Returns all running GPU servers. Includes IP addresses, status, flavor, and expiry time if uptime limits are configured.

{
  "servers": [
    {
      "id": "3f9a1c7e-0b2d-4e6f-8a1b-2c3d4e5f6a7b",
      "name": "my-project-0001",
      "status": "ACTIVE",
      "flavor": "t1-45",
      "ip": ["51.91.120.42"],
      "region": "GRA11",
      "created": "2026-06-07T09:14:03Z",
      "expires_at": "2026-06-08T09:14:03Z",
      "image_name": "ubuntu-22.04-cuda",
      "provision_state": "succeeded",
      "provision_error": null,
      "provision_attempts": 1
    }
  ]
}

provision_state moves through creating → queued → running → succeeded, or ends at create_failed / failed. ip is empty until the instance gets an address; expires_at is null unless an uptime limit is set.

Returns a single server by its ID — the same object shape as an entry in GET /servers. Returns 404 if the ID is not in this project.

{
  "id": "abc123-def456",
  "name": "my-project-0001",
  "status": "ACTIVE",
  "flavor": "t1-45",
  "ip": ["51.91.120.42"],
  "region": "GRA11",
  "created": "2026-06-07T09:14:03Z",
  "expires_at": "2026-06-08T09:14:03Z",
  "image_name": "ubuntu-22.04-cuda",
  "provision_state": "succeeded",
  "provision_error": null,
  "provision_attempts": 1
}

Provision one or more GPU instances. Returns immediately with a list of server objects containing each server's id and provision_state: "creating". IP address and final status become available by polling GET /servers.

{
  "servers": [
    { "id": "3f9a1c7e-0b2d-4e6f-8a1b-2c3d4e5f6a7b", "provision_state": "creating" }
  ]
}

Returns one object per instance immediately. Poll GET /servers for the IP address and final status.

Adds extra hours to a server's uptime limit, pushing back its expires_at. Hours accumulate across repeated calls. Only meaningful when an uptime limit is configured.

{ "server_id": "abc123-def456", "extra_hours": 2.0 }

extra_hours is the running total added to this server so far.

Runs an installation template on an already-running server. The template must be enabled for the project (see GET /templates). The call first confirms SSH access (may take a few seconds), then runs the install in the background — poll GET /servers/{id}/logs for progress.

{
  "status": "installation_started",
  "server_id": "abc123-def456",
  "template": "vllm"
}

Returns 409 if the server is still being created, or 400 if the template is not enabled or SSH authentication to the server fails.

Returns the provisioning and installation log lines for a server, oldest first. Useful for tracking POST /servers and POST /servers/{id}/install progress.

{
  "logs": [
    {
      "id": 1,
      "occurred_at": "2026-06-07T09:14:05Z",
      "level": "info",
      "message": "Provisioned by alice"
    },
    {
      "id": 2,
      "occurred_at": "2026-06-07T09:16:22Z",
      "level": "info",
      "message": "[template:vllm] Installation finished"
    }
  ]
}

Permanently deletes a GPU instance by its server ID (the id returned from POST /servers or GET /servers).

{ "deleted": "abc123-def456" }

If the server is still being created, deletion is deferred and the response is { "deleted": "<id>", "status": "deleting" }. Unknown IDs return 404.

Returns the list of images available for server creation. Use the name field in POST /servers.

{
  "images": [
    {
      "name": "ubuntu-22.04-cuda",
      "ovh_id": "8f1e2d3c-4b5a-6789-0abc-def123456789",
      "pretty_name": "Ubuntu 22.04 + CUDA"
    }
  ]
}

Pass the name value as the image field in POST /servers.

Returns the list of GPU flavors available for server creation including vCPUs, RAM, GPU type, and hourly price. Use the name field in POST /servers.

{
  "flavors": [
    {
      "name": "t1-45",
      "vcpus": 8,
      "ram_gb": 45,
      "disk_gb": 400,
      "gpu_type": "Tesla V100",
      "gpu_count": 1,
      "gpu_vram_gb": 16,
      "price_hour": 1.97
    }
  ]
}

Pass the name value as the flavor field in POST /servers. price_hour is in EUR and may be null if unknown.

Returns the installation templates enabled for this project. Use the folder value as the template field in POST /servers.

{
  "templates": [
    { "folder": "vllm", "name": "vLLM", "description": "vLLM inference server" }
  ],
  "configured": true
}

Pass the folder value as the template field in POST /servers. configured is false when templates are not enabled on this deployment.

Returns current month spend and end-of-month forecast from OVH, broken down by instance.

{
  "ovh_billing": {
    "period": { "from": "2026-06-01T00:00:00Z", "to": "2026-06-07T00:00:00Z" },
    "last_update": "2026-06-07T08:00:00Z",
    "current_total": 142.87,
    "forecast_total": 612.40,
    "instances": [],
    "resources_usage": []
  },
  "internal_billing": {
    "month_start": "2026-06-01T00:00:00+00:00",
    "total_spend": 138.52,
    "per_actor": [
      { "actor": "alice", "spend": 92.10 },
      { "actor": "ci-key", "spend": 46.42 }
    ]
  }
}

ovh_billing covers the whole OVH project; internal_billing is this project's tracked spend for the current month, per actor. If OVH is unreachable, ovh_billing is { "error": "Failed to fetch OVH billing" }.