feat: add OpenRouter + direct OpenAI as alternative LLM providers#442
Open
bettercallzaal wants to merge 3 commits intorecoupable:testfrom
Open
feat: add OpenRouter + direct OpenAI as alternative LLM providers#442bettercallzaal wants to merge 3 commits intorecoupable:testfrom
bettercallzaal wants to merge 3 commits intorecoupable:testfrom
Conversation
Adds a `createModel()` helper that resolves model strings through whichever provider is configured, with priority: 1. Vercel AI Gateway (existing production behavior) 2. OpenRouter (one key, many models) 3. Direct OpenAI (strip provider prefix) This lets contributors run the app locally without needing a Vercel AI Gateway API key — just set OPENROUTER_API_KEY or OPENAI_API_KEY instead. All AI SDK call sites updated to use createModel(). No changes to direct OpenAI API calls (video/transcription/image generation) which remain provider-specific. Also fixes bare model string "gemini-2.5-pro" to use the correct "google/gemini-2.5-pro" provider-prefixed format. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Contributor
|
@bettercallzaal is attempting to deploy a commit to the Recoupable Team on Vercel. A member of the Team first needs to authorize it. |
|
Warning Rate limit exceeded
Your organization is not enrolled in usage-based pricing. Contact your admin to enable usage-based pricing to continue reviews beyond the rate limit, or try again in 23 minutes and 33 seconds. ⌛ How to resolve this issue?After the wait time has elapsed, a review can be triggered using the We recommend that you space out your commits to avoid hitting the rate limit. 🚦 How do rate limits work?CodeRabbit enforces hourly rate limits for each developer per organization. Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout. Please see our FAQ for further information. ℹ️ Review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Run ID: ⛔ Files ignored due to path filters (4)
📒 Files selected for processing (10)
✨ Finishing Touches🧪 Generate unit tests (beta)
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. Comment |
5 tasks
![cubic-dev-ai[bot]](https://avatars.githubusercontent.com/in/1082092?s=60&v=4){kind=small}
There was a problem hiding this comment.
4 issues found across 15 files
Confidence score: 3/5
- There is concrete regression risk in fallback behavior:
createModel("google/gemini-2.5-pro")can be passed directly to OpenAI inlib/chat/toolChains/toolChains.ts, which is likely to fail at runtime with an unknown-model error when onlyOPENAI_API_KEYis configured. - A similarly high-impact path exists in
lib/agents/ImageGenerationAgent/createImageGenerationAgent.ts, where the provider-specificgoogle/gemini-3-pro-imagecan be routed through direct OpenAI and fail for custom image-generation flows. lib/ai/createModel.tscurrently allows non-OpenAI model IDs to reach the OpenAI SDK, so errors are confusing and harder to diagnose instead of failing early with a clear message; this keeps risk in the medium range rather than merge-blocking.- Pay close attention to
lib/chat/toolChains/toolChains.ts,lib/agents/ImageGenerationAgent/createImageGenerationAgent.ts,lib/ai/createModel.ts,lib/ai/getAvailableModels.ts- provider routing and fallback error handling may mask or trigger runtime failures.
Prompt for AI agents (unresolved issues)
Check if these issues are valid — if so, understand the root cause of each and fix them. If appropriate, use sub-agents to investigate and fix each issue separately.
<file name="lib/chat/toolChains/toolChains.ts">
<violation number="1" location="lib/chat/toolChains/toolChains.ts:22">
P1: When only `OPENAI_API_KEY` is set, `createModel("google/gemini-2.5-pro")` passes `"google/gemini-2.5-pro"` verbatim to the OpenAI SDK, which will reject it as an unknown model. The direct-OpenAI fallback in `createModel` only strips the `openai/` prefix and has no mapping or error for non-OpenAI model IDs.</violation>
</file>
<file name="lib/ai/getAvailableModels.ts">
<violation number="1" location="lib/ai/getAvailableModels.ts:36">
P2: When the gateway key is set but the API call fails, returning `DEFAULT_MODELS` silently masks gateway errors. Consider logging the error and/or returning `[]` to preserve the previous failure-is-visible semantics, so operators notice gateway misconfiguration.</violation>
</file>
<file name="lib/ai/createModel.ts">
<violation number="1" location="lib/ai/createModel.ts:26">
P2: Non-OpenAI model IDs (e.g. `"google/gemini-2.5-pro"`) silently pass through to the OpenAI SDK, producing a confusing API error. Consider throwing an explicit error when the direct-OpenAI fallback receives a model that isn't prefixed with `"openai/"`.</violation>
</file>
<file name="lib/agents/ImageGenerationAgent/createImageGenerationAgent.ts">
<violation number="1" location="lib/agents/ImageGenerationAgent/createImageGenerationAgent.ts:11">
P1: Custom agent: **Flag AI Slop and Fabricated Changes**
This Google-specific image generation model (`google/gemini-3-pro-image`) is provider-dependent and will fail when routed through direct OpenAI (`openai("google/gemini-3-pro-image")`). The PR description itself acknowledges image generation is provider-specific and "can't route through OpenRouter," yet this file was included in the mechanical `createModel()` sweep. Keep the original string literal here, or guard this call site so it only resolves via the gateway.</violation>
</file>
Architecture diagram
sequenceDiagram
participant Feature as AI Features (Agents, Utilities, Evals)
participant Factory as NEW: createModel() (lib/ai/createModel.ts)
participant SDK as AI SDK / Providers
participant Gateway as Vercel AI Gateway
participant OpenRouter as NEW: OpenRouter API
participant OpenAI as NEW: OpenAI API
Note over Feature, OpenAI: LLM Request Flow (Priority Cascade)
Feature->>Factory: createModel(modelId)
alt VERCEL_AI_GATEWAY_API_KEY is set
Factory->>SDK: Use gateway(modelId)
SDK-->>Factory: LanguageModel (Gateway)
else OPENROUTER_API_KEY is set
Factory->>SDK: NEW: createOpenRouter()
Factory->>SDK: Use openrouter(modelId)
SDK-->>Factory: LanguageModel (OpenRouter)
else OPENAI_API_KEY is set
Factory->>Factory: NEW: Strip "openai/" prefix if present
Factory->>SDK: Use openai(bareModel)
SDK-->>Factory: LanguageModel (Direct)
else No keys configured
Factory-->>Feature: Throw Error
end
Factory-->>Feature: Return LanguageModel
Feature->>SDK: generateText() / ToolLoopAgent / generateObject()
alt Routing via Gateway
SDK->>Gateway: POST /v1/chat/completions
Gateway-->>SDK: Response
else Routing via OpenRouter
SDK->>OpenRouter: NEW: POST /api/v1/chat/completions
OpenRouter-->>SDK: Response
else Routing via Direct OpenAI
SDK->>OpenAI: NEW: POST /v1/chat/completions
OpenAI-->>SDK: Response
end
SDK-->>Feature: Final AI Output (Text/Object/Tool Call)
Note over Feature, Gateway: Model Discovery Flow
Feature->>SDK: CHANGED: getAvailableModels()
alt VERCEL_AI_GATEWAY_API_KEY is set
SDK->>Gateway: Fetch available models
Gateway-->>Feature: Model list
else Gateway not configured
SDK-->>Feature: NEW: Return hardcoded DEFAULT_MODELS list
end
Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review, or fix all with cubic.
| web_deep_research: "openai/gpt-5.4-mini", | ||
| create_knowledge_base: "openai/gpt-5.4-mini", | ||
| send_email: "openai/gpt-5.4-mini", | ||
| update_account_info: createModel("google/gemini-2.5-pro"), |
There was a problem hiding this comment.
P1: When only OPENAI_API_KEY is set, createModel("google/gemini-2.5-pro") passes "google/gemini-2.5-pro" verbatim to the OpenAI SDK, which will reject it as an unknown model. The direct-OpenAI fallback in createModel only strips the openai/ prefix and has no mapping or error for non-OpenAI model IDs.
Prompt for AI agents
Check if this issue is valid — if so, understand the root cause and fix it. At lib/chat/toolChains/toolChains.ts, line 22:
<comment>When only `OPENAI_API_KEY` is set, `createModel("google/gemini-2.5-pro")` passes `"google/gemini-2.5-pro"` verbatim to the OpenAI SDK, which will reject it as an unknown model. The direct-OpenAI fallback in `createModel` only strips the `openai/` prefix and has no mapping or error for non-OpenAI model IDs.</comment>
<file context>
@@ -18,22 +19,22 @@ export type PrepareStepResult = {
- web_deep_research: "openai/gpt-5.4-mini",
- create_knowledge_base: "openai/gpt-5.4-mini",
- send_email: "openai/gpt-5.4-mini",
+ update_account_info: createModel("google/gemini-2.5-pro"),
+ get_spotify_search: createModel("openai/gpt-5.4-mini"),
+ update_artist_socials: createModel("openai/gpt-5.4-mini"),
</file context>
- Revert createImageGenerationAgent to use bare model string since google/gemini-3-pro-image is provider-specific and can't route through OpenRouter or direct OpenAI - Add explicit error when direct OpenAI fallback receives a non-OpenAI model (e.g. "google/..." or "anthropic/...") instead of silently passing it to the OpenAI API Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add VERCEL_OIDC_TOKEN check to gateway detection in createModel - Return [] on gateway failure instead of DEFAULT_MODELS to avoid masking gateway errors - Add pricing metadata to fallback model entries Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
sweetmantech
requested changes
Apr 17, 2026
Comment on lines
+19
to
+22
| if (process.env.OPENROUTER_API_KEY) { | ||
| const openrouter = createOpenRouter({ apiKey: process.env.OPENROUTER_API_KEY }); | ||
| return openrouter(modelId); | ||
| } |
Contributor
There was a problem hiding this comment.
KISS principle
- Why add OpenRouter when we already support AI Gateway?
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
createModel()helper (lib/ai/createModel.ts) that resolves model strings through whichever provider is configuredcreateModel()(15 files)@openrouter/ai-sdk-providerdependencygetAvailableModels()for non-gateway environments"gemini-2.5-pro"→"google/gemini-2.5-pro"inTOOL_MODEL_MAPMotivation
Contributors without Vercel AI Gateway access can't run the chat/agent features locally. This lets devs run with just an
OPENROUTER_API_KEY(one key, many models) or a directOPENAI_API_KEY.No breaking changes — existing production deployments using the gateway are unaffected.
What's NOT changed
"ai"packageTest plan
VERCEL_AI_GATEWAY_API_KEY→ verify existing behavior unchangedOPENROUTER_API_KEY→ verify chat/agents work through OpenRouterOPENAI_API_KEY→ verify chat/agents work through OpenAI directlypnpm buildpassespnpm testpasses🤖 Generated with Claude Code
Summary by cubic
Add OpenRouter and direct OpenAI as alternative LLM providers via a new
createModel()helper, so local dev works without the Vercel AI Gateway. Existing gateway behavior and media generation APIs remain unchanged.New Features
createModel(modelId)routes by priority: Vercel AI Gateway → OpenRouter → OpenAI; updated AI SDK call sites to use it.getAvailableModels()returns a small default list when no gateway is configured, with pricing metadata.@openrouter/ai-sdk-provider.Bug Fixes
openai/*models.VERCEL_OIDC_TOKEN.getAvailableModels()now returns[]instead of the default list.Written for commit 8f6b5b1. Summary will update on new commits.