Scaffold a provider-neutral experiment interface and first regression fixture

[112-stack]

Context

GPT Lab is a starter repository intended for small, reproducible AI experiments. It needs a minimal foundation that does not lock contributors into one API provider.

Goal

Create the first provider-neutral experiment contract and one deterministic fixture.

Suggested scope

• Define a small interface for sending an input and receiving normalized text plus metadata.
• Add a local fake provider for tests; no network access should be required.
• Store one prompt fixture and expected normalized result.
• Add a test runner and one documented command.
• Add .env.example containing variable names only if environment configuration is needed.

Acceptance criteria

• Tests pass without API keys or network access.
• No real conversations, secrets, or provider credentials are committed.
• Provider-specific response fields do not leak into experiment assertions.
• The README explains how to add another provider.
• The implementation stays intentionally small and well documented.

[Om-Beast]

Hi @112-stack! I'd like to work on this issue if it's available.

My plan is to keep the implementation intentionally small and provider-neutral by introducing a minimal experiment interface, a deterministic fake provider for local testing, a regression fixture, and documentation describing how additional providers can be added without changing the core contract.

I'll ensure the test suite runs without API keys, network access, or provider-specific assertions.

Please let me know if this aligns with your expectations.