Open WebUI Utils

A collection of utilities for Open WebUI

Native Tool Calling Pipe

A powerful pipe implementation that enables seamless OpenAI API-native tool calling with streaming and multi-call support for Open WebUI.

Features

• 🔄 Native OpenAI API tool calling format
• ⚡ Streaming support for real-time responses
• 🛠️ Multi-call support for back-and-forth between tools and the assistant
• 🔌 Seamless integration with Open WebUI
• 🎯 Compatible with any OpenAI-compatible endpoint (OpenAI, Ollama, Openrouter,...)

Demo

CleanShot.2025-01-13.at.08.17.19.mp4

Requirements

• Open WebUI version 0.5.0 or higher
• OpenAI API key

Installation

• Open the function on the community page
• Click Get and add it to your Open WebUI instance
• Configure the Valves
• In a new conversation, choose one of the new models prefixed with native-tool/

Usage

When you choose one of the models prefixed with native-tool/, interaction with the LLM will happen through this pipe and the built-in function calling will be taken over.

Configuration

The pipe accepts the following configuration options:

• OPENAI_API_KEY: Your OpenAI API key
• OPENAI_BASE_URL: OpenAI API base URL (default: "https://api.openai.com/v1")
• MODEL_IDS: List of model IDs to enable (default: ["gpt-4o-mini"])