Skip to main content

mango

PyPi | Read the Docs | Github | mail

Note: This project is still in an early development stage. We appreciate constructive feedback and suggestions for improvement.

mango (modular python agent framework) is a python library for multi-agent systems (MAS). It is written on top of asyncio and is released under the MIT license.

mango allows the user to create simple agents with little effort and in the same time offers options to structure agents with complex behaviour.

The main features of mango are:

  • Container mechanism to speedup local message exchange
  • Message definition based on the FIPA ACL standard
  • Structuring complex agents with loose coupling and agent roles
  • Built-in codecs: JSON and protobuf
  • Supports communication between agents directly via TCP or via an external MQTT broker

A detailed documentation for this project can be found at mango-agents.readthedocs.io

Installation

mango requires Python >= 3.8 and runs on Linux, OSX and Windows.

For installation of mango you should use virtualenv which can create isolated Python environments for different projects.

It is also recommended to install virtualenvwrapper which makes it easier to manage different virtual environments.

Once you have created a virtual environment you can just run pip to install it:

$ pip install mango-agents

Getting started

####Creating an agent

In our first example, we'll create a very simple agent that simply prints the content of all messages it receives:

    from mango import Agent

    class RepeatingAgent(Agent):
        def __init__(self, container):
            # We must pass a ref of the container to "mango.Agent":
            super().__init__(container)
            print(f"Hello world! My id is {self.aid}.")

        def handle_message(self, content, meta):
            # This method defines what the agent will do with incoming messages.
            print(f"Received a message with the following content: {content}")

Agents must be a subclass of Agent. This base class needs a reference to the container the agents live in, so you must forward a container argument to it if you override __init__().

####Creating a container

Agents live in a container, so we need to know how to create a mango container. The container is responsible for message exchange between agents.

    # Containers need to be started via a factory function.
    # This method is a coroutine so it needs to be called using the
    # await statement
    first_container = await create_container(addr=('localhost', 5555))

This is how a container is created. Since the method create_container() is a coroutine we need to await its result.

Running your first agent within a container

To put it all together we will wrap the creation of a container and the agent into a coroutine and execute it using asyncio.run(). The following script will create a RepeatingAgent and let it run within a container for three seconds and then shutdown the container:

    import asyncio
    from mango import Agent, create_container

    class RepeatingAgent(Agent):
            def __init__(self, container):
                # We must pass a ref. to the container to "mango.Agent":
                super().__init__(container)
                print(f"Hello world! My id is {self.aid}.")

            def handle_message(self, content, meta):
                # This method defines what the agent will do with incoming messages.
                print(f"Received a message with the following content: {content}")

    async def run_container_and_agent(addr, duration):
        first_container = await create_container(addr=addr)
        first_agent = RepeatingAgent(first_container)
        await asyncio.sleep(duration)
        await first_agent.shutdown()
        await first_container.shutdown()

    asyncio.run(run_container_and_agent(addr=('localhost', 5555), duration=3))

The only output you should see is Hello world! My id is agent0., because the agent does yet not receive any messages.

Creating a proactive Agent

Let's implement another agent that is able to send a hello world message to another agent:

    from mango import Agent
    from mango.util.scheduling import InstantScheduledTask

        class HelloWorldAgent(Agent):
            def __init__(self, container, other_addr, other_id):
                super().__init__(container)
                self.schedule_instant_acl_message(
                    receiver_addr=other_addr,
                    receiver_id=other_id,
                    content="Hello world!")
                )

            def handle_message(self, content, meta):
                print(f"Received a message with the following content: {content}")

Connecting two agents

We can now connect an instance of a HelloWorldAgent with an instance of a RepeatingAgent and run them.

import asyncio
from mango import Agent, create_container
from mango.util.scheduling import InstantScheduledTask


class RepeatingAgent(Agent):
    def __init__(self, container):
        # We must pass a ref. to the container to "mango.Agent":
        super().__init__(container)
        print(f"Hello world! My id is {self.aid}.")

    def handle_message(self, content, meta):
        # This method defines what the agent will do with incoming messages.
        print(f"Received a message with the following content: {content}")

class HelloWorldAgent(Agent):
    def __init__(self, container, other_addr, other_id):
        super().__init__(container)
        self.schedule_instant_acl_message(
            receiver_addr=other_addr,
            receiver_id=other_id,
            content="Hello world!")
        )

    def handle_message(self, content, meta):
        print(f"Received a message with the following content: {content}")


async def run_container_and_two_agents(first_addr, second_addr):
    first_container = await create_container(addr=first_addr)
    second_container = await create_container(addr=second_addr)
    first_agent = RepeatingAgent(first_container)
    second_agent = HelloWorldAgent(second_container, first_container.addr, first_agent.aid)
    await asyncio.sleep(1)
    await first_agent.shutdown()
    await second_agent.shutdown()
    await first_container.shutdown()
    await second_container.shutdown()


def test_second_example():
    asyncio.run(run_container_and_two_agents(
        first_addr=('localhost', 5555), second_addr=('localhost', 5556))
    )
    

You should now see the following output:

Hello world! My id is agent0.

Received a message with the following content: Hello world!

You have now successfully created two agents and connected them.

Support

License

Distributed under the MIT license.

Metadata

Release files for mango-agents-assume 1.1.4.post10

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for mango-agents-assume 1.1.4.post10
File Size Uploaded
mango_agents_assume-1.1.4.post10.tar.gz 159.5 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for mango-agents-assume 1.1.4.post10
File Interpreter ABI Platform
mango_agents_assume-1.1.4.post10-py3-none-any.whl Python 3 none any Details

Total release size: 221.9 kB

Release files / mango_agents_assume-1.1.4.post10.tar.gz

Download URL mango_agents_assume-1.1.4.post10.tar.gz
Size 159.5 kB
Tags Source
SHA-256 checksum
How to use checksums
d27eeabf6a80974621eea2d0767abf1ca42838f2634a8b79cde9444137a357e2
BLAKE2b-256 checksum
How to use checksums
66264156ccfd74f7738f828aa21d39393f6d87cca8a852ca507569e26f346657
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/5.1.1 CPython/3.12.5

Release files / mango_agents_assume-1.1.4.post10-py3-none-any.whl

Download URL mango_agents_assume-1.1.4.post10-py3-none-any.whl
Size 62.4 kB
Tags Python 3
SHA-256 checksum
How to use checksums
3dab99cc3a6b420a3b63d11d7a26e5545810159963c35cd2143bc37b9397cf68
BLAKE2b-256 checksum
How to use checksums
845dbf5b8de14477a0fafae68dc03c99d9ad5c1218b4aa4eb60f447fd078b729
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/5.1.1 CPython/3.12.5
Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page