Migrate plugins gpt-4 hardcoded calls to gpt-4.1-mini (3 files, ~$280/day) #4690
Description
Problem
Three plugin notification files hardcode model="gpt-4" (resolves to gpt-4-0613 at $30/1M input). These are the remaining source of ~400 reqs/hr of gpt-4-0613 calls traced to the "Omi1 v2" API key, costing ~$280/day.
PR #4675 fixed the backend copy (backend/utils/mentor_notifications.py) but these plugin copies were never updated.
Files to change
| File | Line | Function |
|---|---|---|
plugins/example/notifications/mentor/main.py |
:102 | extract_topics() — highest volume, runs every notification cycle |
plugins/example/notifications/hey_omi.py |
:122 | get_openai_response() |
plugins/example/notifications/drinking_app.py |
:51 | drinking intent detection |
Fix
Change model="gpt-4" → model="gpt-4.1-mini" in all 3 files. Same migration as PR #4675 did for the backend copy.
These are simple classification/extraction tasks (topic extraction, yes/no detection, short Q&A) — gpt-4.1-mini handles them fine.
Impact
- ~$280/day → ~$2/day (gpt-4.1-mini is ~99% cheaper for these call patterns)
- ~400 reqs/hr unaffected (same volume, cheaper model)
Related
- Migrate mentor notification topic extraction from gpt-4 to gpt-4.1-mini #4671 (closed) — same migration for backend copy
- PR fix: migrate mentor topic extraction from gpt-4 to gpt-4.1-mini (#4671) #4675 (merged) — backend fix that missed these plugin copies