Skip to content

Latest commit

 

History

History
121 lines (82 loc) · 4.73 KB

README.md

File metadata and controls

121 lines (82 loc) · 4.73 KB
Raw

AutoGPT-RabbitMQ

This plugin allows you to communicate with your Auto-GPT instance via microservice.

You can try it right now for free https://ai-autopilot.io

📚 Requirements

  1. Python Package: Install the pika Python package:

    pip3 install pika

    or add pika module in the Auto-GPT requirements.txt file.

  2. RabbitMQ: Just start one RabbitMQ server thanks to this download page https://www.rabbitmq.com/download.html

⚙️ Installation

Follow these steps to configure the Auto-GPT RabbitMQ Plugin:

  1. Clone this Repository

    git clone https://github.com/tomtom94/AutoGPT-RabbitMQ.git

  2. Navigate to the folder

    cd AutoGPT-RabbitMQ

  3. Zip the rabbitmq_plugin folder

    • On MacOS, right click the rabbitmq_plugin folder and press Compress.
    • On windows, right click the folder, and press Send to > Compressed (zipped).

  4. Move the zip file

    Move the new rabbitmq_plugin.zip file to the Auto-GPT plugins directory, there should already be a file there titled __PUT_PLUGIN_ZIPS_HERE__.

🔧 Configuration

  1. Add new var env in the .env file in Auto-GPT:

    ################################################################################
### RABBITMQ PLUGIN SETTINGS
################################################################################
RABBITMQ_HOST=localhost
QUEUE_TO_RECEIVE_MESSAGE=service-to-autogpt
QUEUE_TO_SEND_MESSAGE=autogpt-to-service
    • RABBITMQ_HOST: The RabbitMQ connection string
    • QUEUE_TO_RECEIVE_MESSAGE: Receive a message from the service to Auto-GPT via plugin
    • QUEUE_TO_SEND_MESSAGE: Send a message from Auto-GPT to the service via plugin

  2. Update var env in the .env file in Auto-GPT:

    ################################################################################
### CHAT PLUGIN SETTINGS
################################################################################
# CHAT_MESSAGES_ENABLED - Enable chat messages (Default: False)
CHAT_MESSAGES_ENABLED=True
    • CHAT_MESSAGES_ENABLED: By default it's False, this plugin needs this var to be True otherwise this plugin would be useless

  3. Update var env in the .env file in Auto-GPT:

    ################################################################################
ALLOWLISTED PLUGINS
################################################################################

#ALLOWLISTED_PLUGINS - Sets the listed plugins that are allowed (Example: plugin1,plugin2,plugin3)
ALLOWLISTED_PLUGINS=AutoGPTRabbitMQ

► Listen RabbitMQ on your client microservice

  1. Use the same env vars in your service, but reverse their value between QUEUE_TO_RECEIVE_MESSAGE & QUEUE_TO_SEND_MESSAGE

    RABBITMQ_HOST=localhost
QUEUE_TO_RECEIVE_MESSAGE=autogpt-to-service
QUEUE_TO_SEND_MESSAGE=service-to-autogpt

  2. First Auto-GPT talks to you, so listen to it via a consumer in a thread

    This plugin RabbitMQ for AutoGPT has for mission to send all the ouput into the designed for QUEUE_TO_SEND_MESSAGE, The format sent are of type :

    {'role': 'POST_COMMAND', 'content': 'The message Auto-GPT has written, given line by line'}
{'role': 'POST_INSTRUCTION', 'content': '...'}
{'role': 'REPORT', 'content': '...'}

    Your service needs to work the same way the plugin works with RabbitMQ via pika.

    • It means you need to create a consumer of that kind start_consumer which will make a thread of this function run_consumer. Insteand of this code userReply.append(body.decode()) just use straight body.decode() to receive the prompt line by line, and do whatever you want with.
    • And then send message when you wanna reply to the prompt send_message
    • With the same env vars but remember we reversed their value, so just copy past these 3 functions, that's it.

    There are special kill codes, sent sometimes from one procuder to a consumer, depends which process triggers the end :

    • SIGTERM_FROM_SERVICE
    • SIGTERM_FROM_AUTOGPT

    It allows to kill the consumer in a thread from the inside properly. Just need to mirror this pattern on your service side.

  3. Finally Auto-GPT asks you to reply, so send a message in the queue for that

    Once everything has been started properly, and is able to stop properly. Just send a message from your service to Auto-GPT like this function send_message does from Auto-GPT to your service. Just need to mirror this pattern on your service side.