LLM-based Automation Agent

Overview

This project is an automation agent that executes plain-English tasks by leveraging a Large Language Model (LLM). The agent processes log files, reports, and other structured/unstructured data to automate routine tasks within the Continuous Integration (CI) pipeline of DataWorks Solutions. The agent exposes an API that enables users to run specific tasks and retrieve output files.

Features

• Supports multiple deterministic and non-deterministic automation tasks.
• Integrates an LLM (GPT-4o-Mini) for task interpretation and execution.
• Provides API endpoints to execute tasks and read output files.
• Ensures security by restricting data access to the /data directory and preventing file deletion.
• Packaged as a Docker image for easy deployment.

API Endpoints

1. Execute Task

Endpoint: POST /run?task=<task description>
Description: Executes a plain-English task by parsing and performing the required steps.

• Success (200 OK): Task executed successfully.
• Client Error (400 Bad Request): Invalid task description.
• Server Error (500 Internal Server Error): Internal error in execution.

2. Read File Content

Endpoint: GET /read?path=<file path>
Description: Retrieves the content of the specified file.

• Success (200 OK): Returns the file content.
• Not Found (404 Not Found): File does not exist.

Tasks Implemented

Phase A: Operations Tasks

• A1: Install uv (if required) and run datagen.py with user.email as an argument.
• A2: Format /data/format.md using prettier@3.4.2.
• A3: Count the number of Wednesdays in /data/dates.txt and write to /data/dates-wednesdays.txt.
• A4: Sort contacts in /data/contacts.json by last_name and first_name.
• A5: Extract the first line from the 10 most recent .log files in /data/logs/.
• A6: Index Markdown files in /data/docs/ by extracting H1 titles.
• A7: Extract sender’s email address from /data/email.txt using an LLM.
• A8: Extract credit card number from /data/credit-card.png using an LLM.
• A9: Find the most similar pair of comments in /data/comments.txt using embeddings.
• A10: Calculate total sales for "Gold" ticket type from /data/ticket-sales.db.

Installation & Setup

1. Clone the repository:

   git clone https://github.com/mohitrajrathor/AutoLLM-Agent.git
   cd AutoLLM-Agent

2. Install dependencies:

   pip install -r requirements.txt

3. Run the application:

   python main.py

4. Set the environment variable for AI Proxy:

   export AIPROXY_TOKEN=your-token-here

5. Run the application with Docker:

   docker run --rm -e AIPROXY_TOKEN=$AIPROXY_TOKEN -p 8000:8000 user-name/repo-name

License

This project is licensed under the MIT License - see the LICENSE file for details.