Open WebUI Chat (LLM) icon

Open WebUI Chat (LLM)

Send a message (and optional file or knowledge collection) to an Open WebUI model

Open WebUI Chat (LLM)

AI-generated

Summary

Send a user message to a selected Open WebUI model, optionally adding prompts, tools, built-in features, knowledge collections, or a binary file. Choose the API style, response format, and streaming mode; the node returns the assistant result or stream chunks for each input item.

Inputs

  • API Style — Select automatic endpoint detection, the OpenAI-compatible /api/v1 paths, or the legacy /api paths. Defaults to Auto.
  • Model Name or ID (required) — Choose the model or provide its ID with an expression.
  • Tool Names or IDs — Optionally select tools to send with the chat request.
  • Inbuilt Features or IDs — Optionally enable model-supported capabilities such as Web Search, Code Interpreter, or Image Generation. With Function Calling set to Auto, selecting features requests server-side execution.
  • Function Calling — Choose Auto, Legacy, Native, or Model Default. Auto requests Legacy when features are selected; otherwise it leaves the model setting in effect. Native model-driven tool calls do not execute built-in features for API callers.
  • System Prompt — Optional system message added before the user message.
  • User Message (required) — Message sent to the model.
  • Assistant Message — Optional assistant message added between the system prompt and user message.
  • Return — Short returns the assistant content with model, usage, and finish reason where available. Full Response returns the API response; in streaming-chunks mode it returns the raw event payloads.
  • Conversation Knowledge Bases or IDs — Optionally attach knowledge collections for retrieval in the conversation.
  • Upload Target Knowledge Base Name or ID — Optionally select a collection to add an uploaded binary file to.
  • If a Binary Is Provided and a Target Upload Collection Is Selected — When a binary property is set and an upload collection is selected, Add File to the Target Upload Collection uploads and adds the file to that collection; Do Not Add File keeps it out of the collection. In either case, the uploaded file is attached for retrieval.
  • Binary Property (Optional) — Name of the incoming binary property to upload and attach to the request. A file is used only when this property is specified and present on the input item.
  • Output Mode — No Streaming makes a non-streaming call; Final Response streams internally and returns one completed result; Streaming returns each parsed SSE event as a separate item. Defaults to Final Response.
  • Timeout (Seconds) (required) — Maximum wait in seconds for the response; defaults to 600.

Output shape

One or more output items per input item, on the Success output.

No Streaming and Final Response return one item containing either the short result (content, model, usage, finish_reason) or, with Full Response selected, the full API response for non-streaming calls or a chat-completion-shaped object for the completed stream. Streaming returns separate items for each event containing text, role, or finish information; Short mode wraps each as a chunk with data.delta, while Full Response passes through the event payload. If Continue On Fail is enabled, errors are returned as items with error details.

Links

Discussion