Media Agents is a Python-based project that implements Llama-3.1 / GPT4 based LLM assistant which helps journalists to find news leads and materials for news based on CourtListener online base. This system utilizes natural language processing and a graph-based workflow to find newsworthy stories, extract key facts and references, and compose draft with headline from various opinions and sources.

• Automated search of newsworthy stories.
• Building support materials: key facts, key entities (mentioned events & locations, involved people & organizations), references to original legal documents.
• Updates by email.

Before using Media Agents, ensure you have the following:

• Python 3.10 or higher installed.
• media_agents package installed (see the Installation section).
• Setup env variables
• Setup email recipients' list.

Clone the repository:

git clone https://github.com/lliryc/media_agents.git
cd media_agents
python3 setup.py install

The .env file is used to store environment variables that the application needs. Copy and rename .env.example as .env file in the current directory of the project and define values:

LLM_CLIENT=YOUR_LLM_CLIENT_MODEL_NAME # examples: accounts/fireworks/models/llama-v3p1-405b-instruct, gpt-4-turbo
OPENAI_API_KEY=YOUR_OPENAI_API_KEY # for gpt-4 turbo
FIREWORKS_API_KEY=YOUR_FIREWORKS_API_KEY # for llama3.1 405b
SMTP_SERVER=YOUR_SMTP_SERVER # 
SMTP_PORT_SSL=YOUR_SMTP_PORT_SSL
SMTP_USER=YOUR_SMTP_USER
SMTP_PASSWORD=YOUR_SMTP_PASSWORD

The recipients.txt file defines a list of email subscribers. Copy and rename recipients.txt.example in subscriptions folder as recipients.txt and set up your values:

[email protected]  
[email protected]  
[email protected]

To run a program type following command in the console

python3 app/app.py

After that AI assistant will start processing recent updates from CourtListener portal. Usually it takes about 2-3 hours to process recently published legal documents. Once all documents was processed, your recepients will get a email with news leads and support materials for news writing.

To launch a program as a Docker container use following command

bash start_container.sh

Please ensure you logged in on Docker Hub

Feel free to open issues or submit pull requests if you have suggestions or improvements.

This project is licensed under the Apache License (Version 2.0, January 2004) - see the LICENSE file for details.