Argonauts - Track 1 - Submission 12

Our project is a Dockerized AI agent designed to automate the execution of structured natural language tasks using large language models. The agent accepts a JSON file containing multiple prompts as input, processes each task sequentially through the Fireworks AI API, and generates concise, accurate responses in a standardized JSON output format. By packaging the entire application inside a Docker container, we ensure that it runs consistently across different operating systems and environments without requiring users to manually configure dependencies or runtime settings. The system is built in Python using the OpenAI-compatible Fireworks SDK and follows a simple yet robust pipeline: reading task definitions, invoking the selected language model, collecting responses, validating the generated output, and writing the final results to the required output directory. Environment variables are used for securely configuring API credentials and model selection, making the solution portable and easy to deploy. The project emphasizes reproducibility, reliability, and ease of evaluation. Its containerized architecture allows organizers or users to execute the agent using a single Docker command while mounting input and output directories, enabling seamless integration into automated benchmarking pipelines. This lightweight, modular design can also be extended to support additional models, advanced prompting strategies, parallel task execution, or post-processing workflows, making it a flexible foundation for future AI-powered automation tasks.