feat: add ChatOrcaRouter chat model node#6447
Conversation
Add an OrcaRouter chat model node (an OpenAI-compatible LLM router) following the ChatOpenRouter / ChatCometAPI pattern: a credential, the node plus a vision subclass, an icon, and co-located tests. The Model Name dropdown is loaded from the OrcaRouter catalog and is freeSolo. Also register chatOrcaRouter in the server streaming list.
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Code Review
This pull request introduces the ChatOrcaRouter integration, which connects to OrcaRouter, an OpenAI-compatible LLM router. It includes credential configuration, the chat model node implementation, tests, and registration for streaming. The feedback suggests replacing simple truthy checks for numeric parameters (like temperature and penalties) with explicit nullish and empty-string checks to prevent valid values of 0 from being ignored.
| if (temperature) obj.temperature = parseFloat(temperature) | ||
| if (maxTokens) obj.maxTokens = parseInt(maxTokens, 10) | ||
| if (topP) obj.topP = parseFloat(topP) | ||
| if (frequencyPenalty) obj.frequencyPenalty = parseFloat(frequencyPenalty) | ||
| if (presencePenalty) obj.presencePenalty = parseFloat(presencePenalty) | ||
| if (timeout) obj.timeout = parseInt(timeout, 10) |
There was a problem hiding this comment.
Using simple truthy checks (if (temperature)) can lead to bugs when valid values like 0 (for temperature, penalties, or topP) are passed. If these values are resolved as numbers (e.g., from variables or overrides), 0 evaluates to false and the setting is incorrectly ignored. It is safer to explicitly check against null or undefined using loose equality (!= null), and empty strings.
| if (temperature) obj.temperature = parseFloat(temperature) | |
| if (maxTokens) obj.maxTokens = parseInt(maxTokens, 10) | |
| if (topP) obj.topP = parseFloat(topP) | |
| if (frequencyPenalty) obj.frequencyPenalty = parseFloat(frequencyPenalty) | |
| if (presencePenalty) obj.presencePenalty = parseFloat(presencePenalty) | |
| if (timeout) obj.timeout = parseInt(timeout, 10) | |
| if (temperature != null && temperature !== '') obj.temperature = parseFloat(temperature) | |
| if (maxTokens != null && maxTokens !== '') obj.maxTokens = parseInt(maxTokens, 10) | |
| if (topP != null && topP !== '') obj.topP = parseFloat(topP) | |
| if (frequencyPenalty != null && frequencyPenalty !== '') obj.frequencyPenalty = parseFloat(frequencyPenalty) | |
| if (presencePenalty != null && presencePenalty !== '') obj.presencePenalty = parseFloat(presencePenalty) | |
| if (timeout != null && timeout !== '') obj.timeout = parseInt(timeout, 10) |
References
- In JavaScript/TypeScript, use loose equality (
== null) as a standard idiom for a 'nullish' check that covers bothnullandundefined.
Use nullish + empty-string checks instead of truthy checks so a valid 0 for temperature, topP, or the penalties is not dropped. Addresses review feedback.
1 participant
Summary
Adds a new
ChatOrcaRouterchat model node, following the same pattern used forChatOpenRouter(#3865) andChatCometAPI(#5160). OrcaRouter is an OpenAI-compatible LLM router that exposes 150+ models from OpenAI, Anthropic, Google, DeepSeek, xAI, MiniMax, Qwen and others behind a single endpoint, plus anorcarouter/autorouter that picks the best upstream per request.Disclosure: I'm an engineer on the OrcaRouter team.
What changed
packages/components/credentials/OrcaRouterApi.credential.ts, a singlepasswordfield for the OrcaRouter API key (keys begin withsk-orca-).packages/components/nodes/chatmodels/ChatOrcaRouter/, containingChatOrcaRouter.ts(INode implementation),FlowiseChatOrcaRouter.ts(LangChainChatOpenAIsubclass implementingIVisionChatModal),orcaRouter.svg(icon), andChatOrcaRouter.test.ts(co-located jest tests covering the model list, credential wiring, default and custom base URL, attribution headers, blank-temperature behaviour, and Base Options JSON parsing).chatOrcaRoutertostreamAvailableLLMs['Chat Models']inpackages/server/src/utils/index.ts, matching the change made forchatCometAPI.The node wraps
@langchain/openai'sChatOpenAIwithconfiguration.baseURL = 'https://api.orcarouter.ai/v1', exposes the usual chat-model parameters (temperature, max tokens, top P, frequency/presence penalty, timeout, streaming, image uploads), and injectsHTTP-Referer: https://www.orcarouter.ai/andX-Title: Flowiseattribution headers by default (overridable through the existing Base Options JSON field, same shape as the OpenRouter node).The Model Name field is an
asyncOptionsdropdown withfreeSolo: true. ItslistModelsloader fetches OrcaRouter's public model catalog and filters it to chat-completions models, led byorcarouter/auto, so new models appear without a code change. If the fetch fails it falls back to a small set of flagship presets, and users can always type any free-form model id.Companion docs PR
FlowiseAI/FlowiseDocs: FlowiseAI/FlowiseDocs#216
Testing
pnpm --filter flowise-components test: new co-located jest suite passes (ChatOrcaRouter.test.ts).pnpm lintandpnpm buildclean.sk-orca-key: non-stream chat, stream chat, tool calling, 401 on a bad key, and 4xx on an invalid model all behave as expected.pnpm start): dragged the new ChatOrcaRouter node onto a chatflow, wired it to an LLM Chain, sent messages with bothorcarouter/autoand pinned models, confirmed responses and attribution headers in the OrcaRouter console.