researchflow

researchflow is a command-line tool designed to streamline research workflows by automating notifications for script executions and generating AI-powered reviews of research papers.

README in Korean can be found here.

1. Setup

First, clone the repository and install the necessary dependencies.

git clone https://github.com/mjk0618/researchflow.git
cd researchflow
pip install -e .

2. Usage Examples

2.1 alarm: Script Execution Notifier

The alarm command executes a target Python script and sends a notification to a Slack channel with the results (e.g., success or failure, execution time, arguments).

You can test the alarm functionality by running the sample script from the project's root directory:

alarm examples/sample_script.py

• Background Execution (--log): Using the --log flag runs the script in the background and creates a log file, similar to nohup.

• Reporting Arguments: To include the script's arguments in the Slack notification, import and call the report_arguments function from researchflow.core.utils after parsing arguments in your script.

  # In your script (e.g., examples/sample_script.py)
  from researchflow.core.utils import report_arguments
  
  def main():
      # ... (argument parsing code)
      args = parser.parse_args()
      report_arguments(args) # Call this function
      # ... (rest of your script)

2.2 review: AI-Powered Paper Reviewer

The review command uses the Gemini API to review a research paper from an arXiv URL and sends a formatted summary to a Slack channel.

• paper_input: You must provide the arXiv URL for the paper you want to review.
• --user-interests (or -u): You can specify your research interests to tailor the review's evaluation.

Here’s how to review the famous "Attention Is All You Need" paper:

# Basic review
review https://arxiv.org/pdf/1706.03762

# Review with user interests
review https://arxiv.org/pdf/1706.03762 -u "LLM" "NLP"

Note: The review feature is under development and may be unstable.

3. Configuration

3.1 Setting Up Environment Variables

Environment variables are required for Slack notifications and Gemini API integration. Create a .env file in the root directory of the project and add your credentials as shown below.

Directory Structure:

researchflow/
├── examples
├── pyproject.toml
├── README.md
├── researchflow
└── .env (Create this file)

.env file content:

SLACK_WEBHOOK_URL="YOUR_SLACK_WEBHOOK_URL"
GEMINI_API_KEY="YOUR_GEMINI_API_KEY"

3.2 Obtaining a Slack Webhook URL

Follow these steps to generate a Slack Webhook URL, which is required for the alarm and review commands to send messages to a Slack channel.

1. Create or Select a Slack App:
   • Navigate to the Slack API website.
   • Click "Create New App" or select an existing one.
   • When creating a new app, choose "From scratch," provide an app name, and select your workspace.
2. Enable Incoming Webhooks:
   • In the app's settings page, go to "Incoming Webhooks" under the "Features" section in the left sidebar.
   • Toggle the "Activate Incoming Webhooks" switch to "On".
3. Add a New Webhook URL:
   • Click the "Add New Webhook to Workspace" button.
   • Choose the channel where messages will be posted and click "Allow."
4. Copy the Webhook URL:
   • Copy the generated Webhook URL and paste it as the value for SLACK_WEBHOOK_URL in your .env file.

For more details, refer to the official documentation on Sending messages using Incoming Webhooks.

3.3 Obtaining a Gemini API Key

An API key is necessary for the review command. Follow these steps to get your Gemini API key.

1. Go to Google AI Studio:
   • Visit the Google AI Studio website.
2. Generate an API Key:
   • Click "Get API key" in the top right corner.
   • On the next page, click "Create API key".
3. Copy the API Key:
   • Copy the generated key and paste it as the value for GEMINI_API_KEY in your .env file.

For more details, refer to the Generative AI Quickstart guide.