🎉 BOOTH was featured by Python Hub (and a thank you)

[Vedantgitbot]

BOOTH was recently featured by [Python Hub](https://x.com/PythonHub/status/2114752745504686172), just a couple of weeks after its [Python Weekly feature](#20).

It's a small but meaningful milestone for a project that's only three months old. More importantly, BOOTH wouldn't be where it is today without the people who contributed code, examples, experiments, feedback, and ideas along the way.

🙌 A big thank you to our contributors

I want to take a moment to recognize everyone who has helped BOOTH grow:

• @AkashGoswamy — Contributed the Gemini example, which exposed the gap that eventually became call_failed in v0.5.4.
• @opencode13241-eng — Contributed the Anthropic example.
• @sohanpal17 — Contributed the Ollama / Llama example.
• @AK-Lmn — Added py.typed, making BOOTH's type hints available to type checkers.
• @camerontjs-dot — Contributed the keyword-matcher vs. NLI comparator example, including a failure case intentionally left in.
• @Harithalakshmikt — Contributed the LangChain + FAISS + FLAN-T5 example.
• @rk042 — Contributed the mypy CI job and worked on parse_fn (Add parse_fn for structured output (bypass regex/raw_decode JSON scraping) #6).
• @Emart29 — Contributed the confidence-calibration experiment (Measure whether BOOTH's self-reported confidence predicts correct answers (calibration example) #24) and detailed design notes on booth.measure() (Feature Proposal: Add booth.measure() for offline evaluation & confidence calibration benchmarking #21).

Every contribution matters. I'm genuinely grateful to have people willing to test ideas, question assumptions, share results, and help make the project better.

Thank you all! ❤️

🚀 What's happened since the last update?

Quite a bit has changed since the previous update:

• v0.5.4 → v0.5.6: Introduced call_failed, generator and iterator evidence, an upfront check for non-callable validator arguments, and Iterable[str] type hints.
• More real-world examples: Examples with actual output and honest results, including negative ones, now cover Groq, Gemini, Anthropic, Ollama, LangChain, and two comparator studies.
• Better type checking: mypy now runs in CI.
• Confidence calibration: Early results from Emart's experiment suggest that self-reported confidence barely separates correct answers from incorrect ones on small local models. This reinforces the motivation behind evidence-based checks. The findings are being documented in Measure whether BOOTH's self-reported confidence predicts correct answers (calibration example) #24.

🤝 Where community help is most useful

BOOTH is still evolving, and there's plenty of room for new ideas, testing, and contributions.

Here are a few ways you can help:

• Try BOOTH with your provider or model. Share actual output, including failures. Check out the [examples/](https://github.com/Vedantgitbot/booth/tree/main/examples) directory.
• Report edge cases. Minimal, reproducible examples are especially helpful. See [CONTRIBUTING.md](https://github.com/Vedantgitbot/booth/blob/main/CONTRIBUTING.md) for contribution guidelines.
• Join the design discussions. Your feedback is welcome in Proposal: independent-sample agreement check #19 (independent-sample agreement) and Feature Proposal: Add booth.measure() for offline evaluation & confidence calibration benchmarking #21 (booth.measure()).

💬 A question for the community

Which provider and model are you currently using with check() or check_with_evidence()?

And more importantly, what surprised you?

Did something work better than expected? Did you encounter an unexpected failure or an interesting edge case?

Even a simple "it worked fine on X" is useful feedback.

Once again, thank you to everyone who has contributed to BOOTH, whether through code, experiments, examples, feedback, or discussions.

This project is still young, and there's a lot more to explore. Looking forward to building what's next together! 🚀

---

🔗 Project: https://github.com/Vedantgitbot/booth

🐍 Python Hub feature: https://x.com/PythonHub/status/2114752745504686172

📰 Python Weekly feature: #20