# mini_bot — optional environment variables.
# Copy to .env (gitignored) and fill in what you need.
# Loaded by scripts/watchdog.sh on bot startup.

# ---------- Azure Cognitive Services Speech (TTS) ----------
# Free tier (F0): 500,000 characters/month neural voices.
# 1. Go to https://portal.azure.com → 创建资源 → 搜索 "Speech" →
#    选 "Speech services" → 价格层 "Free F0" → 创建。
# 2. 创建好后进入资源 → 左侧 "密钥和终结点" → 复制 KEY1 + 位置/区域。
# 3. 区域格式：eastasia, eastus, westus2, southeastasia, japaneast 等。
# 4. 填到下面两行，重启 bot：scripts/stop.sh && nohup scripts/watchdog.sh &
# 5. 进群发 /tts engine 会显示 azure；/tts style 会自动切到 Azure 预设。
AZURE_SPEECH_KEY=
AZURE_SPEECH_REGION=
# Optional: change the default voice (when no per-chat voice is set).
# Popular: zh-CN-XiaoxiaoNeural / zh-CN-YunxiNeural / en-US-JennyNeural
AZURE_SPEECH_VOICE=zh-CN-XiaoxiaoNeural

# ---------- TTS engine override ----------
# If you have BOTH say (macOS) and azure, the bot prefers azure when
# the Azure vars are set. To force "say" anyway, uncomment:
# TTS_ENGINE=say

# ---------- ASR (语音识别 / Speech-to-Text) ----------
# Voice messages are transcribed to text before being sent to the model.
# Without a backend, voice notes can't be understood (the bot will say so).
# Auto-detect priority: openai → azure → whisper-cpp → faster-whisper.
# Force one with ASR_ENGINE (openai|azure|whisper-cpp|faster-whisper).
# ASR_ENGINE=whisper-cpp
#
# Optional language hint (empty = auto-detect; helps short/noisy clips):
#   zh = Chinese, en = English, ja = Japanese, ...
# ASR_LANG=
#
# -- Backend 1: OpenAI Whisper API (cloud, accurate, ~$0.006/min) --
# Reuses OPENAI_API_KEY. Optional OPENAI_BASE_URL for proxies/compatibles.
# OPENAI_API_KEY=
# OPENAI_BASE_URL=https://api.openai.com/v1
# OPENAI_ASR_MODEL=whisper-1          # or gpt-4o-mini-transcribe
#
# -- Backend 2: Azure Speech-to-Text (cloud) --
# Reuses AZURE_SPEECH_KEY + AZURE_SPEECH_REGION above. Best for clips < 60s.
# (No extra vars needed; defaults language to zh-CN when ASR_LANG is empty.)
#
# -- Backend 3: whisper.cpp (local, offline, free) --
#   brew install whisper-cpp        # provides whisper-cli
#   mkdir -p ~/.qoder/models/whisper && cd $_
#   # small (~488MB, fast) or medium (~1.5GB, more accurate for Chinese):
#   curl -LO https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-small.bin
#   curl -LO https://huggingface.co/ggerganov/whisper.cpp/resolve/main/ggml-medium.bin
# WHISPER_CPP_MODEL=/Users/you/.qoder/models/whisper/ggml-medium.bin
# WHISPER_CPP_BEAM_SIZE=8           # default 8; higher = slower but more accurate
# WHISPER_CPP_INITIAL_PROMPT=以下是普通话的句子。  # context hint for language/style
#
# -- Backend 4: faster-whisper (local python, offline, free) --
#   pip install faster-whisper       # needs ctranslate2 wheels for your python
# FASTER_WHISPER_MODEL=base          # tiny|base|small|medium|large-v3

# ---------- Fuyao AI Gateway (私有推理网关) ----------
# Fuyao 模型通过 opencode 运行，API key 配置在 ~/.config/opencode/opencode.json
# 而非此文件。详见 README.md「Fuyao 私有推理网关」章节。
# OPENCODE_BIN=opencode              # 可选：自定义 opencode 路径

# ---------- API Server (OpenAI-compatible 代理模式) ----------
# 启动后 mini_bot 提供一个 OpenAI 兼容的 HTTP API，外部工具（opencode、IDE 等）
# 可直连使用 mini_bot 支持的所有模型，无需走飞书消息通道。
# Fuyao 模型直连网关（支持 streaming），qodercli 模型通过 qodercli -p 桥接。
# BOT_API_SERVER=1
# BOT_API_PORT=9877
# FUYAO_API_KEY=                     # Fuyao 网关 API key
# FUYAO_BASE_URL=https://fuyao-ai-gateway.xiaopeng.link/v1