Configure the HTTP Request node
Create a TextCortex key in account settings. Use an n8n header-auth credential so the key is not stored as ordinary workflow data.
- Method: POST.
- URL: https://api.textcortex.com/v1/chat/completions
- Authorization header: Bearer followed by your API key.
- Body content type: JSON.
Send a minimal chat request
Replace MODEL_ID_FROM_YOUR_ACCOUNT with an identifier returned by GET /models. Keep stream set to false for a simple request-response workflow.
{
"model": "MODEL_ID_FROM_YOUR_ACCOUNT",
"messages": [
{
"role": "user",
"content": "Hello"
}
],
"stream": false
}Validate the response before continuing
Read the response content from the returned choices. If a later workflow step expects structured data, parse and validate that data before sending it to another system.
Define an error branch for authentication failures, invalid requests, timeouts and rate limits. Avoid retrying a write action just because the model step was repeated.
Use model choices across workflows
A classification workflow and a long-form analysis workflow can use different model identifiers through the same endpoint. Store approved choices in workflow configuration.
Keep EU-hosted routes selected for workflows that require European inference. Review the location of n8n itself and any connected services separately from the model endpoint.