catalog

Get one offer (public)

get/api/v1/offers/{id}

Request

  • Base URL: https://amplerun-api.spiry.org
  • URL: https://amplerun-api.spiry.org/api/v1/offers/{id}
  • Auth: none declared

Path parameters

idstring uuid required
Example:3f0e2a55-8c1b-4d2e-9a6f-0b7c1d2e3f40

Response

Offer summary.

offer_idstring uuid required
machine_idstring uuid required
observed_atstring date-time required
offer_versionstring required

Unsigned base-10 integer as string; never a float.

expected_versionstring required

Unsigned base-10 integer as string; never a float.

readiness'ready' | 'provisioning' | 'unavailable' required
total_rate_micro_per_hourstring required

Unsigned base-10 integer as string; never a float.

available_fromstring date-time required
available_untilstring date-time required
gpu_countinteger
modelstring
vram_bytesstring

Unsigned base-10 integer as string; never a float.

regionstring
countrystring

Additive. ISO 3166-1 alpha-2 country of the machine, read by the controller from the agent's request country as reported by the trusted proxy (never from host input). Absent when unknown.

continent'AF' | 'AN' | 'AS' | 'EU' | 'NA' | 'OC' | 'SA'

Additive. Continent derived from country: AF Africa, AN Antarctica, AS Asia, EU Europe, NA North America, OC Oceania, SA South America. Absent when the country is unknown.

ram_bytesstring

Unsigned base-10 integer as string; never a float.

disk_bytesstring

Unsigned base-10 integer as string; never a float.

measured_bandwidth_mbpsinteger

netfacts lane. Measured DOWNLOAD rate in Mbps; additive. Absent = never measured (never 0).

net_up_mbpsinteger

netfacts lane. Measured UPLOAD rate in Mbps; rides the same measurement as measured_bandwidth_mbps.

net_latency_msinteger

netfacts lane. Median HTTPS latency in ms from the same measurement.

speedtest_atstring date-time
tflops_bf16number

AI Speed. Measured dense BF16 matmul TFLOPS per GPU (the MIN across the machine's GPUs); additive. Absent = never measured (never 0).

tflops_totalnumber

AI Speed. tflops_bf16 x gpu_count, computed, one decimal. Absent with tflops_bf16 or gpu_count.

mem_bw_gbpsnumber

AI Speed. Measured device memory bandwidth in GB/s per GPU (the MIN across the machine's GPUs). Absent = never measured (never 0).

ai_speedinteger

AI Speed. round(100sqrt((tflops_bf16/165.2)(mem_bw_gbps/1008))); RTX 4090 = 100. Absent unless both figures were measured.

bench_atstring date-time
bench_flag'unverified_spec'

AI Speed. The GPU has no datasheet row to check the figures against.

disk_free_bytesstring

Unsigned base-10 integer as string; never a float.

uptime_sstring

Unsigned base-10 integer as string; never a float.

gpu_utilization_pctnumber

Additive. Live GPU load, the PEAK across the machine's devices, from the latest telemetry sample. Absent under the same rules as disk_free_bytes, and when no device reported utilisation. 0 means measured idle.

telemetry_atstring date-time
template_compatibilitystring[]
manifest_hashstring
fits_template_idsUUID[]

Additive. Present only when the request carried fits_for=all; the ids of every published template that fits this offer (the same rule as fit).

platform_fee_bpsinteger

Additive. Platform fee frozen into this offer version, in basis points (500 = 5%), taken out of each charge.

verification_level'unverified' | 'auto_verified' | 'manually_verified'

Additive. How the machine's hardware claims were verified.

admittedboolean

Additive. Whether the machine has passed admission (a passed qualification for its current manifest). false is an unverified listing: shown in the catalog, never quoted or reserved (409 CONFLICT).

tier'container' | 'inference'

Additive (D-23). container: SSH containers under the launch profile. inference: managed models on a signed, hash-pinned runtime with no renter code, reached over the host's outbound session (community inference: don't send sensitive data).

backendsstring[]

Additive. Backends the machine's devices serve, in preference order (cuda > rocm > sycl > vulkan > metal). A template runs here with the first one it has a runtime for.

availability_class'dedicated' | 'interruptible'

D-22. dedicated (the default) is never preempted. interruptible is sold only from idle or co-listed hosts: the host may take the machine back after preemption_notice_s seconds of notice, and billing stops at the notice.

preemption_notice_sinteger nullable

D-22. Notice in seconds (SIGTERM plus /work/.amplerun-preempt) before the host stops an interruptible rental; null for dedicated.

Example response

{
  "offer_id": "8b0c2f4e-1d3a-4c5b-9e6f-7a8b9c0d1e2f",
  "machine_id": "5d6e7f80-9a1b-4c2d-8e3f-4a5b6c7d8e9f",
  "observed_at": "2026-09-25T12:00:00Z",
  "offer_version": "3",
  "expected_version": "12",
  "readiness": "ready",
  "total_rate_micro_per_hour": "450000",
  "platform_fee_bps": 500,
  "included_quotas": {
    "included_disk_gib": 100,
    "included_egress_gib": 50
  },
  "available_from": "2026-09-25T00:00:00Z",
  "available_until": "2026-10-25T00:00:00Z",
  "maintenance_windows": [],
  "gpu_count": 1,
  "model": "NVIDIA GeForce RTX 4090",
  "vram_bytes": "25757220864",
  "region": "us-east-1",
  "manifest_hash": "sha256:9f86d081884c7d659a2feaa0c55ad015a3bf4f1b2b0b822cd15d6c15b0f00a08",
  "verification_level": "auto_verified",
  "admitted": true,
  "fit": {
    "fits": true,
    "reason": "ok"
  }
}

Changes

No recorded changes to this operation across all 1 revision of this API.