AmitKumar2801/pytorch-hackathon-2026

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

Advanced Cognitive Agent Environment (OpenEnv)

Meta PyTorch Hackathon (Scaler School of Technology) - Round 1 Submission

📌 Overview

This repository hosts a robust, containerized environment designed to evaluate autonomous LLM reasoning capabilities. Built using the OpenEnv specification, it simulates a dynamic real-world sequential task execution environment suitable for AI agent interaction and training.

🚀 Key Features

• —Standardized Logging: Implements professional logging modules utilizing [START], [STEP], and [END] markers for precise automated scoring.
• —API Readiness: Built with FastAPI, exposing a /health endpoint for orchestration and readiness probes.
• —Configurable Dynamics: Fully integrated with API_BASE_URL and MODEL_NAME via environment variables.
• —Dockerized: Packaged for seamless deployment on Hugging Face Spaces.

📁 Repository Structure

• —inference.py: Core execution script managing the agent's evaluation sequence and state logging.
• —openenv.yaml: Metadata configuration defining the environment's capabilities and sequential task definitions.
• —main.py: FastAPI server script.
• —Dockerfile & requirements.txt: Environment dependencies and container configuration.

🛠️ Usage

To run the environment locally:

pip install -r requirements.txt
python inference.py