Connect Langflow

This guide uses Langflow’s OpenAI Compatible provider. Use an existing installation with access to its model provider settings. Older releases may show a custom Base URL on an OpenAI component instead; follow the documentation matching your installed version.

1. Configure the provider

Click your profile icon, open Settings → Model Providers, and select OpenAI Compatible.

Field	Value
Base URL	The service Base URL above, including /v1
API Key	Your service key, entered privately without Bearer

Use the API root; Langflow handles /models/ discovery and the text request path. Do not paste /chat/completions into Base URL. Keep other provider and security settings.

Select Save. Langflow checks the models endpoint and lists the discovered models. Enable the text models you want to use. Compare their exact IDs with the authenticated models endpoint; a discovered row does not establish embedding or image support.

2. Select the model in a flow

Open a draft flow. In a Language Model field, choose OpenAI Compatible, then select an enabled text model. Connect it to the flow’s prompt and chat output as your flow requires, and save.

Use the flow’s Playground to send a short text prompt and confirm the request in your service dashboard. Running the flow spends balance; agent loops and multiple model nodes can make several requests.

Provider settings persist after saving. Reopen the flow’s model picker if the list has not refreshed. If discovery fails, check the /v1 suffix, key and network access from the Langflow server.

Environment settings

For deployments managed through environment variables, the provider documents OPENAI_COMPATIBLE_BASE_URL and OPENAI_COMPATIBLE_API_KEY. Set them privately in the existing deployment environment, then restart that deployment. Model enabling and flow selection still use the UI in this guide. Avoid changing another provider’s generic OpenAI variables.

Tool calling requires a capable model and the flow’s explicit tool configuration. This text setup does not configure embeddings or other media.

See the official OpenAI Compatible provider guide and Language Model component.