fix: allow npm consumers to install the package

[Silu Panda (SiluPanda)]

Fixes #212

Summary

• remove package-manager restrictions from the published engines metadata so npm consumers with engine-strict=true can install autoevals
• keep repository development pinned to pnpm through the existing packageManager, root-only preinstall check, and lockfile CI gate

npm validates every published engines entry for consumers, so the pnpm/npm/yarn-only metadata made a normal dependency install fail before the package could be used.

Validation

• fail-first: packed current main, then installed the tarball from an npm consumer with engine-strict=true; npm 11.6.2 failed with EBADENGINE
• packed this branch and repeated the same consumer install; installation succeeded and import("autoevals") exposed ExactMatch
• pnpm exec vitest run on the 9 offline JS test files: 66 passed
• pnpm run build
• git diff --check

The remaining live-API Vitest files require OpenAI/Braintrust credentials and were not run locally.

Disclosure: this contribution was prepared with OpenAI Codex.

[Frieder Bluemle (friederbluemle)]

Can someone please review and merge this? I'm seeing this warning in a project that uses npm:

npm warn EBADENGINE Unsupported engine {
npm warn EBADENGINE   package: 'autoevals@0.3.0',
npm warn EBADENGINE   required: {
npm warn EBADENGINE     npm: 'please-use-pnpm',
npm warn EBADENGINE     pnpm: '>=10.27.0',
npm warn EBADENGINE     yarn: 'please-use-pnpm'
npm warn EBADENGINE   },
npm warn EBADENGINE   current: { node: 'v24.19.0', npm: '11.17.0' }
npm warn EBADENGINE }

[Silu Panda (SiluPanda)]

Abhijeet Prasad (@AbhiPrasad), could you review this npm-install fix when you have a chance? Another npm consumer has requested review and merge after encountering the same EBADENGINE warning. Thank you!

[github-actions]

Braintrust eval report

Autoevals (main-1790175338)

Score	Average	Improvements	Regressions
NumericDiff	78.5% (+1pp)	8 🟢	10 🔴
Time_to_first_token	11.06tok (+1.3tok)	66 🟢	147 🔴
Llm_calls	1.55 (+0)	-	-
Tool_calls	0 (+0)	-	-
Errors	0 (+0)	-	-
Llm_errors	0 (+0)	-	-
Tool_errors	0 (+0)	-	-
Prompt_tokens	511.28tok (+8.77tok)	32 🟢	28 🔴
Prompt_cached_tokens	0tok (+0tok)	-	-
Prompt_cache_creation_tokens	0tok (+0tok)	-	-
Prompt_cache_creation_5m_tokens	0tok (+0tok)	-	-
Prompt_cache_creation_1h_tokens	0tok (+0tok)	-	-
Completion_tokens	473.64tok (+20.87tok)	103 🟢	107 🔴
Completion_reasoning_tokens	354.91tok (+15.42tok)	82 🟢	89 🔴
Completion_accepted_prediction_tokens	0tok (+0tok)	-	-
Completion_rejected_prediction_tokens	0tok (+0tok)	-	-
Completion_audio_tokens	0tok (+0tok)	-	-
Total_tokens	984.91tok (+29.65tok)	106 🟢	105 🔴
Estimated_cost	0$ (+0$)	70 🟢	64 🔴
Duration	11.07s (+1.33s)	67 🟢	148 🔴
Llm_duration	13.9s (+3.5s)	57 🟢	157 🔴