Hi LiteLLM team,

I recently published [`llm-auth`](https://github.com/omry/llm-auth), a small MIT-licensed local credential manager for LLM developer tooling. The use case is managing multiple named LLM auth surfaces for local scripts, benchmarks, agents, evals, and experiments without scattering credentials across many ad hoc `.env` files or provider-specific token locations.

Today `llm-auth` supports:

- API-key surfaces, for example `research=openai` or `evals=openrouter`. - ChatGPT subscription OAuth via LiteLLM's ChatGPT provider. - Redacted `status`. - Lightweight `test`. - `renew` for refreshable OAuth state. - Repo-local `.env` metadata envelopes that describe provider, model, API base, env var, and auth mode.

Example:

```bash llm-auth add-api-key research openai --model gpt-4.1-mini llm-auth login chatgpt llm-auth status llm-auth test research ```

The broader use case is that local LLM apps often need a stable abstraction like:

```text surface name -> provider -> auth mode -> model/API base -> secret location ```

rather than every tool needing to know directly whether credentials came from `OPENAI_API_KEY`, a ChatGPT OAuth token, OpenRouter, a project `.env`, or som…