Prevent potential AttributeError when accessing reasoning_model #34
Description
Description of the bug:
Currently, there's a potential for an AttributeError to occur within the agent's graph logic when configurable.reasoning_model is accessed. This issue arises specifically if the reasoning_model field is not explicitly defined in the Configuration class located at backend/src/agent/configuration.py.
The line causing the potential issue is in backend/src/agent/graph.py:
reasoning_model = state.get("reasoning_model") or configurable.reasoning_model
If Configuration is initialized without reasoning_model being explicitly set (e.g., through environment variables or RunnableConfig), configurable.reasoning_model becomes an invalid attribute. This could lead to a runtime AttributeError and cause the application to crash if a value for reasoning_model is not provided through other means.
Actual vs expected behavior:
When reasoning_model is not defined in Configuration and not provided via state or environment variables, accessing configurable.reasoning_model in backend/src/agent/graph.py results in an AttributeError. This causes the application to fail at runtime.
Expected Behavior:
configurable.reasoning_model should always return a valid string value. This ensures that the agent's reasoning model can be reliably configured without causing a runtime error, even if an explicit value isn't provided during configuration. It should gracefully default to a sensible value if no other configuration is given.
Any other information you'd like to share?
No response