[nicho810]

Adds cheapaimodel as a built-in provider card in the extended catalog.
It uses the same pattern as the Cheaper Inference card (#3268) and the Opper card (#3278).
cheapaimodel is an OpenAI-compatible, pay-per-token gateway for open-weight models (DeepSeek, GLM, Kimi) and Grok, prepaid in USD.

• Card id: cheapaimodel, category router, disabled by default.
• Base URL: https://api.cheapaimodel.ai/v1.
• Default model: deepseek-v4-1-flash (1000000 tokens), vision on. Suggestions: deepseek-v4-1-flash, glm-5.3-flash, glm-5.3, grok-4.7 (bare ids as listed on https://cheapaimodel.ai/models, 2026-11-02).
• Vision depends on the model: use deepseek-v4-1-flash or grok-4.7. glm-5.3 is text-only, and glm-5.3-flash rejects WebBrain's vision request (see Live test). The docs section says so, as the Atlas Cloud section does.
• Chrome and Firefox catalogs stay identical.

Known limitation: kimi-k3 is also served, but it only accepts temperature 1 or 0.6, and the OpenAI-compatible path sends options.temperature ?? 0.7 unless the model is on the default-temperature list. So kimi-k3 is left out of the suggestions and the docs note it. This PR does not change the request code.

Files

• src/{chrome,firefox}/src/providers/provider-catalog.js: one new entry, before cheaperinference (alphabetical).
• src/{chrome,firefox}/icons/providers/cheapaimodel.svg: cheapaimodel logo mark (currentColor).
• src/{chrome,firefox}/icons/providers/SOURCES.md: icon source row.
• test/run.js: id list, counts (85 extra cards, 120/119 defaults) and a defaults check.
• docs/providers-and-models.md: id table, counts and a short setup section.
• README.md: card count 119/118.

Tests

• node --check on both catalogs and test/run.js: OK.
• The Chrome and Firefox catalogs are byte-identical (cmp).
• Both SVGs parse as XML and carry xmlns and viewBox, with no href, <image> or <script>.
• git diff --check: clean.
• node test/provider-catalog-integrations.mjs: passed.
• node test/provider-model-limits.mjs: passed.
• WEBBRAIN_TEST_FILTER=provider node test/run.js: 119 passed, 1 failed. The failure is "the Providers page LLM tab has a search box beside provider filters" (providerDisplayLabel is not defined). It fails the same way on the base commit (22721a4) without this change, so it is not caused by this PR.
• Run with Node 24.16.0 after npm ci --ignore-scripts.

Live test

Not run through the extension. I replayed WebBrain's requests against https://api.cheapaimodel.ai/v1/chat/completions with a test key.

• GET /v1/models lists exactly the five ids above plus kimi-k3; plain chat returns 200 on all of them.
• Vision: for a non-DeepSeek-direct endpoint, visionGenerationOptions sends reasoning_effort: "none", reasoning_tokens: 0, chat_template_kwargs: {enable_thinking: false}, temperature: 0 and max_tokens: 800 with an image_url part. deepseek-v4-1-flash, grok-4.7 and kimi-k3 described a test image correctly. glm-5.3-flash returned HTTP 400 "This model always engages in thinking and cannot be disabled" (the same request shape also failed this way in a separate gortex test); the message names none of the fields unsupportedVisionGenerationControl looks for, so WebBrain does not retry without the controls. Without those fields the same image works on glm-5.3-flash. glm-5.3 rejects image parts (text only).
• Tool calls: two-round tools / tool exchanges returned 200 on all five models.

Disclosure: I work on cheapaimodel. This PR was prepared with AI assistance (an AI coding agent wrote the change).

🤖 Generated with Claude Code

[vercel]

@nicho810 is attempting to deploy a commit to the esokullu's projects Team on Vercel.

A member of the Team first needs to authorize it.

[esokullu]

Hi @nicho810,

Thank you for taking the time to put this pull request together and for following WebBrain's catalog and documentation conventions!

At this time, we've decided not to add cheapaimodel to the built-in provider catalog. For built-in cards, we focus on direct AI lab endpoints and established, verified compute infrastructure providers. We avoid adding third-party proxy relays and token resellers to ensure long-term endpoint reliability, consistent model parameter handling, and data privacy for our users.

That said, any WebBrain user who wants to use cheapaimodel can easily connect to it via WebBrain's generic OpenAI-Compatible provider card simply by setting the Base URL to https://api.cheapaimodel.ai/v1 and providing their API key.

Thank you again for your contribution and support of WebBrain!