250class Valves(BaseModel):251 """Configuration parameters for the n8n pipe connector."""252 253 model_config = ConfigDict(validate_assignment=True)254 255 n8n_host: str = Field(256 default="http://localhost:5678",257 description="Base URL of the n8n server, e.g. https://n8n.example.com",258 )259 n8n_webhook_id: str = Field(260 default="your-webhook-id",261 description="Path of the n8n Webhook node (the part after /webhook/)", Connect Open-WebUI to n8n: each user message is posted to an n8n Webhook node and the workflow answer is shown as the assistant reply. Build agents, RAG pipelines or automations in n8n and use them from the chat like any other model.
What you get
- Text, images and uploaded documents forwarded to the webhook (JSON, or multipart when there are attachments)
- A stable session id per chat for your agent memory, with separate sessions for Open-WebUI background tasks (title, tags)
- Metadata with every call: user, chat, message, task, model
- Timeouts and retries that behave: the timeout valve applies to every request, retries only on connection errors and HTTP 502/503/504, never after a timeout, so a slow workflow is not started twice
- Live status in the chat while n8n works, clear error messages in English or French
- Optional streaming with the n8n "Streaming response" mode
- Safe by default: remote image download off, attachment size and type limits, bearer token optional
Setup
- In n8n, add a Webhook node (POST, Header Auth recommended) and return {"output": "..."} with a Respond to Webhook node.
- Set the valves: n8n_host, n8n_webhook_id (the webhook path), n8n_bearer_token.
- Select the pipe as a model and start chatting.
Ready-to-import example workflow, full valves reference and troubleshooting: https://github.com/sboily/open-webui-n8n-pipe
Based on coleam/n8n_pipe. Issues and contributions are welcome on GitHub.
Connect Open-WebUI to n8n: each user message is posted to an n8n Webhook node and the workflow answer is shown as the assistant reply. Build agents, RAG pipelines or automations in n8n and use them from the chat like any other model.
What you get
- Text, images and uploaded documents forwarded to the webhook (JSON, or multipart when there are attachments)
- A stable session id per chat for your agent memory, with separate sessions for Open-WebUI background tasks (title, tags)
- Metadata with every call: user, chat, message, task, model
- Timeouts and retries that behave: the timeout valve applies to every request, retries only on connection errors and HTTP 502/503/504, never after a timeout, so a slow workflow is not started twice
- Live status in the chat while n8n works, clear error messages in English or French
- Optional streaming with the n8n "Streaming response" mode
- Safe by default: remote image download off, attachment size and type limits, bearer token optional
Setup
- In n8n, add a Webhook node (POST, Header Auth recommended) and return {"output": "..."} with a Respond to Webhook node.
- Set the valves: n8n_host, n8n_webhook_id (the webhook path), n8n_bearer_token.
- Select the pipe as a model and start chatting.
Ready-to-import example workflow, full valves reference and troubleshooting: https://github.com/sboily/open-webui-n8n-pipe
Based on coleam/n8n_pipe. Issues and contributions are welcome on GitHub.

