# ACE-Step Environment Configuration
# Copy this file to .env and modify as needed
#
# This file is used by:
# - Python scripts (acestep_v15_pipeline.py, api_server.py, etc.)
# - Windows launcher (start_gradio_ui.bat)
# - Linux/macOS launchers (start_gradio_ui.sh, start_gradio_ui_macos.sh)
#
# Settings in .env will survive repository updates, unlike hardcoded values
# in launcher scripts which get overwritten on each update.

# ==================== Model Settings ====================
# DiT model path
ACESTEP_CONFIG_PATH=acestep-v15-turbo

# LM model path (used when LLM is enabled)
# Available: acestep-5Hz-lm-0.6B, acestep-5Hz-lm-1.7B, acestep-5Hz-lm-4B
ACESTEP_LM_MODEL_PATH=acestep-5Hz-lm-1.7B

# Device selection: auto, cuda, cpu, xpu
ACESTEP_DEVICE=auto

# LM backend: vllm (faster) or pt (PyTorch native)
ACESTEP_LM_BACKEND=vllm

# ==================== LLM Initialization ====================
# Controls whether to initialize the Language Model (LLM/5Hz LM)
#
# Flow: GPU detection (full) → ACESTEP_INIT_LLM override → Model loading
# GPU optimizations (offload, quantization, batch limits) are ALWAYS applied.
# ACESTEP_INIT_LLM only overrides the "should we try to load LLM" decision.
#
# Values:
#   auto (or empty)    = Use GPU auto-detection result (default)
#   true/1/yes         = Force enable LLM after GPU detection (may cause OOM)
#   false/0/no         = Force disable LLM (pure DiT mode, faster)
#
# Examples:
#   ACESTEP_INIT_LLM=auto       # Let GPU detection decide (default)
#   ACESTEP_INIT_LLM=           # Same as auto
#   ACESTEP_INIT_LLM=true       # Force enable even on low VRAM GPU
#   ACESTEP_INIT_LLM=false      # Force disable for pure DiT mode
#
# When LLM is disabled, these features are unavailable:
#   - Thinking mode (thinking=true)
#   - Chain-of-Thought caption/language detection
#   - Sample mode (generate from description)
#   - Format mode (LLM-enhanced input)
#
# Default: auto (based on GPU VRAM detection). For a 12 GB Forge deployment,
# setting this to false is a deliberate core-generation profile: it avoids
# loading the optional 5 Hz planner / LLM. It does not select or reserve a GPU.
ACESTEP_INIT_LLM=auto

# ==================== Model Storage ====================
# Shared checkpoints directory (useful when running multiple installations)
# Point all installations to the same folder to avoid duplicate model downloads.
# Example: ACESTEP_CHECKPOINTS_DIR=~/ace-step-models
# ACESTEP_CHECKPOINTS_DIR=

# ==================== Download Settings ====================
# Preferred download source: auto, huggingface, modelscope
# ACESTEP_DOWNLOAD_SOURCE=auto

# ==================== API Server Settings ====================
# API key for authentication (optional)
# ACESTEP_API_KEY=sk-your-secret-key

# ==================== Docker Forge ====================
# ACE-Step Forge is available at http://localhost:3000 by default.
# FORGE_PORT=3000
# API_PORT=8001
# GRADIO_PORT=7860
# Restrict ACE-Step containers to one host GPU. Use `nvidia-smi -L` to choose
# an index or, preferably, a stable GPU UUID; the Compose default is GPU 0.
# Docker Desktop can enumerate numeric GPU indices differently from the host,
# so a UUID is the safer choice for multi-GPU machines.
# ACESTEP_GPU_DEVICE_ID=0
# 12 GB Forge core-generation profile (uncomment after replacing the UUID):
# ACESTEP_GPU_DEVICE_ID=GPU-<uuid-from-nvidia-smi-L>
# ACESTEP_INIT_LLM=false
# Streamable HTTP MCP endpoint for Claude Code, Codex, and other MCP clients.
# Docker Compose exposes this only on 127.0.0.1 by default.
# MCP_PORT=8002
# When API_PORT or the public API URL changes, set this so MCP returns usable
# audio URLs from the shared Library.
# ACESTEP_MCP_PUBLIC_API_BASE_URL=http://localhost:8001
# Optional bearer token required by MCP clients. This is distinct from
# ACESTEP_API_KEY, which protects the underlying REST API.
# ACESTEP_MCP_API_KEY=replace-with-a-long-random-token
# Add a Tailscale/production hostname here before exposing MCP remotely.
# ACESTEP_MCP_ALLOWED_HOSTS=localhost:*,127.0.0.1:*,[::1]:*,madesk.example.ts.net
# Override only when the local Docker network already routes this range.
# ACESTEP_DOCKER_SUBNET=172.16.43.0/24
# Forge's REST API starts with lazy model initialization by default. Set this
# to false only when you want to download/load models before Forge becomes
# healthy. ACESTEP_INIT_SERVICE below is for the legacy Gradio profile.
# ACESTEP_NO_INIT=true

# ==================== Gradio UI Settings ====================
# Server port (default: 7860)
# PORT=7860

# Server name/host (default: 127.0.0.1 for local only, 0.0.0.0 for network access)
# SERVER_NAME=127.0.0.1

# UI language: en, zh, he, ja (default: en)
# LANGUAGE=en

# Default batch size for generation (1 to GPU-dependent max)
# When not specified, defaults to min(2, GPU_max)
# ACESTEP_BATCH_SIZE=2

# ==================== Startup Settings ====================
# By default models are lazy-loaded on first request (fast server startup).
# Set to false to force eager model loading at startup.
# ACESTEP_NO_INIT=true