ractogateway.exceptions

Unified exception hierarchy for RactoGateway.

All provider-specific SDK errors (openai.APITimeoutError, anthropic.APITimeoutError, google.api_core.exceptions.*, etc.) are caught by the adapters and re-raised as one of these classes so callers never need to import provider packages just to handle errors.

Example:

from ractogateway.exceptions import RactoGatewayTimeoutError

try:
    response = kit.chat(config)
except RactoGatewayTimeoutError:
    # Retry or surface a user-friendly message.
    ...

exception ractogateway.exceptions.RactoGatewayError[source]

Bases: Exception

Base class for every RactoGateway runtime error.

exception ractogateway.exceptions.RactoGatewayTimeoutError[source]

Bases: RactoGatewayError

The upstream provider did not respond within the allowed time.

exception ractogateway.exceptions.RactoGatewayAPIError(message, *, status_code=None)[source]

Bases: RactoGatewayError

The upstream provider returned an error response.

status_code: HTTP status code returned by the provider, when available.

exception ractogateway.exceptions.RactoGatewayAuthError[source]

Bases: RactoGatewayError

API key missing, invalid, or not authorised for the requested resource.

exception ractogateway.exceptions.ResponseModelValidationError(message, *, attempts, last_error, raw_response=None)[source]

Bases: RactoGatewayError

Raised when response_model validation cannot be satisfied.

The LLM returned structurally valid JSON but Pydantic rejected it (e.g. a field value was out of range) and all automatic retry attempts were exhausted.

attempts: Total number of API calls made (1 initial + N retries).

last_error: The final pydantic.ValidationError that triggered this exception.

raw_response: The LLM’s raw text from the last attempt, available for inspection or manual recovery.

Example::

from ractogateway.exceptions import ResponseModelValidationError

try:: response = kit.chat(config)
except ResponseModelValidationError as e:: print(e.last_error) # Pydantic ValidationError details print(e.raw_response) # raw JSON string from the LLM print(e.attempts) # how many times we tried