Multi-Agent Orchestrator for LLM Workflows

Live demo

https://vtayyab6-multi-agent-orchestrator.hf.space/?__theme=light

A Gradio application to configure a coordinator, sub-agents, and safe tools, then run a visible multi-agent workflow on Gemini with live progress, plan, execution trace, and final synthesized answer.

Features

Feature	Details
Workflow design in UI	Configure Task, Agent, Sub-Agents, Tools, and API Config from tabs
Multi-agent orchestration	1 coordinator + up to 3 sub-agents with delegated subtasks
Safe tools	Wikipedia Search, Keyword Extractor, Text Summarizer, Word Counter, Calculator
Live execution visibility	Live status, plan, execution trace, agent I/O, sources, architecture
Model flexibility	Select Gemini model from dropdown (default: gemini-3.1-flash-lite-preview)
Final answer utility	Dedicated final answer tab with copy action
Secure key handling	API key entered in UI, used only for run-time requests, not stored

Screenshots (EU AI Act Example)

1) Before running

2) Sample task input

3) Agent configuration

4) Sub-agent configuration

5) Tools configuration

6) Results - plan

7) Results - execution trace

8) Results - agent input/output

9) Results - architecture

10) Results - final answer

Project Structure

multi-agent-orchestrator/
├── app.py
├── requirements.txt
├── README.md
├── screenshots/
│   ├── 01-before-running.png
│   ├── 02-sample-task-run.png
│   ├── 03-agent-tab.png
│   ├── 04-subagents-tab.png
│   ├── 05-tools-tab.png
│   ├── 06-results-plan-tab.png
│   ├── 07-results-execution-trace-tab.png
│   ├── 08-results-agent-io-tab.png
│   ├── 09-results-architecture-tab.png
│   └── 10-results-final-answer-tab.png
└── src/
    ├── __init__.py
    ├── config.py
    ├── agent_builder.py
    ├── gemini_client.py
    ├── orchestrator.py
    ├── tool_builder.py
    └── tools.py

Local Setup

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
python app.py