A minimal, provider-neutral sandbox for reproducible GPT and AI experiments.
Project status: starter repository. The initial roadmap is intentionally open so contributors can help shape a small, testable foundation.
- Keep experiments easy to understand and reproduce.
- Separate providers from experiment logic.
- Store prompts and expected outputs as reviewable fixtures.
- Avoid committing secrets, private conversations, or large generated artifacts.
- Document limitations and evaluation methods alongside results.
- a provider-neutral example interface, prompt regression fixtures, evaluation utilities, and a secure environment-variable template.
- A lightweight project structure and test runner.
- A documented
.env.examplecontaining names only, never credentials. - One small end-to-end example with expected output.
Read CONTRIBUTING.md before proposing a new framework or provider.
Report vulnerabilities privately as described in SECURITY.md.
Released under the MIT License.