Setup fails on servers with no array disks (null parity fields, string capacity values)

[itskevinb]

Description

Integration v1.5.1 fails with "Received invalid data from Unraid" on servers that have no array disks and have never run a parity check. The GraphQL API returns types that the Pydantic models don't accept.

Additionally, the call_api method in api/__init__.py has no timeout on the session.post() call. When HA's event loop is busy during startup, the setup task gets cancelled (asyncio.CancelledError / "Setup cancelled") before the request completes.

Environment

• Home Assistant 2026.4.1
• Unraid API integration v1.5.1
• Unraid 7.2.4 (server with no array disks configured)
• Unraid API version 4.32.1

Root Cause

The GraphQL API on a server with no array disks returns:

{
  "array": {
    "capacity": {
      "kilobytes": {
        "free": "0",    // string, not int
        "used": "0",    // string, not int
        "total": "0"    // string, not int
      }
    },
    "parityCheckStatus": {
      "date": null,        // null, model expects datetime
      "duration": null,    // null, model expects int
      "speed": "0",        // string, model expects float
      "status": "NEVER_RUN",
      "errors": null,
      "progress": 0
    }
  }
}

The strict Pydantic models in api/v4_20.py reject these types, causing a ValidationError which is caught as UnraidApiInvalidResponseError.

Fix

Three files need changes:

1. api/__init__.py — Add timeout to call_api

from aiohttp import ClientTimeout

# In call_api method:
response = await self.session.post(
    self.endpoint,
    json={"query": query, "variables": variables or {}},
    headers={...},
    timeout=ClientTimeout(total=30),  # Add this
)

2. api/v4_20.py — Widen model types

class ArrayCapacityKilobytes(BaseModel):
    free: int | str    # was: int
    used: int | str    # was: int
    total: int | str   # was: int

class ParityCheck(BaseModel):
    date: datetime | None    # was: datetime
    duration: int | None     # was: int
    speed: float | str       # was: float
    status: ParityCheckStatus
    errors: int | None
    progress: int

And convert in query_metrics_array:

capacity_free=int(response.array.capacity.kilobytes.free),
capacity_used=int(response.array.capacity.kilobytes.used),
capacity_total=int(response.array.capacity.kilobytes.total),
parity_check_speed=float(response.array.parity_check.speed) if response.array.parity_check.speed is not None else 0.0,

3. models.py — Match nullable types

@dataclass
class MetricsArray:
    ...
    capacity_free: int | str    # was: int
    capacity_used: int | str    # was: int
    capacity_total: int | str   # was: int
    ...
    parity_check_date: datetime | None    # was: datetime
    parity_check_duration: int | None     # was: int
    parity_check_speed: float | str       # was: float

The same changes apply to api/v4_26.py which inherits from v4_20.

Workaround

Apply the patches above manually inside the HA container at /config/custom_components/unraid_api/. They will be overwritten on the next HACS update.

[AquaCobell]

same here