[binyangzhu000-sudo]

Summary

Adds Atlas Cloud as a built-in provider — an OpenAI-compatible gateway to 300+ models (OpenAI, Anthropic, DeepSeek, Qwen, GLM, Kimi, MiniMax, ...) behind one endpoint.

Followed docs/adding-new-provider.md and the existing SiliconFlow/Groq shape: a thin OpenAICompatible subclass with a fixed apiHost, registered through defineProvider().

Changes

• src/shared/types/provider.ts — ModelProviderEnum.AtlasCloud
• src/shared/providers/definitions/models/atlascloud.ts — the model class (OpenAICompatible subclass, apiHost: 'https://api.atlascloud.ai/v1')
• src/shared/providers/definitions/atlascloud.ts — the defineProvider() call
• src/shared/providers/builtin-registration.ts — the side-effect import (appended at the end, per the file's own "preserve this import order" comment)
• src/shared/providers/index.contract.test.ts — extended the exact provider-order assertion (this is the registration-order contract test, not new coverage I added on top)
• src/shared/models/index.ts — added the missing aiProviderNameHash entry. This is the legacy backward-compat map docs/adding-new-provider.md's "Migration Notes" say is superseded by defineProvider(), but tsc's exhaustive Record<ModelProviderEnum, string> check still requires every enum member to have one — npx tsc --noEmit catches this immediately if it's missing.
• src/renderer/components/icons/ProviderIcon.tsx — the real Atlas Cloud logo mark, sourced from @lobehub/icons-static-svg (the icon set already backing several entries in this same file), same inline-<path> pattern as the SiliconFlow/Bedrock/xAI entries above it.

Curated model list (9 models) and their context/max-output numbers are taken from the live GET /v1/models catalog (2026-09-20), not guessed. Capability flags are conservative: tool_use is set on all of them (the OpenAI-origin models don't advertise it in supported_features, but they do support it in practice — verified live, see below); reasoning/vision are only set where the catalog's own supported_features/input_modalities actually say so.

No per-provider unit test file added — no sibling OpenAI-compatible provider (Groq, SiliconFlow, DeepSeek, ...) has one; the registry-order contract test is this codebase's real coverage mechanism for a new provider registration, and that's what's extended here.

Testing

• npx biome lint — clean
• npx vitest run src/shared/ src/renderer/utils src/renderer/components/icons — 1186 passed, 6 skipped (no regressions)
• npx tsc --noEmit (full workspace) — 0 errors
• Live, against the real gateway: createOpenAICompatible({ baseURL: 'https://api.atlascloud.ai/v1' }) + streamText() with the exact model-id shape this provider uses (openai/gpt-4.1-mini) — the same @ai-sdk/openai-compatible + ai package primitives AtlasCloud.getProvider()/getChatModel() call internally — returned a real streamed response with real token usage:

  [stream] "chatbox atlascloud ok"
  [usage] { inputTokens: 16, outputTokens: 5, totalTokens: 21, ... }
  

---

Disclosure: I work at Atlas Cloud.

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Advanced

Run ID: 7ad9a166-4f50-403a-ae81-5d551132fd7b

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}