# Environment Configuration for Qwen3-VL-2B-FP8 Deployment
# Copy this file to .env and update with your values

# ============================================
# HUGGINGFACE CONFIGURATION
# ============================================

# Your HuggingFace access token (required for some models)
# Get your token from: https://huggingface.co/settings/tokens
HF_TOKEN=your_huggingface_token_here

# ============================================
# VLLM CONFIGURATION
# ============================================

# GPU memory utilization (0.0 to 1.0)
# Default: 0.85
VLLM_GPU_MEMORY_UTILIZATION=0.85

# Maximum model length in tokens
# Default: 8192 (Qwen3-VL supports up to 131072)
VLLM_MAX_MODEL_LEN=8192

# Maximum number of concurrent sequences
# Default: 8
VLLM_MAX_NUM_SEQS=8

# Host to bind the vLLM server
# Default: 0.0.0.0
VLLM_HOST=0.0.0.0

# Port for vLLM API
# Default: 8000
VLLM_PORT=8000

# ============================================
# STREAMLIT CONFIGURATION
# ============================================

# Port for Streamlit app
# Default: 8501
STREAMLIT_PORT=8501

# ============================================
# NETWORK CONFIGURATION
# ============================================

# Tailscale network (if using)
# Example: 100.x.x.x
TAILSCALE_NETWORK=

# ============================================
# MONITORING (Optional)
# ============================================

# Enable Prometheus metrics
PROMETHEUS_ENABLED=false
PROMETHEUS_PORT=9090