feat(loop-init): add cheaperinference model provider

[aiapienthusiast]

Summary

This PR adds Cheaper Inference as a model provider in loop-init.
It uses the same pattern as #569 (OrcaRouter).

Cheaper Inference is an OpenAI-compatible LLM gateway.
One API key gives access to models from several labs.

Usage:

npx @cobusgreyling/loop-init . -p ci-sweeper -t claude --with-foundry --model-provider cheaperinference
npx @cobusgreyling/loop-init . -p ci-sweeper -t claude --with-foundry --model-provider cheaperinference --model claude-sonnet-5

The default model is gpt-5.4-mini.
Set CHEAPER_INFERENCE_API_KEY to a key from https://cheaperinference.com/signup.

Related issue

None.

Changes

• New pattern or starter (followed templates/pattern-template.md + updated registry.yaml)
• Doc / example / story improvement
• Tool / CLI change under tools/
• Test only
• Other

Checklist (from CONTRIBUTING)

• Links work from README or the relevant index
• No secrets, tokens, or internal company URLs
• STATE.md* examples use .example suffix
• Safety-related content references docs/safety.md
• If you touched a package under tools/<pkg>: cd tools/<pkg> && npm ci && npm test
• If you touched the registry: root npm ci && npm run validate:registry

Testing / Dogfood

• npx @cobusgreyling/loop audit . (or package-local tests) on affected paths
• Manual review of generated state / skill output if scaffolding changed

Details

Files:

• tools/loop-init/src/cli.ts: add the cheaperinference provider and the model/cheaperinference primitive.
• tools/loop-init/dist/cli.js: rebuilt with npm run build.
• tools/loop-init/README.md: document the flag, models and endpoints.
• tools/loop-init/test/cli.test.mjs: 4 new tests, same as the OrcaRouter tests.

The primitive writes these endpoints:

• OpenAI-compatible: https://api.cheaperinference.com/v1
• Anthropic-compatible: https://api.cheaperinference.com (clients add /v1/messages)

Model options: gpt-5.4-mini, gpt-5.4, claude-sonnet-5, gemini-3.1-pro, deepseek-v4-flash, glm-5.3.

Tests:

• cd tools/loop-init && npm ci && npm test: 68 of 69 pass. All 4 new tests pass.
• The one failure is loop-init prints Loop Ready score after scaffold. It also fails on main in my environment.
• npm run check:loop-init: OK.

Live check: gpt-5.4-mini returned a reply through both endpoints in the generated .foundry/stack.yaml.
All model ids in the primitive are in the live GET /v1/models list.

[github-actions]

This PR changes paths that must run the real validate and audit workflows (tools, patterns, scripts, or CI).

Fork PRs from first-time contributors start with those workflows waiting for approval. A maintainer needs to open the Checks tab and click Approve and run workflows. Until that happens, branch protection will show the PR as blocked even after a review.

Content-only PRs (docs/, examples/, stories/, skills/, root markdown) skip this step — required checks are posted from this workflow instead.

— loop-engineering fork-pr-gate