附录 C · 多模态 Skill 完整代码示例
本附录包含第十二章（自定义 Skill 开发）中 12.9.2–12.9.6 节多模态 Skill 的完整代码实现。正文中仅保留核心逻辑与架构说明，完整版本请参阅本节。
C.1 PPT 生成 Skill 完整代码
python
# src/ppt_generator.py
from pptx import Presentation
from pptx.util import Inches, Pt
from pptx.dml.color import RgbColor
from pptx.enum.text import PP_ALIGN
class PPTGenerator:
"""PPT 生成器"""
def __init__(self, template_dir: str, style: str = "business"):
self.prs = Presentation()
self.style = self._load_style(style)
def generate(self, content: dict, max_slides: int = 15) -> str:
"""生成 PPT 文件"""
# 1. 创建标题页
self._add_title_slide(
title=content["title"],
subtitle=content.get("subtitle", "")
)
# 2. 创建内容页
for section in content["sections"][:max_slides - 1]:
self._add_content_slide(section)
# 3. 保存
output_path = f"outputs/{content['filename']}.pptx"
self.prs.save(output_path)
return output_path
def _add_title_slide(self, title: str, subtitle: str):
"""添加标题页"""
slide_layout = self.prs.slide_layouts[0] # Title Slide
slide = self.prs.slides.add_slide(slide_layout)
# 设置标题
title_shape = slide.shapes.title
title_shape.text = title
title_para = title_shape.text_frame.paragraphs[0]
title_para.font.size = Pt(44)
title_para.font.bold = True
title_para.font.color.rgb = self.style["primary_color"]
# 设置副标题
if subtitle:
subtitle_shape = slide.placeholders[1]
subtitle_shape.text = subtitle
def _add_content_slide(self, section: dict):
"""添加内容页"""
slide_layout = self.prs.slide_layouts[1] # Title and Content
slide = self.prs.slides.add_slide(slide_layout)
# 标题
title_shape = slide.shapes.title
title_shape.text = section["heading"]
# 内容要点
body_shape = slide.placeholders[1]
tf = body_shape.text_frame
tf.clear()
for i, point in enumerate(section["points"][:5]):
p = tf.add_paragraph()
p.text = f"• {point}"
p.font.size = Pt(18)
p.space_after = Pt(12)
p.level = 0
```python
---
## C.2 播客生成 Skill 完整代码
```python
# src/podcast_generator.py
from TTS.api import TTS
from pydub import AudioSegment
import tempfile
class PodcastGenerator:
"""播客生成器"""
def __init__(self, voice_model: str = "tts_models/en/vctk/vits"):
self.tts = TTS(model_name=voice_model)
async def generate(
self,
content: str,
style: str = "single",
duration_minutes: int = 15
) -> str:
"""生成播客音频"""
# 1. 生成播客脚本
script = self._generate_script(content, style, duration_minutes)
# 2. 分段 TTS
segments = []
for segment in script["segments"]:
audio_path = await self._synthesize_segment(segment)
segments.append(audio_path)
# 3. 合并音频 + 添加背景音乐
final_audio = self._merge_segments(segments)
# 4. 保存
output_path = f"outputs/podcast_{script['title']}.mp3"
final_audio.export(output_path, format="mp3", bitrate="128k")
return output_path
async def _synthesize_segment(self, segment: dict) -> str:
"""合成单个语音片段"""
text = segment["text"]
speaker = segment.get("speaker", "default")
# 使用 TTS 生成
with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as f:
self.tts.tts_to_file(
text=text,
speaker=speaker,
file_path=f.name
)
return f.name
def _merge_segments(self, segment_paths: list) -> AudioSegment:
"""合并音频片段"""
combined = AudioSegment.empty()
for path in segment_paths:
segment = AudioSegment.from_wav(path)
# 添加 0.5s 停顿
combined += segment + AudioSegment.silent(duration=500)
return combinedC.3 图片生成 Skill 完整代码
python
# src/image_generator.py
from openai import AsyncOpenAI
from PIL import Image
import aiohttp
import io
class ImageGenerator:
"""图片生成器"""
def __init__(self, api_key: str, default_model: str = "dall-e-3"):
self.client = AsyncOpenAI(api_key=api_key)
self.default_model = default_model
async def generate(
self,
prompt: str,
style: str = "vivid",
size: str = "1024x1024",
n: int = 1
) -> list[str]:
"""生成图片"""
# 1. 优化提示词
optimized_prompt = self._enhance_prompt(prompt, style)
# 2. 调用 API
response = await self.client.images.generate(
model=self.default_model,
prompt=optimized_prompt,
size=size,
n=n,
quality="standard" if size == "1024x1024" else "hd"
)
# 3. 下载并保存
paths = []
for i, img_data in enumerate(response.data):
image_url = img_data.url
async with aiohttp.ClientSession() as session:
async with session.get(image_url) as resp:
image_bytes = await resp.read()
# 保存
output_path = f"outputs/generated_image_{i}.png"
with open(output_path, "wb") as f:
f.write(image_bytes)
paths.append(output_path)
return paths
def _enhance_prompt(self, prompt: str, style: str) -> str:
"""增强提示词"""
style_prefixes = {
"realistic": "High quality, photorealistic, detailed: ",
"illustration": "Digital illustration, artistic, vibrant colors: ",
"3d": "3D render, octane render, cinematic lighting: ",
"pixel": "Pixel art, retro game style, 16-bit: "
}
prefix = style_prefixes.get(style, "")
return f"{prefix}{prompt}"
```python
---
## C.4 视频生成 Skill 完整代码
```python
# src/video_generator.py
from moviepy.editor import (
ImageClip, TextClip, CompositeVideoClip,
AudioFileClip, concatenate_videoclips
)
from moviepy.video.fx.all import fadein, fadeout
class VideoGenerator:
"""视频生成器"""
def __init__(self, resolution: tuple = (1920, 1080), fps: int = 30):
self.resolution = resolution
self.fps = fps
async def generate(
self,
scenes: list[dict],
audio_path: str = None,
subtitle_style: dict = None
) -> str:
"""生成视频"""
clips = []
for scene in scenes:
# 创建画面
if "image_path" in scene:
clip = ImageClip(scene["image_path"])
else:
# 纯色背景 + 文字
clip = self._create_text_scene(scene)
# 设置时长
clip = clip.set_duration(scene["duration"])
# 添加转场
clip = fadein(clip, 0.5)
clip = fadeout(clip, 0.5)
clips.append(clip)
# 合并
final_video = concatenate_videoclips(clips, method="compose")
# 添加音频
if audio_path:
audio = AudioFileClip(audio_path)
audio = audio.subclip(0, final_video.duration)
final_video = final_video.set_audio(audio)
# 保存
output_path = "outputs/generated_video.mp4"
final_video.write_videofile(
output_path,
fps=self.fps,
codec="libx264",
audio_codec="aac"
)
return output_path
def _create_text_scene(self, scene: dict) -> ImageClip:
"""创建文字场景"""
# 创建背景
bg = ImageClip("templates/blank_bg.png")
bg = bg.resize(self.resolution)
# 添加文字
text = TextClip(
scene["text"],
fontsize=60,
color="white",
font="Arial-Bold",
size=self.resolution,
method="caption"
).set_duration(scene["duration"])
return CompositeVideoClip([bg, text])C.5 看板生成 Skill 完整代码
python
# src/kanban_generator.py
from jinja2 import Template
import json
class KanbanGenerator:
"""看板生成器"""
def __init__(self, template_dir: str = "templates"):
self.template_dir = template_dir
def generate_html(self, tasks: list[dict], columns: list[str] = None) -> str:
"""生成 HTML 看板"""
if columns is None:
columns = ["Backlog", "To Do", "In Progress", "Done"]
# 按状态分组
grouped = {col: [] for col in columns}
for task in tasks:
status = task.get("status", "Backlog")
if status in grouped:
grouped[status].append(task)
# 渲染模板
template_str = self._load_template("kanban.html")
template = Template(template_str)
html = template.render(
columns=columns,
tasks=grouped,
title="Project Kanban"
)
# 保存
output_path = "outputs/kanban.html"
with open(output_path, "w", encoding="utf-8") as f:
f.write(html)
return output_path
def generate_image(self, html_path: str) -> str:
"""将 HTML 看板转为图片"""
from html2image import Html2Image
hti = Html2Image(size=(1920, 1080))
output_path = "outputs/kanban.png"
hti.screenshot(
html_file=html_path,
save_as=output_path
)
return output_path