Return typed Pydantic models from external API calls and centralize request logging by status class

[Copilot]

External API responses were being propagated as untyped dicts, requiring repeated ad-hoc parsing in services. This change introduces centralized response-model validation and consistent per-request logging with status-based severity (2xx→INFO, 4xx/5xx→ERROR).

• Centralized response parsing and validation

  • Extended http_utils.extract_response_data(...) with an optional model parameter.
  • When provided, response payloads are validated/converted via pydantic.TypeAdapter, enabling direct return of typed models (including collections like List[official.Pet]).
  • Validation errors now surface as structured 500s with schema mismatch detail.

• Centralized external request logging

  • Added request logging in http_utils.raise_for_status(...) using response metadata (method, url, status_code).
  • Enforced log levels:
    • INFO for 2xx
    • ERROR for 4xx/5xx

• Service-layer adoption of typed models

  • Updated external API service methods to pass explicit model types into extract_response_data(...) across:
    • dashboard, reports, households, devices, pets
  • Removed now-redundant local re-validation paths where responses are already typed.
  • Routed auth and timeline error handling through shared HTTP utilities to ensure logging and status handling are consistent.

• Related consistency cleanup

  • Normalized endpoint usage in devices service to settings.endpoint for all device/tag calls.

# Before
data = http_utils.extract_response_data(response)          # dict/list payload

# After
pet = http_utils.extract_response_data(response, model=official.Pet)
pets = http_utils.extract_response_data(response, model=List[official.Pet])

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.