feat(providers): add Atomic Chat OpenAI-compatible provider

[yanalialiuk]

Summary

Adds Atomic Chat as a local OpenAI-compatible provider for Roo Code. Atomic Chat exposes a local OpenAI-style HTTP API (default: http://127.0.0.1:1337) with support for GET /v1/models and POST /v1/chat/completions, allowing Roo Code to use locally hosted models through the same workflow as other OpenAI-compatible providers.

Project: https://github.com/AtomicBot-ai/Atomic-Chat

What changed

• Added new provider id: atomic-chat
• Added provider settings:
  • atomicChatModelId
  • atomicChatBaseUrl
  • atomicChatApiKey
• Updated:
  • localProviders
  • provider catalog
  • model-id maps
  • secret key handling
  • extension message types for model refresh
• Added AtomicChatHandler
• Uses OpenAI SDK with configurable baseURL
• Uses automatic /v1 routing
• Reuses shared Anthropic → OpenAI message conversion
• Uses the same streaming flow as existing OpenAI-compatible local providers
• Added model discovery support:
  • getAtomicChatModels
  • modelCache integration branch
• Added webview request/response flow:
  • requestAtomicChatModels
  • atomicChatModels
• Uses the same refresh/discovery pattern as LM Studio and related local providers
• Added webview support:
  • provider settings UI
  • validation
  • model picker support
  • localization strings (EN)

How to test

1. Start Atomic Chat (or another compatible OpenAI-style local server) using a base URL such as:

http://127.0.0.1:1337

2. Ensure the server exposes:

• GET /v1/models
• POST /v1/chat/completions
• Expected model response format:

{
  "data": [
    {
      "id": "model-name"
    }
  ]
}

3.Open Roo Code settings.
4.Select provider: Atomic Chat
5.Configure:
-base URL (optional if using default)
-API key (optional)
-model
6.Verify:
-models populate correctly
-chat completion succeeds
-streaming responses function normally

Notes

• Default base URL matches a common local setup: http://127.0.0.1:1337
• Users can override the endpoint as needed
  -Model metadata currently follows the same defaults used by other OpenAI-compatible providers until richer metadata -----becomes available from the upstream API
  -The implementation intentionally follows existing OpenAI-compatible provider patterns to minimize maintenance overhead and behavioral differences

Interactively review PR in Roo Code Cloud