[SkyAgent] Upload initial refactored code

[SumanthRH]

What does this PR do?

Uploading our initial refactored code for SkyAgent

[gemini-code-assist]

Summary of Changes

Hello @SumanthRH, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request represents a substantial refactoring and rebranding effort, transforming the 'SkyRL' project into 'SkyAgent'. The primary goal is to establish a more generic and modular framework for developing and evaluating AI agents. This involved a comprehensive restructuring of the codebase, including a new core agent framework, consolidated environment configurations, a complete overhaul of build and deployment processes, and a complete refresh of examples and documentation to align with the new vision.

Highlights

• Project Rebranding and Scope Shift: The project has been rebranded from 'SkyRL' to 'SkyAgent', reflecting a shift towards a more generic agent layer for training and evaluating AI agents. The README has been completely rewritten to reflect this new focus and modular design.
• Core Agent Framework Overhaul: A new modular agent framework has been introduced under skyagent/skyagent/agents/, including base classes (AgentRunner, BaseTrajectory), a registry for agent and trajectory types, and specific implementations for OpenHands CodeAct and ReAct agents. This forms the new architectural foundation.
• Consolidated Environment and Build System: Environment variable management has been centralized into a new .env file, replacing older, specific configurations. All previous Dockerfiles have been removed, indicating a complete change in the build and deployment strategy.
• Restructured Examples and Documentation: Almost all old example scripts and their associated data preprocessing/configuration files have been removed. New example directories (run_openai, run_skyrl, run_verl) have been added, showcasing integrations with OpenAI-compatible servers, SkyRL-train, and VERL frameworks. The entire docs directory has also been removed, suggesting a complete documentation overhaul.
• Dependency and Build System Update: The pyproject.toml file has been significantly modified to reflect the project name change, update dependencies, and restructure optional dependencies, aligning with the new modular architecture and integrations.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in issue comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments or fill out our survey to provide feedback.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request introduces a significant refactoring and initial upload of the SkyAgent codebase. The changes are extensive, including file additions, deletions, and modifications, establishing a new baseline for the project.

My review has identified several issues that should be addressed to improve code quality, maintainability, and security:

• A critical security vulnerability due to committing a .env file to the repository.
• A critical build issue in pyproject.toml related to incorrect package versioning configuration.
• A high-severity bug in a data processing script that will likely lead to runtime errors.
• Several medium-severity issues related to code quality, such as dead code, magic numbers, and hardcoded paths.

I've provided specific comments and suggestions for each issue. Addressing these points will be crucial for the stability and security of the new codebase.

[gemini-code-assist]

Committing a .env file to version control, even with placeholder values, is a critical security risk. It encourages developers to place secrets in a tracked file, which can be accidentally committed.

The standard practice is to:

1. Rename this file to .env.example to serve as a template for other developers.
2. Add .env to your .gitignore file to prevent accidental commits of files containing real secrets.

[gemini-code-assist]

The [tool.setuptools.dynamic] section specifies that the version should be read from skyagent.__version__. This configuration has two issues:

1. The skyagent/__init__.py file does not define a __version__ attribute, which will cause a build failure.
2. According to PEP 621, if version is specified in the [project] table (as it is on line 16), it must not also be specified in the dynamic table.

To fix this, you should remove the version key from the [tool.setuptools.dynamic] table. The static version from the [project] table will then be used correctly.

[tool.setuptools.dynamic]
#version = {file = "verl/version/version"}

[gemini-code-assist]

The function extract_func_part and the variable method_set are defined but are not used anywhere in the script. This appears to be dead code and should be removed to keep the codebase clean and reduce clutter.

[gemini-code-assist]

The number 40 used here, and 2000 on line 76, are magic numbers. They make the code harder to understand and maintain because their purpose is not immediately clear.

Please define them as constants with descriptive names at the top of the script. For example:

MAX_ENTRIES_PER_IMAGE = 40
MAX_TRAIN_INSTANCES_TO_PROCESS = 2001 # or 2000 depending on the logic

[gemini-code-assist]

The variable input_dataset is being reassigned here. It was previously assigned on line 43. Reassigning variables in this way can make the code harder to follow and debug. To improve clarity, consider using a more descriptive name for this dataset, such as validation_input_dataset.

[gemini-code-assist]

The modules AsyncLLMEngine and AsyncEngineArgs are imported but are not used in this script. These unused imports should be removed to keep the code clean.

Suggested change

	from vllm.engine.async_llm_engine import AsyncLLMEngine
	from vllm import AsyncEngineArgs
	import os
	from skyagent import AutoAgentRunner
	from pathlib import Path
	from transformers import AutoTokenizer
	import datasets
	import asyncio

[gemini-code-assist]

The dataset path is hardcoded, which makes the script less flexible and harder to reuse in different environments. It's better to pass this path as a command-line argument, for instance, by using Python's argparse module.

[gemini-code-assist]

This assert statement is commented out with a note to "remove the assert". If it's no longer needed, it should be removed entirely to improve code clarity and reduce clutter.