Add Docker virtual environments for reproducible development

.dockerignore

@@ -0,0 +1,86 @@
+# Docker build context exclusions for AgentGym-RL
+# Reduces build context size and speeds up image builds
+
+# Git
+.git
+.gitignore
+.gitmodules
+
+# IDE and editor files
+.vscode
+.idea
+*.swp
+*.swo
+.DS_Store
+
+# Python bytecode and caches
+__pycache__
+*.py[cod]
+*$py.class
+*.pyo
+.pytest_cache
+.mypy_cache
+.ruff_cache
+
+# Virtual environments
+.venv
+venv
+env
+ENV
+
+# Build artifacts
+*.egg-info
+*.egg
+dist
+build
+*.whl
+
+# Model checkpoints and large data (mount these instead)
+*.pt
+*.ckpt
+*.safetensors
+saves/
+checkpoints/
+models/*.bin
+models/*.safetensors
+
+# Training outputs
+wandb/
+outputs/
+logs/
+**/executer_logs/
+
+# Environment data (mount instead)
+AgentGym/agentenv-*/retrieve_data/
+milvus_data/
+
+# Documentation build
+docs/_build
+
+# Test artifacts
+htmlcov
+.coverage
+coverage.xml
+
+# Temporary files
+*.tmp
+*.temp
+*.log
+wget-log
+
+# Jupyter
+.ipynb_checkpoints
+
+# Archive files
+*.tar.gz
+*.zip
+*.tar
+
+# Lock files
+*.lock
+uv.lock
+
+# Local configuration
+.env
+.env.local
+*.code-workspace

.env.example

@@ -0,0 +1,73 @@
+# AgentGym-RL Environment Configuration
+# Copy this file to .env and fill in your values
+
+# =============================================================================
+# API KEYS
+# =============================================================================
+
+# OpenAI API (required for evaluation)
+OPENAI_API_KEY=your-openai-api-key-here
+OPENAI_BASE_URL=https://api.openai.com/v1
+
+# Weights & Biases (optional for training logging)
+WANDB_API_KEY=
+WANDB_MODE=offline
+
+# HuggingFace (optional for model uploads)
+HF_TOKEN=
+
+# =============================================================================
+# ENVIRONMENT CONFIGURATION
+# =============================================================================
+
+# Environment to use: searchqa, babyai, sciworld, textcraft, webarena
+ENV=searchqa
+
+# Environment server port
+ENV_PORT=36001
+
+# Environment server URL (for evaluation)
+ENV_SERVER_BASE=http://localhost:36001
+
+# =============================================================================
+# MODEL CONFIGURATION
+# =============================================================================
+
+# Model name for evaluation
+MODEL_NAME=gpt-4o-mini
+
+# Maximum interaction rounds
+MAX_ROUND=10
+
+# =============================================================================
+# TRAINING CONFIGURATION
+# =============================================================================
+
+# VLLM settings
+VLLM_USE_MODELSCOPE=0
+VLLM_WORKER_MULTIPROC_METHOD=spawn
+VLLM_ATTENTION_BACKEND=XFORMERS
+
+# PyTorch CUDA settings
+PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True
+
+# =============================================================================
+# SCRIPTS/UTILITIES CONFIGURATION
+# =============================================================================
+
+# Model merger settings
+LOCAL_DIR=saves/checkpoint
+SAVE_DIR=
+HF_UPLOAD_PATH=
+
+# =============================================================================
+# DATA CONFIGURATION
+# =============================================================================
+
+# Evaluation settings
+INFERENCE_FILE=searchqa_eval_sample.json
+DATA_LEN=100
+TIMEOUT=2400
+MAX_TOKENS=4096
+TEMPERATURE=0.7
+TOP_P=1