[heydryft]

Thanks for keeping the provider presets in one small, well-validated list: adding one was a single entry plus a test.

Summary

Adds a Runcrate inference preset. Runcrate is an AI cloud with an OpenAI-compatible chat completions API at https://api.runcrate.ai/v1 that serves open models. The preset uses the same openai-compatible protocol as Together, Fireworks and Novita, so no new request path is involved.

The default model is Qwen/Qwen3-235B-A22B-Instruct-2507. I picked it because its entry in GET https://api.runcrate.ai/v1/models lists tools in supported_features, and Browser Kitten needs tool calls for the Inference provider.

Changes

• src/agent/providerPresets.js: new runcrate entry in PROVIDER_PRESETS (base URL, default model, openai-compatible).
• tests/provider-presets.test.mjs: asserts the preset's id, name, base URL, protocol and default model.
• CHANGELOG.md: one line under Unreleased / Added.

Validation

• node --test tests/provider-presets.test.mjs: 8 passed, 0 failed.
• node --test tests/*.test.mjs: 814 passed, 0 failed.
• git diff --check: clean.
• I did not load the extension in Chrome or run the Playwright smoke test, and I did not send a live chat request through the extension.

Disclosure: I work on Runcrate.

This change was prepared with AI assistance (omp with Claude): the code, test, changelog line and this description were written with it, and the commands above were run as part of preparing it.