Repository navigation
Description
Heating POST requests start returning 403 during sequential multi-zone control
Environment
- Home Assistant Core 2026.11.1, built-in Hive integration; no custom Hive component.
- That Core release pins
pyhive-integration==1.0.10. Installed dependency files have not been independently inspected. - Multiple Hive radiator valves and a Hive thermostat. The official Hive app remains usable.
- No account identifiers, device identifiers, credentials or unredacted diagnostics are included here.
Observed behaviour
An individually issued target change and a separate Manual-mode change were independently confirmed at a physical radiator valve and in the Hive app. Return to Schedule was also confirmed, allowing for device delivery delay.
A later multi-zone sequence used ordinary Home Assistant climate services:
- Set a radiator to
heat(Hive Manual). - Set its target temperature.
- Continue to the next radiator.
Controller writes were sequential, with at least five seconds between writes and additional room-level pauses. Normal integration polling remained enabled.
During the captured failure, three valves completed both calls and a fourth completed its mode call. The fourth valve's temperature call then raised HTTP 403. The retained HA script trace locates the error at that temperature service call. Four subsequent attempts to return the affected valves to Auto/Schedule also received HTTP 403. The operator restored the valves through the official Hive app.
The logged response was:
HTTP 403
{"message":"Forbidden"}
HiveAuthError: Token expired or forbidden calling [redacted URL]
This exception text does not establish actual token expiry. No HTTP 429 was observed, and the sequence does not establish a fixed request-count limit or throttling cause.
HA continued to report the entities available during the failure. That is not proof that contemporaneous cloud GET requests succeeded; cached state may be involved. A later integration reload recovered fresh state matching the restored schedules. No further device writes were attempted after that final recovery.
Source observations
Read-only inspection of the published 1.0.10 wheel found that setState() lets POST HiveAuthError escape without rejected-write authentication recovery. Heating setters check token age before writing. In contrast, getDevices() has a GET authentication-rejection login/retry path.
The published 2.0.0b5 wheel was also inspected without installing it. Its state-change path likewise exposes POST 401/403 without a forced-refresh/retry path. Its request headers match the inspected stable version. This is a source observation, not a claim that either behaviour causes this incident or that a beta upgrade fixes it.
References: Core Hive manifest, stable distribution, beta distribution.
Guidance requested
- How should a client distinguish invalid/revoked authentication, permission rejection and throttling when the response is this generic 403?
- Is there a supported authentication-recovery or pacing policy for heating writes?
- Which response headers or additional diagnostics can safely clarify the rejection? Any collection would exclude tokens, cookies, account/device identifiers and unredacted URLs.
Automatic control has been disabled pending reliable writes and restoration. No blind 403 replay, beta upgrade or custom authentication patch has been deployed.
Activity