This project demonstrates a multi-agent system (MAS) implemented in Python, where five specialized agents collaborate asynchronously to solve tasks. The agents include Coordinator, ResearchAgent, PlannerAgent, ExecutorAgent, and ReporterAgent, each handling distinct responsibilities. The system integrates tools such as a web fetcher, calculator, and file writer, and is designed to be LangGraph-friendly for future integration. The demo highlights modular design, clear separation of roles, asynchronous coordination, and extensibility, making it a concise and presentable showcase of multi-agent collaboration.
Multi-agent systems (MAS) provide an effective paradigm for distributing complex tasks among autonomous entities. Each agent operates semi-independently but communicates and coordinates with others to achieve shared objectives. This project presents a small-scale MAS designed for demonstration purposes, emphasizing:
Role specialization
Asynchronous task execution
Tool-based operations (web fetch, calculation, file writing)
Optional LangGraph integration
The system aims to showcase collaboration, modularity, and scalability, providing a foundation for more advanced MAS implementations in AI workflows.
Agent Architecture
CoordinatorAgent: Assigns tasks, monitors progress, and orchestrates communication.
ResearchAgent: Retrieves information from external sources using a web fetcher.
PlannerAgent: Converts research into structured action plans.
ExecutorAgent: Performs calculations, executes tasks, and writes files using integrated tools.
ReporterAgent: Compiles outputs and generates reports or summaries.
System Design
Agents communicate asynchronously via an in-memory message bus.
Tools are modular, enabling easy reuse and extension.
LangGraph detection is built-in to log integration potential without requiring installation.
Python’s asyncio library is used to run agents concurrently and avoid blocking operations.
Setup
Five agents were instantiated and coordinated via the orchestrator (main.py).
Sample tasks included fetching data, planning workflows, executing calculations, and generating summaries.
Functional tests were performed for web fetching, calculations, and file writing.
Test Scenarios
Simple Task Execution: Agents collaborated to fetch, calculate, and report results.
Multi-Step Workflow: Dependent tasks were executed sequentially across agents.
LangGraph Detection: Verified optional integration hooks without LangGraph installation.
Task Completion: All assigned tasks were executed successfully.
Asynchronous Operation: Agents ran concurrently without deadlocks or conflicts.
Tool Integration: Web fetch, calculator, and file writing worked seamlessly across agents.
Modularity & Extensibility: New agents or tools could be added with minimal changes.
LangGraph Readiness: System correctly logged potential LangGraph integration points.
The multi-agent demo successfully illustrates a collaborative, asynchronous MAS with specialized agents and modular tools. It demonstrates clear role separation, effective communication, and seamless integration of tools, making it an ideal showcase for employers. The system is designed to be extensible and LangGraph-friendly, allowing future integration into AI pipelines. Potential enhancements include full LangGraph integration, additional agent specializations, advanced task orchestration, and improved logging/reporting. Overall, this demo provides a concise, practical example of multi-agent collaboration in AI systems.