gemini-2.0-flash returns 404 for new API keys — default model should be updated to gemini-2.5-flash #9
Description
Hey there,
New users setting up ApplyPilot with a freshly generated Google AI Studio API key will encounter silent 404 errors during scoring. Every job scores 0, and the pipeline completes without error, making it appear successful even though it did nothing.
The root cause is that gemini-2.0-flash (the current default in .env) is no longer available to new Google AI Studio API keys. Google returns HTTP 404 with the message "This model is no longer available to new users," which ApplyPilot logs as a generic LLM error without surfacing the real reason to the user.
applypilot doctor reports the API key as OK, giving no indication that anything is wrong. The failure only becomes apparent when you notice all 2,000+ jobs scored 0.
To reproduce:
Generate a new Google AI Studio API key
Set LLM_MODEL=gemini-2.0-flash in .env
Run applypilot run -w 4
All jobs will score 0 silently
Fix: Update the default model in the documentation and/or .env template to gemini-2.5-flash, which works correctly with new API keys.