AiKey: AI FinOps Infrastructure — Virtual Keys, Security Vault & Anti-Nerfing

[slim steppe]

Hey devs! I'm sharing AiKey, a project I built to move AI credential management from the "Stone Age" to a proper FinOps architecture.

The Pain Points:

Attribution: Who burned those tokens? Traditional dashboards don't tell you at a project/team level.
Model Nerfing: Are you actually getting the flagship performance you're paying for?
Security: Sharing .env files with raw master keys is a disaster for scaling teams.
How AiKey Fixes This:

Virtual Key Orchestration: Issue metadata-tagged keys for instant cost attribution.
Quality Radar: Protocol-level verification to ensure the responding model matches your request (no more silent downgrades).
Vault & Runtime Injection: Keep master keys secure and inject them dynamically via CLI with zero code changes.
GitHub: https://github.com/aikeylabs/launch

I'm around if you have questions about the implementation or how to set up AI cost governance for your team!

[sullen flint]

This is timely — I was just discussing per-node cost attribution in LangGraph pipelines yesterday and the lack of proper tooling is a real gap.
The model nerfing point is interesting — have you built in any way to benchmark actual vs expected model performance so you can detect when a provider quietly downgrades the model behind the same API endpoint?