Skip to content

Latest commit

 

History

History
149 lines (117 loc) · 4.93 KB

README.md

File metadata and controls

149 lines (117 loc) · 4.93 KB

Concurrent Agents Pattern

This demo showcases the Concurrent Agents pattern using the Azure AI Agent Framework. Concurrent workflows execute multiple agents in parallel (fan-out) and aggregate their results (fan-in), enabling parallel processing for improved efficiency.

Features

  • Fan-Out Execution: Same prompt dispatched to multiple agents simultaneously
  • Fan-In Aggregation: Results from all agents combined into final output
  • Parallel Processing: Multiple domain experts work concurrently
  • Custom Executors: Extend with specialized agent executor classes
  • Custom Aggregators: Control how results are combined

Samples Included

File Description
concurrent_agents.py Basic concurrent workflow with researcher, marketer, and legal agents
concurrent_custom_agent_executors.py Using custom executor classes that own ChatAgents
concurrent_custom_aggregator.py Custom aggregation logic for combining results
concurrent_participant_factory.py Participant factories for stateful workflows
concurrent_workflow_as_agent.py Wrap concurrent workflow as a reusable agent

Prerequisites

  • Python 3.10 or later
  • Azure OpenAI resource with a deployed model
  • Azure CLI authenticated (az login)

Setup

See the main README for complete setup instructions. Quick start:

# From project root
cp .env.example .env  # Then edit with your Azure OpenAI settings
python -m venv venv
venv\Scripts\activate  # Windows (or source venv/bin/activate on Linux/Mac)
pip install -r requirements.txt
az login

Running the Demos

# Basic concurrent workflow
python concurrent_agents.py

# With custom agent executors
python concurrent_custom_agent_executors.py

# With custom aggregator
python concurrent_custom_aggregator.py

# With participant factories
python concurrent_participant_factory.py

# Workflow as reusable agent
python concurrent_workflow_as_agent.py

How It Works

Concurrent Fan-Out/Fan-In Pattern

                    User Input
                        │
                   [Dispatcher]
                   /    │    \
                  ↓     ↓     ↓
            Researcher Marketer Legal
                  \     │     /
                   [Aggregator]
                        │
                   Final Output

Architecture

  1. ConcurrentBuilder: Constructs the workflow with parallel participants
  2. Default Dispatcher: Fans out the same user prompt to all agents
  3. Parallel Execution: All agents process the input simultaneously
  4. Default Aggregator: Combines results into list[ChatMessage]

Sample Output

===== Final Aggregated Conversation (messages) =====
------------------------------------------------------------
01 [user]:
We are launching a new budget-friendly electric bike for urban commuters.
------------------------------------------------------------
02 [researcher]:
**Insights:**
- Target Demographic: Urban commuters seeking affordable, eco-friendly transport
- Market Trends: E-bike sales growing with urbanization and sustainability focus
...
------------------------------------------------------------
03 [marketer]:
**Value Proposition:**
"Freedom to commute without breaking the bank"
...
------------------------------------------------------------
04 [legal]:
**Compliance Considerations:**
- Consumer safety regulations for electric vehicles
- Battery disposal and recycling requirements
...

Key Concepts

  • Domain Agents: Each agent specializes in a specific perspective (research, marketing, legal)
  • Custom Executors: Create classes that own ChatAgents with @handler decorators
  • AgentExecutorRequest/Response: Standard input/output types for custom executors
  • Workflow Completion: Completes when all participants become idle

Use Cases

  1. Multi-Perspective Analysis: Get insights from different domain experts
  2. Parallel Research: Multiple researchers investigating different aspects
  3. Review Processes: Legal, technical, and business reviews in parallel
  4. Content Generation: Multiple writers creating variations simultaneously

Troubleshooting

Authentication errors:

az login --tenant <your-tenant-id>

Module not found: Ensure root-level virtual environment is activated:

cd ..  # Go to project root
venv\Scripts\activate
cd 02-ConcurrentAgents
python concurrent_agents.py

Azure OpenAI errors:

  • Verify AZURE_OPENAI_ENDPOINT is correct
  • Ensure your deployment name matches AZURE_OPENAI_CHAT_DEPLOYMENT_NAME
  • Check that your Azure subscription has access to Azure OpenAI

Learn More