OpenWebUI-OpenAI-Agent

Overview

OpenWebUI-OpenAI-Agent is a multi-agent pipeline implementation based on the OpenAI Agent SDK. It enables flexible multi-agent conversations, tool-calling, and advanced reasoning for OpenWebUI. The project supports multiple models, agent collaboration, and both standard OpenAI SDK and MCP (Multi-Component Pipeline) backends.

Features

• Multi-agent collaboration: Built-in General Agent, Reasoning Agent, Research Agent, and Triage Agent for automatic task routing.
• Tool calling: Register and execute custom tools (FunctionTool) asynchronously. Includes built-in support for image generation.
• Streaming output: Real-time event streaming for responsive user experience.
• Extensible: Easily integrate custom models, tools, and pipeline logic.
• MCP support: Optional integration with MCP servers for advanced tool execution and environment isolation.
• Observability: Optional tracing support for debugging and monitoring agent behavior.

Main Files

• openai_agent_sdk_pipe.py: Main pipeline using the OpenAI Agent SDK, including agent definitions, tool registration, and event stream handling.
• openai_agent_sdk_mcp_pipe.py: An advanced pipeline with MCP (Multi-Component Pipeline) server support for robust tool execution and environment management.

Agents

• General Agent: Handles general questions and basic coding tasks.
• Reasoning Agent: Handles complex reasoning and advanced coding tasks.
• Research Agent: Performs in-depth research and answers questions requiring current information.
• Triage Agent: Automatically routes user requests to the appropriate agent.

Usage

1. Configure API Keys and Models
   Set your OpenAI API key and model preferences in the pipeline valves or environment variables.

2. Register Tools
   Define and register custom tools (functions) to extend agent capabilities.

3. Run the Pipeline
   Integrate the pipeline with OpenWebUI or call it directly in your Python code.

4. (Optional) MCP Configuration
   For advanced tool execution, configure MCP servers in openai_agent_sdk_mcp_pipe.py and provide a valid MCP config file.

Configuration

• Edit the Valves class in each pipeline file to set API keys, model IDs, proxy settings, MCP options, and other parameters.
• For MCP, provide a JSON config file with your MCP server definitions.

License

MIT License.

---

This project is built with the OpenAI Agent SDK.
For SDK usage and advanced features, please refer to the official documentation.