, beta2, beta3
Purpose: Primary persistent storage class
Required: Yes (if you have persistent storage)

- key: capabilities/storage/1/class
  value: beta3

Value	Storage type
`beta1`	HDD
`beta2`	SSD
`beta3`	NVMe

capabilities/storage/1/persistent

Values: "true", "false"
Purpose: Indicate that storage class 1 provides persistent storage
Required: Yes (if you have persistent storage)

- key: capabilities/storage/1/persistent
  value: "true"

Complete persistent storage example:

- key: feat-persistent-storage
  value: true
- key: capabilities/storage/1/class
  value: beta3
- key: capabilities/storage/1/persistent
  value: "true"

Important: You can only advertise one persistent storage class per provider. Choose either beta1 (HDD), beta2 (SSD), or beta3 (NVMe) based on what you configured in Rook-Ceph.

feat-shm

Values: true, false
Purpose: Advertise shared memory (SHM) support via the ram storage class
Recommended: Yes (all providers should support SHM)

- key: feat-shm
  value: true

capabilities/storage/2/class (SHM/Shared Memory)

Value: ram
Purpose: SHM storage class (separate from persistent storage at index 1)
Required: Yes (if advertising SHM)

- key: capabilities/storage/2/class
  value: ram

capabilities/storage/2/persistent (SHM)

Value: "false"
Purpose: SHM is non-persistent
Required: Yes (if advertising SHM)

- key: capabilities/storage/2/persistent
  value: "false"

Complete storage example with persistent storage and SHM:

# Persistent storage (index 1)
- key: feat-persistent-storage
  value: true
- key: capabilities/storage/1/class
  value: beta3
- key: capabilities/storage/1/persistent
  value: "true"

# SHM/Shared Memory (index 2)
- key: feat-shm
  value: true
- key: capabilities/storage/2/class
  value: ram
- key: capabilities/storage/2/persistent
  value: "false"

Note: All providers should support SHM for deployments requiring shared memory. This is configured in the inventory operator during provider installation.

feat-endpoint-ip

Values: true, false
Purpose: Advertise IP lease support

- key: feat-endpoint-ip
  value: true

feat-endpoint-custom-domain

Values: true, false
Purpose: Advertise custom domain support

- key: feat-endpoint-custom-domain
  value: true

GPU Capabilities

For GPU providers, advertise models and their RAM/interface variants using the standardized capability key format. Valid keys are synced from the provider-configs GPU database.

GPU Attribute Format

capabilities/gpu/vendor/<vendor>/model/<model>
capabilities/gpu/vendor/<vendor>/model/<model>/ram/<size>
capabilities/gpu/vendor/<vendor>/model/<model>/interface/<iface>
capabilities/gpu/vendor/<vendor>/model/<model>/ram/<size>/interface/<iface>

Example for NVIDIA RTX 4090 (PCIe, 24Gi):

- key: capabilities/gpu/vendor/nvidia/model/rtx4090
  value: "true"
- key: capabilities/gpu/vendor/nvidia/model/rtx4090/ram/24Gi
  value: "true"
- key: capabilities/gpu/vendor/nvidia/model/rtx4090/interface/pcie
  value: "true"
- key: capabilities/gpu/vendor/nvidia/model/rtx4090/ram/24Gi/interface/pcie
  value: "true"

GPU Vendors

nvidia — NVIDIA GPUs
amd — AMD GPUs (for example mi100, mi60)

GPU Interface Types

pcie — PCIe connected (most common)
sxm — SXM form factor (data center)
mig — Multi-Instance GPU

RAM Sizes

Common values: 8Gi, 12Gi, 16Gi, 24Gi, 32Gi, 40Gi, 48Gi, 80Gi

Multiple GPU Models

If you have different GPU models across nodes, list each model with its RAM and interface keys:

attributes:
  # RTX 4090
  - key: capabilities/gpu/vendor/nvidia/model/rtx4090
    value: "true"
  - key: capabilities/gpu/vendor/nvidia/model/rtx4090/ram/24Gi
    value: "true"
  - key: capabilities/gpu/vendor/nvidia/model/rtx4090/interface/pcie
    value: "true"
  - key: capabilities/gpu/vendor/nvidia/model/rtx4090/ram/24Gi/interface/pcie
    value: "true"

  # RTX 3090
  - key: capabilities/gpu/vendor/nvidia/model/rtx3090
    value: "true"
  - key: capabilities/gpu/vendor/nvidia/model/rtx3090/ram/24Gi
    value: "true"
  - key: capabilities/gpu/vendor/nvidia/model/rtx3090/interface/pcie
    value: "true"
  - key: capabilities/gpu/vendor/nvidia/model/rtx3090/ram/24Gi/interface/pcie
    value: "true"

Important: Each node should only have one GPU model. Don’t mix GPU types on a single node.

Adding New GPU Models

If your GPU isn’t in the provider-configs database, submit it:

Visit the provider-configs repository
Follow the GPU submission process in the README
Wait for approval (typically 1–3 business days)
Update your provider attributes with the approved capability keys

Complete Example

Here’s a complete provider.yaml attributes section:

attributes:
  # Standard
  - key: host
    value: akash
  - key: tier
    value: community
  - key: organization
    value: my-gpu-provider
  - key: email
    value: [email protected]
  - key: discord-username
    value: mygpuprovider
  - key: website
    value: https://mygpuprovider.com

  # Location
  - key: location-region
    value: na-us-west
  - key: country
    value: US
  - key: city
    value: LAX
  - key: timezone
    value: utc-8
  - key: location-type
    value: datacenter
  - key: hosting-provider
    value: equinix

  # Hardware
  - key: capabilities/cpu
    value: amd
  - key: capabilities/cpu/arch
    value: x86-64
  - key: capabilities/gpu
    value: nvidia
  - key: capabilities/memory
    value: ddr5

  # Network
  - key: network-provider
    value: level3
  - key: network-speed-up
    value: 10000
  - key: network-speed-down
    value: 10000

  # Features
  - key: feat-persistent-storage
    value: true
  - key: feat-shm
    value: true
  - key: feat-endpoint-ip
    value: true
  - key: feat-endpoint-custom-domain
    value: true

  # Persistent Storage (required if you have Rook-Ceph)
  - key: capabilities/storage/1/class
    value: beta2
  - key: capabilities/storage/1/persistent
    value: "true"

  # SHM/Shared Memory (recommended for all providers)
  - key: capabilities/storage/2/class
    value: ram
  - key: capabilities/storage/2/persistent
    value: "false"

  # GPU Capabilities
  - key: capabilities/gpu/vendor/nvidia/model/rtx4090
    value: "true"
  - key: capabilities/gpu/vendor/nvidia/model/rtx4090/ram/24Gi
    value: "true"
  - key: capabilities/gpu/vendor/nvidia/model/rtx4090/interface/pcie
    value: "true"
  - key: capabilities/gpu/vendor/nvidia/model/rtx4090/ram/24Gi/interface/pcie
    value: "true"
  - key: cuda
    value: "12.7"

Verify Attributes

After updating your provider, verify attributes are registered:

provider-services query provider get <your-provider-address>

Look for your attributes in the attributes section of the output.

Provider Installation
GPU Support Setup
Provider Audit — Official audit process and attribute checklist
Provider Attributes Schema (canonical key list)
Provider Configs Repository