From 99983004d4cebc19bf98d2e53775b50434c7a50c Mon Sep 17 00:00:00 2001 From: cncoder Date: Sun, 3 May 2026 04:14:36 +0800 Subject: [PATCH] docs(podcast-generator): unify python3 commands + declare python in requires.bins - Commands in README all used python3.12 even in Linux section, but Linux install says pip3.11. Use python3 uniformly with a prerequisites note explaining Mac=3.12 / Linux=3.11. - Add python3 to metadata.openclaw.requires.bins so doctor.py's hard dep is declared. - Replace placeholder with a realpath-based SKILL export (resolves against SKILL.md dir). - Add python3.11-on-Linux troubleshooting row. --- skills/podcast-generator/SKILL.md | 36 +++++++++++++++++++------------ 1 file changed, 22 insertions(+), 14 deletions(-) diff --git a/skills/podcast-generator/SKILL.md b/skills/podcast-generator/SKILL.md index 05ba404..02c656a 100644 --- a/skills/podcast-generator/SKILL.md +++ b/skills/podcast-generator/SKILL.md @@ -5,7 +5,9 @@ metadata: openclaw: emoji: "🎙️" requires: - bins: ["ffprobe", "ffmpeg"] + bins: ["ffprobe", "ffmpeg", "python3"] + # Python 版本要求: Mac=python3.12 (MLX 依赖), Linux=python3.11 (qwen-tts 兼容). + # python3.14 不兼容 (mlx C API / ABI)。doctor.py 会自动检查。 --- # Skill: podcast-generator @@ -30,12 +32,17 @@ metadata: ## 前置安装（只做一次） +> **Python 版本**：Mac 用 `python3.12`（MLX 依赖），Linux 用 `python3.11`（qwen-tts 兼容）。`python3.14` 不兼容。下文命令统一用 `python3`，请先 `alias python3=python3.12`（Mac）或 `alias python3=python3.11`（Linux），或直接把命令里的 `python3` 换成对应版本。 + ### Mac (Apple Silicon) — MLX 后端 ```bash -pip3.12 install mlx mlx-audio numpy soundfile pydub pyloudnorm pyyaml boto3 brew install ffmpeg python@3.12 # 已有可跳过 -export SKILL="/podcast-generator/scripts" +python3.12 -m pip install mlx mlx-audio numpy soundfile pydub pyloudnorm pyyaml boto3 + +# SKILL 路径（相对 SKILL.md 所在目录） +export SKILL="$(cd "$(dirname "$(realpath ~/.agents/skills/podcast-generator/SKILL.md 2>/dev/null || echo .)")" && pwd)/scripts" +# 或手动设定: export SKILL=~/.agents/skills/podcast-generator/scripts ``` ### Linux x86_64 (AL2023 / Ubuntu) — Torch 后端 @@ -48,9 +55,9 @@ sudo dnf install -y https://download1.rpmfusion.org/free/el/rpmfusion-free-relea sudo dnf install -y --allowerasing ffmpeg # Python 包（qwen-tts 自带 torch 2.11 + transformers 4.57.3） -pip3.11 install qwen-tts numpy soundfile pydub pyloudnorm pyyaml boto3 +python3.11 -m pip install qwen-tts numpy soundfile pydub pyloudnorm pyyaml boto3 -export SKILL="/podcast-generator/scripts" +export SKILL=~/.agents/skills/podcast-generator/scripts ``` 首次运行会自动下载 `Qwen/Qwen3-TTS-12Hz-0.6B-Base`（~1.5GB）到 `~/.cache/huggingface/`。 @@ -62,11 +69,11 @@ export SKILL="/podcast-generator/scripts" ## 一键生成（推荐） ```bash -# 0. 环境自检（1 秒） -python3.12 $SKILL/doctor.py +# 0. 环境自检（1 秒，自动识别平台 + Python 版本） +python3 $SKILL/doctor.py # 1. 一键生成 -python3.12 $SKILL/one_shot.py "AI 能否取代程序员" --duration 5min --style debate +python3 $SKILL/one_shot.py "AI 能否取代程序员" --duration 5min --style debate # 2. 打开 HTML 播放 open output/*/*/index.html @@ -182,7 +189,7 @@ VOICE_PAN = { 如果你不走 `one_shot.py`，想单独把 MP3 + 脚本拼成播放器： ```bash -python3.12 $SKILL/generate_player_html.py /tmp/demo.mp3 /tmp/demo.txt /tmp/demo.html \ +python3 $SKILL/generate_player_html.py /tmp/demo.mp3 /tmp/demo.txt /tmp/demo.html \ --title "AI 能不能取代程序员" ``` @@ -202,15 +209,15 @@ python3.12 $SKILL/generate_player_html.py /tmp/demo.mp3 /tmp/demo.txt /tmp/demo. ```bash # 1. 检查 AWS 凭证 -python3.12 $SKILL/publish_to_cdn.py check +python3 $SKILL/publish_to_cdn.py check # 2. 起全球唯一 bucket 名 BUCKET="my-podcast-cdn-$(whoami)-$(date +%s)" -python3.12 $SKILL/publish_to_cdn.py provision --bucket "$BUCKET" --region us-east-1 +python3 $SKILL/publish_to_cdn.py provision --bucket "$BUCKET" --region us-east-1 # 配置自动缓存到 ~/.podcast-generator/publish.json # 3. 等 CloudFront Deployed 后发布 -python3.12 $SKILL/publish_to_cdn.py publish \ +python3 $SKILL/publish_to_cdn.py publish \ --mp3 podcast.mp3 --html index.html --slug ai-vs-programmers ``` @@ -240,8 +247,9 @@ python3.12 $SKILL/publish_to_cdn.py publish \ | 症状 | 修复 | |------|------| | TTS 超时 | 重新运行，断点续传自动恢复（复用 `_tts_chunks/`） | -| 音频质量差 | 确认无并行 TTS（`ps aux \| grep python3.12`） | -| python3.12 找不到 | `brew install python@3.12` | +| 音频质量差 | 确认无并行 TTS（`ps aux \| grep python3`） | +| python3.12 找不到 (Mac) | `brew install python@3.12` | +| python3.11 找不到 (Linux) | `sudo dnf install python3.11` (AL2023) 或 `apt install python3.11` (Ubuntu) | | ffmpeg 找不到 | `brew install ffmpeg` | | MP3 太大发不出去 | `ffmpeg -i podcast.mp3 -b:a 64k compressed.mp3` |