A critical configuration flaw exists in the current codebase: non-standardized environment variable names are used for the Gemini API key, which introduces fragility into the model initialization and API communication pipeline.
Different modules expect inconsistent identifiers for the same environment variable:
Some modules reference GEMINI_API_KEY
Others (e.g., model_loader.py) reference GOOGLE_API_KEY
This inconsistency violates 12-factor app principles regarding strict config uniformity and creates a hidden source of runtime failure, especially in containerized environments (e.g., Docker, Kubernetes) or CI/CD pipelines.