Implemented
0
Which credential type fits an endpoint that charges per call?
Source: n8n-io/n8n#39546 · opened by @TheCheetah11
The AI nodes in n8n's LangChain package each sit beside a credential that holds a key, and OpenRouter shows the whole shape in two files. The node itself is built on the OpenAI client, and packages/@n8n/nodes-langchain/credentials/OpenRouterApi.credentials.ts is what stands behind it: an API Key marked required and stored as a password, a base URL that stays hidden, and an authenticate block that turns that key into a bearer header before a request leaves. The test a user runs when saving the credential then proves it against the vendor. Pointing that plumbing at an OpenAI-compatible endpoint is where it gets rough, and your tracker is carrying the reports. There is an open thread about OpenAI credentials routed through compatible gateways coming apart, with the model prefix stripped and header fields dropped (#39227), another about the Gemini compatible endpoint in the assistant, and a third where a five word question ran to about 137k input tokens and $0.42 (#38130). BlockRun …
No pledges yet. Be the first to back this.
Comments
Similar requests
Zendesk OAuth2 credential cannot connect to OAuth clients marked as Public (PKCE required)
0 votes · 0 comments
HTTP Request does not refresh credentials when the API reports expiry in the response body
0 votes · 0 comments
When custom Ollama credentials are configured and the endpoint is inaccessible, the '/home/chat' page keeps displaying a loading status when selecting a model, resulting in a timeout for the API endpoint /rest/chat/models.
0 votes · 0 comments
AI Agent: return tool-schema validation failures to the model as recoverable errors instead of terminating the execution
0 votes · 0 comments
Feature: Add configurable request-body validation to the Webhook node
0 votes · 0 comments
No comments yet.