Silent Deprecation: Discovering a Dead SDK Through Startup Warnings

LEO's backend used the google.generativeai Python package to call the Gemini API. Google deprecated this package and created a replacement: google.genai. The startup log captured this deprecation warning firing every time the application started.

From backend/startup.log (lines 22-30):

FutureWarning: google.generativeai package is deprecated. Please install google-genai instead.
  import google.generativeai as genai
...
FutureWarning was raised and it indicates the library you are using is deprecated
and may be removed from a future release. Please install google-genai library.

The warning was emitted from within the package at import time, appearing before any application initialization. This warning fires on every startup, creating persistent noise in startup logs.

• backend/startup.log lines 22-30 — FutureWarning from google.generativeai
• backend/requirements.txt — google-generativeai in the dependencies list
• Absence of google-genai in requirements.txt

Established: google.generativeai is deprecated. The startup log confirms the warning fires on every startup. google-genai is not in the requirements.

Likely: The migration from google.generativeai to google.genai requires API changes beyond a simple import swap — the two packages have different class names and call signatures. This explains why the migration was not a quick fix.

Unknown: Whether the application would break immediately upon package removal, or whether the deprecated package would continue to function until explicitly removed from PyPI.

No documented migration.

SDK deprecation warnings in startup logs are easy to dismiss but should be tracked as technical debt. A FutureWarning on every startup creates log noise that can mask other important warnings. The migration cost is often higher than the import change alone — API differences between old and new SDKs require code-level updates throughout.

SDK deprecation warnings should trigger a scheduled migration task, not just log acknowledgment. The window between deprecation warning and removal is finite, and the longer a migration is deferred, the more disruptive the eventual forced upgrade becomes.

• backend/startup.log (lines 22-30)
• backend/requirements.txt