Agent & Workflow Builder

Natural-language powered AI Agent & Workflow builder using Microsoft Agent Framework Skills.

Create, edit, visualize, and test-run AI agents & multi-step workflows — entirely from a chat interface.

Note: This is a Proof-of-Concept (PoC) project. Not intended for production deployments without further hardening.

---

Features

Feature	Description
Natural Language Agent Creation	Describe the agent you want in plain text; the system generates YAML definition + executable Python code
Natural Language Workflow Creation	Describe a multi-step process; automatic generation of executors, edges, and conditional branching
Chat-based Editing	Iteratively refine agents and workflows through follow-up chat messages
Visual Workflow Graph	Interactive React Flow visualization with animated edge traversal
Playground	Live execution of agents (LLM streaming) and workflows (step-by-step SSE)
Build Validation	YAML schema, Python AST, and graph connectivity checks
Dark / Light Mode	Toggle between themes; preference is persisted in localStorage

Architecture

┌─────────────────────────────────────────┐
│          Frontend (React + TS)          │
│  ┌────────────┐  ┌───────────────────┐  │
│  │ Agent Builder│ │ Workflow Builder  │  │
│  │ (NL / Manual)│ │ (NL / Edit)      │  │
│  └─────┬──────┘  └────────┬──────────┘  │
└────────┼──────────────────┼─────────────┘
         │  REST + SSE      │
┌────────┼──────────────────┼─────────────┐
│        ▼                  ▼             │
│  ┌────────────────────────────────┐     │
│  │      Orchestrator Agent        │     │
│  │  (Azure OpenAI GPT-4o / 4.1)  │     │
│  └──────┬──────────────┬─────────┘     │
│         │              │               │
│  ┌──────▼─────┐ ┌──────▼──────────┐   │
│  │ agent-     │ │ workflow-       │   │
│  │ creator    │ │ creator         │   │
│  │ Skill      │ │ Skill           │   │
│  └────────────┘ └─────────────────┘   │
│        Backend (FastAPI + Python)       │
└────────────────────────────────────────┘

Skills Design

This project implements the Microsoft Agent Framework Skills specification:

agent-creator Skill

• SKILL.md — Agent creation guidelines & prompts
• scripts/validate_agent.py — YAML + Python validation
• references/AGENT_SCHEMA.md — Agent definition schema

workflow-creator Skill

• SKILL.md — Workflow creation & editing guidelines
• scripts/validate_workflow.py — YAML + Python validation (incl. graph connectivity)
• references/WORKFLOW_PATTERNS.md — Workflow pattern catalog

Quick Start

Prerequisites

• Python 3.11+
• Node.js 18+
• Azure OpenAI resource with a GPT-4o (or GPT-4.1) deployment
• Azure CLI (az login) for DefaultAzureCredential, or an API key

1. Clone the Repository

git clone https://github.com/<your-username>/agent-workflow-builder.git
cd agent-workflow-builder

2. Backend Setup

# Create virtual environment
python -m venv .venv

# Activate  (Windows PowerShell)
.\.venv\Scripts\Activate.ps1
# Activate  (macOS / Linux)
# source .venv/bin/activate

# Install dependencies
pip install -r backend/requirements.txt

# Configure environment
cp backend/.env.example backend/.env
# Edit backend/.env with your Azure OpenAI settings

3. Frontend Setup

cd frontend
npm install
cd ..

4. Run

# Terminal 1 — Backend
cd backend
uvicorn main:app --reload --port 8000

# Terminal 2 — Frontend
cd frontend
npm run dev

Open http://localhost:5173 in your browser.

Configuration

Copy backend/.env.example to backend/.env and set these variables:

Variable	Description	Default
AZURE_OPENAI_ENDPOINT	Your Azure OpenAI endpoint URL	(required)
AZURE_OPENAI_DEPLOYMENT	Model deployment name	gpt-4o
AZURE_OPENAI_API_VERSION	API version	2024-12-01-preview
AZURE_OPENAI_API_KEY	API key (optional if using DefaultAzureCredential)	—

Authentication: The backend tries DefaultAzureCredential first (via az login). If that fails, it falls back to the API key.

API Reference

Method	Endpoint	Description
GET	/api/health	Health check
POST	/api/agents	Create agent (NL prompt or explicit definition)
GET	/api/agents	List all agents
GET	/api/agents/{name}	Get agent detail (YAML + code)
PUT	/api/agents/{name}/edit	Edit agent via natural language
DELETE	/api/agents/{name}	Delete agent
POST	/api/agents/{name}/run	Run agent in playground (SSE stream)
POST	/api/workflows	Create workflow (NL prompt or explicit definition)
GET	/api/workflows	List all workflows
GET	/api/workflows/{name}	Get workflow detail (YAML + code)
PUT	/api/workflows/{name}/edit	Edit workflow via natural language
DELETE	/api/workflows/{name}	Delete workflow
POST	/api/workflows/{name}/run	Run workflow in playground (SSE stream)

Project Structure

.
├── backend/
│   ├── main.py              # FastAPI application & endpoints
│   ├── orchestrator.py       # LLM orchestration logic
│   ├── llm_client.py         # Azure OpenAI client wrapper
│   ├── models.py             # Pydantic request/response models
│   ├── config.py             # Environment configuration
│   ├── requirements.txt      # Python dependencies
│   ├── .env.example          # Environment template
│   ├── generated/            # Runtime output (gitignored)
│   └── skills/
│       ├── agent-creator/    # Agent creation skill
│       └── workflow-creator/ # Workflow creation skill
├── frontend/
│   ├── src/
│   │   ├── App.tsx           # Main layout, routing, dark mode
│   │   ├── index.css         # All styles (light + dark themes)
│   │   ├── components/
│   │   │   ├── AgentBuilderPage.tsx
│   │   │   ├── WorkflowBuilderPage.tsx
│   │   │   └── HomePage.tsx
│   │   └── api/
│   │       └── client.ts     # API client
│   ├── package.json
│   ├── tsconfig.json
│   └── vite.config.ts
├── test_api.py               # Integration tests
├── .gitignore
├── LICENSE
├── CONTRIBUTING.md
└── README.md

Tech Stack

Layer	Technology
Frontend	React 19, TypeScript, Vite 6, React Flow (@xyflow/react)
Backend	FastAPI, Python 3.11, Uvicorn
LLM	Azure OpenAI (GPT-4o / GPT-4.1)
Skills	Microsoft Agent Framework Skills specification
Storage	Local filesystem (YAML + Python files)

Running Tests

# Start the backend first, then:
python test_api.py

The test script exercises agent creation (NL & manual), workflow creation, editing, listing, and playground execution.

Contributing

See CONTRIBUTING.md for guidelines.

License

This project is licensed under the MIT License.