Agent Orchestrator

Technology

Built with Microsoft Agent Framework

A2A Protocol v0.3.0 Compliant Server

Powered by the Microsoft Agent Framework, this orchestrator enables seamless agent-to-agent communication using the standardized A2A protocol. Built on top of FastAPI and the official A2A SDK, it provides dynamic workflow orchestration, multi-agent coordination, and comprehensive health monitoring for distributed agent systems.

Uses Agent Framework Workflows with a graph-based architecture for defining complex agent interactions. Workflows are composed of nodes (representing agents) and edges (representing connections and message flow between agents), enabling sophisticated multi-step reasoning and task delegation patterns.

GitHub Repository Documentation

API Endpoints & Documentation

Explore the A2A protocol endpoints, agent cards, and workflow orchestration APIs.

API Documentation Health Status A2A Agent Card

Agent Orchestrator Dashboard

Live monitoring of all subsystems and services. Updates every 30 seconds.

Loading system status...

🚀 Quick Start: Chat with Agent

Use this curl command to send a message to the orchestrator agent via the A2A protocol. The agent will return a task ID that you can use to check the status and results.

Terminal Command

curl -X POST http://localhost:5015/invoke \
-H "Content-Type: application/json" \
-d '{"input": {"message": "Explain quantum computing in simple terms"}}'

✅ Response:

{
  "task_id": "abc-123...",
  "task_url": "http://localhost:5015/tasks/abc-123...",
  "status": {"state": "running"}
}

📝 Next Step: Click or curl the task_url to check status and get results.

⚙️ Workflow Configuration

Configure the agent orchestrator to run multi-agent workflows.

Required Environment Variables

Configure these parameters to enable full orchestrator functionality:

Parameter	Required	Default	Description
Workflow Configuration
`WORKFLOW_NODES`	Required	`[]`	JSON array defining workflow nodes. Each node represents a step in the workflow (start, agent, end).
`WORKFLOW_EDGES`	Required	`[]`	JSON array defining connections between nodes. Controls workflow execution flow.
`WORKFLOW_AGENT_LIST`	Required	`[]`	JSON array of A2A agents. Defines available agents with their URLs and capabilities.
Azure OpenAI Configuration
`LLM_ENDPOINT`	Required	`""`	Azure OpenAI endpoint URL (e.g., https://your-resource.openai.azure.com/)
`LLM_KEY`	Required	`""`	Azure OpenAI API key for authentication
`LLM_DEPLOYMENT_NAME`	Optional	`gpt-4o`	Azure OpenAI deployment name
`LLM_VERSION`	Optional	`2024-02-01`	Azure OpenAI API version
`LLM_MODEL`	Optional	`gpt-4`	Azure OpenAI model name
Embedding Service Configuration
`EMBEDDING_BASE_URL`	Optional	`""`	Base URL for embedding service
`EMBEDDING_API_KEY`	Optional	`""`	API key for embedding service authentication
`EMBEDDING_PROVIDER`	Optional	`azure`	Embedding provider (azure, openai, etc.)
Database Configuration
`DB_HOST` or `PG_HOST`	Optional	`localhost`	PostgreSQL database host
`DB_PORT` or `PG_PORT`	Optional	`5432`	PostgreSQL database port
`DB_NAME`	Optional	`orchestrator`	PostgreSQL database name
`DB_USERNAME` or `PG_USER`	Optional	`postgres`	PostgreSQL database username
`DB_PASSWORD` or `PG_PASSWORD`	Optional	`""`	PostgreSQL database password
Server Configuration
`PORT`	Optional	`8080`	Server port (5015 for local development)
`HOST`	Optional	`0.0.0.0`	Server host address
`ENVIRONMENT`	Optional	`development`	Application environment (development, production)
`DEBUG`	Optional	`false`	Enable debug mode

Example Configuration

.env file

# Workflow Configuration
WORKFLOW_NODES='[{"id":"start","type":"start"},{"id":"agent1","type":"agent","agent_ref":"research-agent"}]'
WORKFLOW_EDGES='[{"source":"start","target":"agent1"}]'
WORKFLOW_AGENT_LIST='[{"agentName":"research-agent","agentUrl":"http://localhost:5016"}]'

# Azure OpenAI Configuration
LLM_ENDPOINT=https://your-openai.openai.azure.com/
LLM_KEY=your-api-key-here
LLM_DEPLOYMENT_NAME=gpt-4o

📚 Need Help?
For detailed workflow configuration and examples, see the DETAILS.md documentation