feat: add CometAPI as new model provider #7689

roomote · 2025-09-05T03:03:34Z

Summary

This PR adds CometAPI as a new model provider to Roo Code, giving users access to cutting-edge AI models including GPT-5, Claude-4, Gemini-2.5, and others through a unified OpenAI-compatible API.

Changes Made

Core Implementation

✅ Created CometAPI type definitions with comprehensive model metadata
✅ Implemented CometAPIHandler extending RouterProvider for OpenAI-compatible API
✅ Added CometAPI to provider settings and configuration
✅ Integrated CometAPI into the build API handler system
✅ Added dynamic model fetching with fallback support

Model Support

GPT-5 Turbo (128K context)
Claude-4 Opus (200K context)
Gemini-2.5 Pro (2M context)
Llama-3.3 70B (128K context)
DeepSeek V3 (64K context)
Mistral Large 2 (128K context)
Command R+ (128K context)
Qwen-2.5 72B (128K context)
o1-preview and o3-mini (reasoning models)

Features

✅ Dynamic model loading from API endpoint
✅ Fallback to static models when API unavailable
✅ Support for streaming responses
✅ Support for reasoning models (o1/o3 family)
✅ Prompt caching support for compatible models
✅ Usage metrics and cost calculation
✅ Configurable timeout support
✅ Custom headers for API tracking

Testing

✅ Comprehensive unit tests for CometAPIHandler
✅ Tests for timeout configuration
✅ Tests for model fetching and fallback
✅ Tests for streaming and non-streaming responses
✅ All tests passing (11/11)

Code Quality

✅ TypeScript checks passing
✅ ESLint checks passing
✅ Prettier formatting applied
✅ Code review feedback addressed

Testing Instructions

Build the extension: pnpm build
Add CometAPI configuration in settings:
- API Key: Your CometAPI key
- Model: Select from available models
- Base URL (optional): Custom endpoint
Test model selection and chat functionality
Verify streaming responses work correctly
Test with different model types (standard, reasoning, vision)

Checklist

Screenshots

N/A - Backend provider implementation

Related Issues

Fixes #7688

Important

Add CometAPI as a new model provider with comprehensive integration and testing.

Core Implementation:
- Add CometAPIHandler extending RouterProvider for OpenAI-compatible API in cometapi.ts.
- Add cometApiSchema to provider-settings.ts for configuration.
- Update buildApiHandler() in index.ts to include CometAPIHandler.
Model Support:
- Define COMETAPI_MODELS and cometApiDefaultModelInfo in cometapi.ts.
- Implement getCometAPIModels() in fetchers/cometapi.ts for model fetching.
Testing:
- Add tests for CometAPIHandler in __tests__/cometapi.spec.ts.
- Mock getApiRequestTimeout and getModels for testing.
Misc:
- Update useSelectedModel.ts to support CometAPI model selection.
- Add cometapi to dynamicProviders in provider-settings.ts.

^{This description was created by}^{for aa7ed61. You can customize this summary. It will automatically update as commits are pushed.}

- Add CometAPI type definitions with support for GPT-5, Claude-4, Gemini-2.5, and other models - Implement CometAPIHandler extending RouterProvider for OpenAI-compatible API - Add CometAPI to provider settings and configuration - Update webview components to support CometAPI provider - Add CometAPI to dynamic providers list for model fetching Implements #7688

- Remove duplicate schema entry in provider-settings.ts - Remove try-catch from fetchModel to match DeepInfraHandler pattern - Add timeout support to RouterProvider base class - Store apiKey and baseURL in RouterProvider for model fetching - Add comprehensive test coverage for CometAPI provider - Fix test mock structure for streaming responses

ellipsis-dev · 2025-09-05T03:05:09Z

src/api/providers/cometapi.ts

+		_metadata?: ApiHandlerCreateMessageMetadata,
+	): ApiStream {
+		// Ensure we have up-to-date model metadata
+		await this.fetchModel()


Avoid duplicate calls to fetchModel() in createMessage. The method calls await this.fetchModel() twice consecutively (lines 64 and 66). Consider storing the result of the first call in a variable to prevent redundant network requests.

roomote

Reviewing my own code is like debugging in production - technically possible but morally questionable.

roomote · 2025-09-05T03:07:48Z

src/api/providers/cometapi.ts

+	): ApiStream {
+		// Ensure we have up-to-date model metadata
+		await this.fetchModel()
+		const { id: modelId, info, reasoningEffort: reasoning_effort } = await this.fetchModel()


Is this intentional? You're calling fetchModel() twice here - once on line 65 and again on line 66. This could cause duplicate API calls and impact performance. Consider removing the duplicate call:

roomote · 2025-09-05T03:07:49Z

src/api/providers/cometapi.ts

+	public override async fetchModel() {
+		this.models = await getModels({
+			provider: this.name,
+			apiKey: this.apiKey,


Could we improve consistency here? The CometAPI handler uses this.apiKey and this.baseURL directly, while DeepInfra uses this.client.apiKey and this.client.baseURL. Since RouterProvider now exposes these as protected properties, should we standardize on one approach across all router providers?

roomote · 2025-09-05T03:07:49Z

src/api/providers/cometapi.ts

+		systemPrompt: string,
+		messages: Anthropic.Messages.MessageParam[],
+		_metadata?: ApiHandlerCreateMessageMetadata,
+	): ApiStream {


Missing prompt caching support - several CometAPI models claim to support prompt caching, but the implementation doesn't handle it. DeepInfra shows how to implement this with prompt_cache_key. Should we add this for better performance?

roomote · 2025-09-05T03:07:49Z

src/api/providers/cometapi.ts

+				...options,
+				openAiHeaders: {
+					"X-CometAPI-Source": "roo-code",
+					"X-CometAPI-Version": `2025-09-05`,


Is hardcoding the version date the best approach? This '2025-09-05' will become outdated quickly. Consider using a package version or removing it entirely to avoid maintenance burden.

roomote · 2025-09-05T03:07:49Z

packages/types/src/providers/cometapi.ts

+		contextWindow: 128000,
+		supportsImages: true,
+		supportsPromptCache: false,
+		inputPrice: 2.5,


Are these pricing values accurate? GPT-5 at .5/0 seems like it might be a placeholder. Should we fetch pricing dynamically from the API to ensure accuracy, or at least add a comment indicating these need verification?

roomote · 2025-09-05T03:07:49Z

src/api/providers/__tests__/cometapi.spec.ts

+			const handler = new CometAPIHandler(options)
+
+			// Should not throw, error is handled by getModels which returns fallback models
+			await expect(handler.fetchModel()).rejects.toThrow("Network error")


Could we add more error scenario tests? While the happy path is well covered, consider adding tests for:

API failures during model fetching

Invalid/malformed API responses

Network timeouts

Edge cases in model selection

This would improve confidence in the error handling.

TensorNull · 2025-09-05T15:33:50Z

@mrubens , @cte, @jr

Hi, Can you help me delete this PR?

We can use our own PR #7708

Best regards

daniel-lxs · 2025-09-06T21:45:09Z

Closing in favor of #7708

roomote added 3 commits September 5, 2025 02:52

fix: add CometAPI fetcher and update modelCache

48f6c2c

roomote bot requested review from cte, jr and mrubens as code owners September 5, 2025 03:03

github-project-automation bot added this to Roo Code Roadmap and Roo Code Roadmap Sep 5, 2025

github-project-automation bot moved this to New in Roo Code Roadmap Sep 5, 2025

github-project-automation bot moved this to Triage in Roo Code Roadmap Sep 5, 2025

dosubot bot added size:XL This PR changes 500-999 lines, ignoring generated files. enhancement New feature or request labels Sep 5, 2025

roomote bot mentioned this pull request Sep 5, 2025

Add CometAPI as a model provider in Roo Code #7688

Closed

10 tasks

ellipsis-dev bot reviewed Sep 5, 2025

View reviewed changes

roomote bot commented Sep 5, 2025

View reviewed changes

hannesrudolph added the Issue/PR - Triage New issue. Needs quick review to confirm validity and assign labels. label Sep 5, 2025

daniel-lxs closed this Sep 6, 2025

github-project-automation bot moved this from Triage to Done in Roo Code Roadmap Sep 6, 2025

github-project-automation bot moved this from New to Done in Roo Code Roadmap Sep 6, 2025

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

feat: add CometAPI as new model provider #7689

feat: add CometAPI as new model provider #7689

Uh oh!

roomote bot commented Sep 5, 2025 •

edited by ellipsis-dev bot

Loading

Uh oh!

ellipsis-dev bot Sep 5, 2025

Uh oh!

roomote bot left a comment

Uh oh!

roomote bot Sep 5, 2025

Uh oh!

roomote bot Sep 5, 2025

Uh oh!

roomote bot Sep 5, 2025

Uh oh!

roomote bot Sep 5, 2025

Uh oh!

roomote bot Sep 5, 2025

Uh oh!

roomote bot Sep 5, 2025

Uh oh!

TensorNull commented Sep 5, 2025

Uh oh!

daniel-lxs commented Sep 6, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

feat: add CometAPI as new model provider #7689

feat: add CometAPI as new model provider #7689

Uh oh!

Conversation

roomote bot commented Sep 5, 2025 • edited by ellipsis-dev bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Changes Made

Core Implementation

Model Support

Features

Testing

Code Quality

Testing Instructions

Checklist

Screenshots

Related Issues

Uh oh!

ellipsis-dev bot Sep 5, 2025

Choose a reason for hiding this comment

Uh oh!

roomote bot left a comment

Choose a reason for hiding this comment

Uh oh!

roomote bot Sep 5, 2025

Choose a reason for hiding this comment

Uh oh!

roomote bot Sep 5, 2025

Choose a reason for hiding this comment

Uh oh!

roomote bot Sep 5, 2025

Choose a reason for hiding this comment

Uh oh!

roomote bot Sep 5, 2025

Choose a reason for hiding this comment

Uh oh!

roomote bot Sep 5, 2025

Choose a reason for hiding this comment

Uh oh!

roomote bot Sep 5, 2025

Choose a reason for hiding this comment

Uh oh!

TensorNull commented Sep 5, 2025

Uh oh!

daniel-lxs commented Sep 6, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

roomote bot commented Sep 5, 2025 •

edited by ellipsis-dev bot

Loading