初步实现 LLM调用生成海报文案
This commit is contained in:
parent
45ecb5cb4b
commit
f38e88d04b
1
.env
Normal file
1
.env
Normal file
@ -0,0 +1 @@
|
||||
MOONSHOT_API_KEY=sk-Xzv3OqwLoM1RRsXIQuEhLMNATrMv2JsLwf7LbM97Ha6sFDun
|
42
README.md
42
README.md
@ -20,4 +20,44 @@
|
||||
python generate_layout.py
|
||||
```
|
||||
- 该脚本会自动调用 DeepSeek API 生成 React 组件代码,并将其保存到指定路径。
|
||||
默认为``output/generated_code.jsx``,可以根据需要修改。
|
||||
默认为``output/generated_code.jsx``,可以根据需要修改。
|
||||
|
||||
## 图层叠加和PSD导出部分
|
||||
### 功能组件
|
||||
- **图层合成**:根据生成的布局配置,将各个图层叠加在一起,生成完整的海报图像。
|
||||
- **PSD导出**:将合成的海报图像保存为标准的PSD格式,同时保留各个图层的独立性,方便后续编辑。
|
||||
### 如何调用
|
||||
1. **环境准备**:
|
||||
```bash
|
||||
conda create -n ai_service python=3.8
|
||||
conda activate ai_service
|
||||
pip install -r requirements.txt
|
||||
```
|
||||
2. **运行脚本**:
|
||||
```bash
|
||||
python export_psd.py #简单生成
|
||||
python export_psd_from_json.py #从json文件加载布局并生成
|
||||
python PSD_test.py #在run_pipline.py中的调用方法
|
||||
```
|
||||
该脚本会自动调用图层合成模块生成完整海报图像,并将其保存为标准PSD格式。
|
||||
|
||||
## 文案生成
|
||||
### 功能组件
|
||||
使用 kimi API 提供的 LLM(大型语言模型)功能,用于生成海报文案设计。当前实现的功能包括:
|
||||
|
||||
- **标题生成**:根据海报主题生成适合的标题。
|
||||
- **副标题生成**:提供与主题相关的副标题或说明文案。
|
||||
- **字体与颜色建议**:为生成的文案提供适合的字体和颜色建议。
|
||||
|
||||
### 如何调用
|
||||
1. **环境准备**:
|
||||
- 确保已安装 Python 环境和必要依赖(`openai`、`python-dotenv`等)。
|
||||
- 在项目根目录的 `.env` 文件中配置 `MOONSHOT_API_KEY`。
|
||||
|
||||
2. **运行脚本**:
|
||||
- 直接运行 `generate_text.py` 脚本:
|
||||
```bash
|
||||
python generate_text.py
|
||||
```
|
||||
- 该脚本会自动调用 kimi API 生成适合海报主题的标题、副标题和 Logo 文案,并提供字体和颜色建议,并将其保存到指定路径。
|
||||
默认为``output/poster_content.json``,可以根据需要修改。
|
83
configs/font.yaml
Normal file
83
configs/font.yaml
Normal file
@ -0,0 +1,83 @@
|
||||
# 颜色常量
|
||||
NAMING_COLORS:
|
||||
NANKAI_PURPLE: "#7E0C6E"
|
||||
LOGO_WHITE: "#FFFFFF"
|
||||
DARK_TEXT_ON_LIGHT_BG: "#000000" # 示例深色文字
|
||||
LIGHT_TEXT_ON_DARK_BG: "#f4efd9" # 示例浅色文字
|
||||
|
||||
# 可用字体列表
|
||||
# name: 字体文件在系统中的名称或前端能识别的名称
|
||||
# tags: 用于规则匹配的标签
|
||||
|
||||
available_fonts:
|
||||
- name: "FZLanTingHei-ExtraBold-GB" # 方正兰亭特黑简体 (假设这是实际的font-family名)
|
||||
displayName: "方正兰亭特黑简体"
|
||||
tags: ["现代", "力量感", "标题", "正式", "无衬线", "醒目"]
|
||||
roles: ["title"]
|
||||
|
||||
- name: "Adobe Song Std L" # adobe宋体 (L 通常表示 Light weight)
|
||||
displayName: "Adobe 宋体 Std L"
|
||||
tags: ["传统", "正文", "经典", "宋体"]
|
||||
roles: ["subtitle", "content"]
|
||||
|
||||
- name: "SimHei" # 黑体 简
|
||||
displayName: "黑体 (简体)"
|
||||
tags: ["通用", "简约", "正文", "无衬线", "清晰"]
|
||||
roles: ["title", "subtitle", "content"] # 黑体用途广泛
|
||||
|
||||
- name: "Hiragino Sans GB W3" # 冬青黑体 (W3 通常表示字重)
|
||||
displayName: "冬青黑体 W3"
|
||||
tags: ["现代", "清晰", "正文", "优雅", "无衬线"]
|
||||
roles: ["subtitle", "content"]
|
||||
|
||||
# 核心样式规则
|
||||
# 主题 -> 图层ID -> 样式定义
|
||||
STYLE_RULES:
|
||||
"世界读书日": # 对应海报主题
|
||||
"layer6_title": # 对应文字图层6:大标题
|
||||
font_prefs: # 字体选择偏好
|
||||
- name: "ImpactfulModernSans" # 首选这个名字的字体
|
||||
- tags: ["title", "bold", "modern_sans"] # 如果首选找不到,则找带这些标签的
|
||||
font_size_category: "largest"
|
||||
color_on_light_bg: DARK_TEXT_ON_LIGHT_BG # 使用上面定义的颜色名
|
||||
color_on_dark_bg: LIGHT_TEXT_ON_DARK_BG
|
||||
position_hint: "图片上方、居中、不遮挡图像"
|
||||
"layer7_subtitle": # 对应文字图层7:小字/文案
|
||||
font_prefs:
|
||||
- name: "ReadableBodySans"
|
||||
- tags: ["body", "readable_sans"]
|
||||
font_size_category: "small_below_title"
|
||||
color_on_light_bg: DARK_TEXT_ON_LIGHT_BG
|
||||
color_on_dark_bg: LIGHT_TEXT_ON_DARK_BG
|
||||
position_hint: "标题下方"
|
||||
|
||||
"南开校庆":
|
||||
"layer6_title":
|
||||
font_prefs:
|
||||
- name: "南开官方标准黑体"
|
||||
- tags: ["formal", "title"]
|
||||
font_size_category: "largest"
|
||||
color_on_light_bg: NANKAI_PURPLE
|
||||
color_on_dark_bg: LOGO_WHITE
|
||||
position_hint: "图片上方、居中、庄重"
|
||||
|
||||
"通用默认": # 如果主题未匹配,则使用这里的规则
|
||||
"default_title": # 给一个默认的标题图层ID
|
||||
font_prefs:
|
||||
- tags: ["title", "bold"]
|
||||
font_size_category: "largest"
|
||||
color_on_light_bg: "#1A1A1A"
|
||||
color_on_dark_bg: "#E5E5E5"
|
||||
"default_body":
|
||||
font_prefs:
|
||||
- tags: ["body", "readable_sans"]
|
||||
font_size_category: "normal"
|
||||
color_on_light_bg: "#000000"
|
||||
color_on_dark_bg: "#ffffff"
|
||||
|
||||
# Logo的颜色规则 (独立于普通文字图层,像是对图像素材的选择)
|
||||
LOGO_RULES:
|
||||
"layer5_nankai_logo":
|
||||
description: "南开大学官方logo,位于四角,预留空隙"
|
||||
color_on_light_bg: NANKAI_PURPLE
|
||||
color_on_dark_bg: LOGO_WHITE
|
27
nohup.out
27
nohup.out
@ -1,27 +0,0 @@
|
||||
python3: can't open file '/var/www/ai_service/your_server.py': [Errno 2] No such file or directory
|
||||
python3: can't open file '/var/www/ai_service/aihelp.py': [Errno 2] No such file or directory
|
||||
2025-04-10 10:12:52,702 - INFO - Starting server on port 9000
|
||||
2025-04-10 10:13:17,949 - ERROR - Port 9000 is already in use
|
||||
Traceback (most recent call last):
|
||||
File "/var/www/ai_service/apihelp.py", line 4, in <module>
|
||||
Traceback (most recent call last):
|
||||
File "/var/www/ai_service/apihelp.py", line 4, in <module>
|
||||
from c import HTTPServer, BaseHTTPRequestHandler
|
||||
from c import HTTPServer, BaseHTTPRequestHandler
|
||||
ModuleNotFoundError: No module named 'c'
|
||||
ModuleNotFoundError: No module named 'c'
|
||||
Error in sys.excepthook:
|
||||
Traceback (most recent call last):
|
||||
File "/usr/lib/python3/dist-packages/apport_python_hook.py", line 228, in partial_apport_excepthook
|
||||
return apport_excepthook(binary, exc_type, exc_obj, exc_tb)
|
||||
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
File "/usr/lib/python3/dist-packages/apport_python_hook.py", line 144, in apport_excepthook
|
||||
os.open(pr_filename, os.O_WRONLY | os.O_CREAT | os.O_EXCL, 0o640), "wb"
|
||||
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
||||
FileExistsError: [Errno 17] File exists: '/var/crash/_var_www_ai_service_apihelp.py.0.crash'
|
||||
|
||||
Original exception was:
|
||||
Traceback (most recent call last):
|
||||
File "/var/www/ai_service/apihelp.py", line 4, in <module>
|
||||
from c import HTTPServer, BaseHTTPRequestHandler
|
||||
ModuleNotFoundError: No module named 'c'
|
@ -1,51 +0,0 @@
|
||||
import React from 'react';
|
||||
|
||||
const DragonBoatFestivalPoster = () => {
|
||||
return (
|
||||
<div style={{ position: 'relative', width: '1080px', height: '1920px' }}>
|
||||
{/* 背景图层 */}
|
||||
<div style={{ position: 'absolute', top: 0, left: 0, width: '100%', height: '100%' }}></div>
|
||||
|
||||
{/* 主体图层 */}
|
||||
<div style={{ position: 'absolute', top: '25%', left: '50%', transform: 'translate(-50%, -50%)' }}>
|
||||
<div></div>
|
||||
<div></div>
|
||||
</div>
|
||||
|
||||
{/* 活动亮点 */}
|
||||
<div style={{
|
||||
position: 'absolute',
|
||||
bottom: '25%',
|
||||
left: '50%',
|
||||
transform: 'translate(-50%, 0)',
|
||||
display: 'grid',
|
||||
gridTemplateColumns: 'repeat(3, 1fr)',
|
||||
gap: '20px'
|
||||
}}>
|
||||
<div>
|
||||
<div></div>
|
||||
<div></div>
|
||||
<div></div>
|
||||
</div>
|
||||
<div>
|
||||
<div></div>
|
||||
<div></div>
|
||||
<div></div>
|
||||
</div>
|
||||
<div>
|
||||
<div></div>
|
||||
<div></div>
|
||||
<div></div>
|
||||
</div>
|
||||
</div>
|
||||
|
||||
{/* 页脚 */}
|
||||
<div style={{ position: 'absolute', bottom: 0, left: '50%', transform: 'translateX(-50%)' }}>
|
||||
<div></div>
|
||||
<div></div>
|
||||
</div>
|
||||
</div>
|
||||
);
|
||||
};
|
||||
|
||||
export default DragonBoatFestivalPoster;
|
16
outputs/poster_content.json
Normal file
16
outputs/poster_content.json
Normal file
@ -0,0 +1,16 @@
|
||||
{
|
||||
"layer5_logo_content": {
|
||||
"text": "",
|
||||
"color": "#000000"
|
||||
},
|
||||
"layer6_title_content": {
|
||||
"content": "世界读书日",
|
||||
"font_name": "FZLanTingHei-ExtraBold-GB",
|
||||
"color": "#000000"
|
||||
},
|
||||
"layer7_subtitle_content": {
|
||||
"content": "阅古今中外,启智慧之门",
|
||||
"font_name": "Adobe Song Std L",
|
||||
"color": "#4F4F4F"
|
||||
}
|
||||
}
|
@ -1,140 +0,0 @@
|
||||
import os
|
||||
from openai import OpenAI
|
||||
from dotenv import load_dotenv
|
||||
import time
|
||||
import logging
|
||||
|
||||
|
||||
# === Config LLM call ===
|
||||
load_dotenv()
|
||||
deepseek_url = 'https://api.deepseek.com/v1' # set to be compatible with the OpenAI API
|
||||
deepseek_api = os.getenv("DEEPSEEK_API_KEY")
|
||||
if not deepseek_api:
|
||||
raise ValueError("DEEPSEEK_API_KEY not set!")
|
||||
|
||||
# 配置日志
|
||||
logging.basicConfig(level=logging.INFO)
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
# === prompts and parameters ===
|
||||
def call_deepseek(
|
||||
messages=None,
|
||||
system_prompt="你是一个擅长前端开发的AI,专注于生成Vue.js代码。",
|
||||
prompt=None,
|
||||
model='deepseek-chat',
|
||||
temperature=0.6,
|
||||
max_tokens=1000,
|
||||
stream=False,
|
||||
max_retries=3,
|
||||
):
|
||||
"""
|
||||
调用 DeepSeek API,支持多轮对话和流式/非流式响应
|
||||
|
||||
Args:
|
||||
messages (list): 对话消息列表,格式为[{"role": "user", "content": "..."},...]。
|
||||
如果提供,则优先使用,否则根据 system_prompt和 prompt 构造.
|
||||
system_prompt (str): 系统提示词,仅在 message 未提供时使用.
|
||||
prompt (str): 用户提示词,仅在 message 未提供时使用.
|
||||
model (str): 模型名称,默认为 'deepseek-chat'。
|
||||
temperature (float): 温度参数,控制生成文本的随机性,默认为 0.6。
|
||||
max_tokens (int): 最大生成 token 数量,默认为 1000。
|
||||
stream (bool): 是否使用流式响应,默认为 False。
|
||||
max_retries (int): 最大重试次数,默认为 3。
|
||||
|
||||
Returns:
|
||||
tuple: (result, usage)
|
||||
- result (str): 非流式返回字符串(回复内容),流式返回生成器(逐块内容).
|
||||
- usage (dict): token使用统计(非流式返回 Usage 对象,流式返回 None).
|
||||
|
||||
Raises:
|
||||
ValueError: 如果 message 和 prompt 都为空或参数无效
|
||||
Exception: 如果 API 调用失败且达到最大重试次数
|
||||
"""
|
||||
# 初始化 OpenAI 客户端
|
||||
client = OpenAI(
|
||||
api_key=deepseek_api,
|
||||
base_url=deepseek_url,
|
||||
)
|
||||
|
||||
# 参数验证
|
||||
if messages is None:
|
||||
if not prompt:
|
||||
raise ValueError("必须提供 message 或 prompt 参数")
|
||||
messages = [
|
||||
{"role": "system", "content": system_prompt},
|
||||
{"role": "user", "content": prompt}
|
||||
]
|
||||
elif not isinstance(messages, list) or not messages:
|
||||
raise ValueError("message 参数必须是非空列表")
|
||||
|
||||
# 模型验证
|
||||
models = ["deepseek-chat", "deepseek-reasoner"]
|
||||
if model not in models:
|
||||
raise ValueError(f"无效的模型名称: {model},可用模型: {models}")
|
||||
|
||||
# 调用 API
|
||||
for attempt in range(max_retries):
|
||||
try:
|
||||
response = client.chat.completions.create(
|
||||
model=model,
|
||||
messages=messages,
|
||||
temperature=temperature,
|
||||
max_tokens=max_tokens,
|
||||
stream=stream
|
||||
)
|
||||
|
||||
# 流式响应
|
||||
if stream:
|
||||
def stream_generator():
|
||||
usage = None
|
||||
for chunk in response:
|
||||
if chunk.choices[0].delta.content is not None:
|
||||
yield chunk.choices[0].delta.content
|
||||
return usage
|
||||
return stream_generator(), None
|
||||
else:
|
||||
# 非流式响应
|
||||
content = response.choices[0].message.content
|
||||
return content, response.usage
|
||||
|
||||
except Exception as e:
|
||||
if hasattr(e, 'status_code') and e.status_code == 429: # 限流
|
||||
logger.warning(f"请求过于频繁,正在重试... (尝试 {attempt + 1}/{max_retries})")
|
||||
time.sleep(2 ** attempt) # 指数退避
|
||||
else:
|
||||
logger.error(f"API 调用失败:{str(e)}")
|
||||
raise
|
||||
raise Exception("达到最大重试次数,API 调用失败")
|
||||
|
||||
def generate_vue_code(prompt=None):
|
||||
prompt = (
|
||||
"生成一个React组件代码,用于端午节活动海报,包含以下部分并指定排版位置:"
|
||||
"1. 背景图层:div,占据整个组件区域。"
|
||||
"2. 主体图层:div,位于顶部1/4处,居中,包含标题和副标题。"
|
||||
"3. 活动亮点:div,位于底部1/4处,居中,使用网格布局展示三项活动(每项包含图标、标题和描述)。"
|
||||
"4. 页脚:div,位于底部,居中,包含主办单位信息和logo图片。"
|
||||
"组件尺寸为1080x1920px,布局使用absolute定位,仅关注排版位置,不包含任何样式描述(如颜色、字体、阴影、动画等)。"
|
||||
"仅生成React代码本身,不包含说明性文字、注释或Markdown格式。"
|
||||
)
|
||||
system_prompt = (
|
||||
"你是一个擅长前端开发的AI,专注于生成React.js代码。"
|
||||
"生成的代码仅关注排版位置,使用absolute定位,不包含任何样式描述(如颜色、字体、阴影、动画等)。"
|
||||
"确保代码符合React最佳实践,仅生成代码本身。"
|
||||
)
|
||||
|
||||
|
||||
result,usage = call_deepseek(prompt=prompt, system_prompt=system_prompt,temperature=0.4)
|
||||
# print(result)
|
||||
# print(usage)
|
||||
return result
|
||||
|
||||
def save_code(code,file_path="../outputs/generated_code.jsx"):
|
||||
os.makedirs(os.path.dirname(file_path), exist_ok=True)
|
||||
with open(file_path, "w", encoding="utf-8") as f:
|
||||
f.write(code)
|
||||
|
||||
if __name__ == "__main__":
|
||||
vue_code = generate_vue_code()
|
||||
save_code(vue_code)
|
||||
print("React组件代码已生成并保存到outputs/generated_code.jsx")
|
272
scripts/generate_text.py
Normal file
272
scripts/generate_text.py
Normal file
@ -0,0 +1,272 @@
|
||||
# -*- coding: utf-8 -*-
|
||||
import os
|
||||
from openai import OpenAI
|
||||
from dotenv import load_dotenv
|
||||
import json
|
||||
import yaml
|
||||
|
||||
# 加载 .env 文件
|
||||
load_dotenv()
|
||||
|
||||
# 从环境变量中获取 API Key
|
||||
MOONSHOT_API_KEY = os.getenv("MOONSHOT_API_KEY")
|
||||
|
||||
# --- 可配置的常量 ---
|
||||
DEFAULT_LOGO_TEXT = ""
|
||||
AVAILABLE_FONTS = []
|
||||
NAMING_COLORS = {}
|
||||
STYLE_RULES = {}
|
||||
LOGO_RULES = {}
|
||||
|
||||
def load_config_from_file(file_path):
|
||||
"""从 YAML 文件加载配置"""
|
||||
global DEFAULT_LOGO_TEXT, AVAILABLE_FONTS, NAMING_COLORS, STYLE_RULES, LOGO_RULES
|
||||
try:
|
||||
with open(file_path, 'r', encoding='utf-8') as file:
|
||||
config_data = yaml.safe_load(file)
|
||||
|
||||
DEFAULT_LOGO_TEXT = config_data.get("default_logo_text", DEFAULT_LOGO_TEXT)
|
||||
AVAILABLE_FONTS = config_data.get("available_fonts", [])
|
||||
NAMING_COLORS = config_data.get("NAMING_COLORS", {})
|
||||
STYLE_RULES = config_data.get("STYLE_RULES", {})
|
||||
LOGO_RULES = config_data.get("LOGO_RULES", {})
|
||||
|
||||
if not AVAILABLE_FONTS:
|
||||
print("警告:未能从 YAML 配置中加载可用字体列表,或列表为空。")
|
||||
AVAILABLE_FONTS = [{"name": "SimHei", "displayName": "黑体 (简体)", "tags": ["通用"], "roles": ["title", "subtitle", "content"]}]
|
||||
|
||||
print("YAML 配置文件加载成功。")
|
||||
print(f"默认 Logo 文字: {DEFAULT_LOGO_TEXT if DEFAULT_LOGO_TEXT else '未设置'}")
|
||||
except yaml.YAMLError as e:
|
||||
print(f"解析 YAML 文件时发生错误: {e}")
|
||||
AVAILABLE_FONTS = [{"name": "SimHei", "displayName": "黑体 (简体)", "tags": ["通用"], "roles": ["title", "subtitle", "content"]}]
|
||||
print("已使用内部备用字体列表。")
|
||||
except FileNotFoundError:
|
||||
print(f"错误:未找到文件 {file_path}")
|
||||
AVAILABLE_FONTS = [{"name": "SimHei", "displayName": "黑体 (简体)", "tags": ["通用"], "roles": ["title", "subtitle", "content"]}]
|
||||
print("已使用内部备用字体列表。")
|
||||
|
||||
if not MOONSHOT_API_KEY:
|
||||
print("错误:未能从环境变量中获取 MOONSHOT_API_KEY,请检查 .env 文件。")
|
||||
exit()
|
||||
|
||||
# 初始化 OpenAI 客户端
|
||||
try:
|
||||
client = OpenAI(api_key=MOONSHOT_API_KEY, base_url="https://api.moonshot.cn/v1")
|
||||
print("Kimi客户端初始化成功。")
|
||||
except Exception as e:
|
||||
print(f"初始化OpenAI客户端失败: {e}")
|
||||
exit()
|
||||
|
||||
def extract_parameters_from_input(user_input):
|
||||
"""
|
||||
使用模型从用户输入中提取主题、风格、元素、背景颜色和自定义Logo文字等信息。
|
||||
"""
|
||||
extraction_prompt = f"""
|
||||
你是一个专业的自然语言理解助手,请从以下用户输入中提取海报设计的关键信息,并严格按照JSON格式返回:
|
||||
用户输入:{user_input}
|
||||
输出格式:
|
||||
{{
|
||||
"poster_theme": "这里是海报的主题,例如:世界读书日、南开大学校庆",
|
||||
"style_desc": "这里是海报的风格描述,例如:现代、简约、有文化气息",
|
||||
"elements_to_include": ["这里是需要包含的元素,例如:南开大学教学楼", "书本", "阅读的人"],
|
||||
"background_is_light": true,
|
||||
"custom_logo_text": "这里是用户在输入中明确指定的、用作Logo的文字内容。例如,如果用户说“Logo文字用‘技术创新研讨会’”,则提取“技术创新研讨会”。如果用户没有明确指定Logo文字,则返回null或空字符串。"
|
||||
}}
|
||||
注意:对于 "background_is_light",如果用户输入明确提及背景色深浅,则据此判断;若未提及或无法判断,默认为true (浅色)。
|
||||
对于 "custom_logo_text",仅当用户明确指定logo文字时才提取,否则应为null或空字符串。
|
||||
"""
|
||||
|
||||
try:
|
||||
completion = client.chat.completions.create(
|
||||
model="moonshot-v1-8k",
|
||||
messages=[
|
||||
{"role": "system", "content": "你是一个专业的自然语言理解助手,专门负责从用户输入中提取海报设计的关键参数,并严格以JSON格式输出。"},
|
||||
{"role": "user", "content": extraction_prompt}
|
||||
],
|
||||
temperature=0.2
|
||||
)
|
||||
response_content = completion.choices[0].message.content
|
||||
print(f"--- 模型返回的参数提取结果 ---\n{response_content}\n--------------------------\n")
|
||||
|
||||
json_str = response_content.strip()
|
||||
if "```json" in json_str:
|
||||
json_str = json_str.split("```json")[1].split("```")[0].strip()
|
||||
elif json_str.startswith("```") and json_str.endswith("```"):
|
||||
json_str = json_str[3:-3].strip()
|
||||
|
||||
extracted_params = json.loads(json_str)
|
||||
|
||||
# 提供默认值
|
||||
if "background_is_light" not in extracted_params:
|
||||
extracted_params["background_is_light"] = True
|
||||
if "custom_logo_text" not in extracted_params:
|
||||
extracted_params["custom_logo_text"] = None
|
||||
|
||||
return extracted_params
|
||||
except Exception as e:
|
||||
print(f"调用模型提取参数或解析JSON时发生错误: {e}")
|
||||
return {
|
||||
"poster_theme": "默认主题",
|
||||
"style_desc": "通用风格",
|
||||
"elements_to_include": [],
|
||||
"background_is_light": True,
|
||||
"custom_logo_text": None,
|
||||
"error": f"参数提取失败: {e}"
|
||||
}
|
||||
|
||||
def generate_text_content_for_layers(poster_theme, style_desc, background_is_light, logo_text_to_display, elements_to_include=None):
|
||||
"""
|
||||
调用 Kimi API 生成文案内容和设计建议(具体字体名称、颜色),包括指定的Logo文字。
|
||||
AI 将从 AVAILABLE_FONTS (来自YAML) 中选择字体。
|
||||
"""
|
||||
global AVAILABLE_FONTS # 确保能访问到从YAML加载的字体列表
|
||||
|
||||
if elements_to_include is None: elements_to_include = []
|
||||
if not client: return {"error": "Kimi客户端未初始化."}
|
||||
|
||||
background_color_desc = '浅色' if background_is_light else '深色'
|
||||
|
||||
# 构建字体列表供AI参考
|
||||
font_list_for_prompt = "当前可用的字体列表如下 (请从中选择 'name' 作为 font_name 的值):\n"
|
||||
if AVAILABLE_FONTS:
|
||||
for font in AVAILABLE_FONTS:
|
||||
font_list_for_prompt += (
|
||||
f"- 名称(name): '{font['name']}' (显示名: {font.get('displayName', font['name'])}) - "
|
||||
f"风格标签: [{', '.join(font.get('tags', []))}] - "
|
||||
f"建议角色: [{', '.join(font.get('roles', []))}]\n"
|
||||
)
|
||||
else:
|
||||
font_list_for_prompt += "- 名称(name): 'SimHei' (显示名: 黑体) - 风格标签: [通用] - 建议角色: [标题, 内容]\n"
|
||||
font_list_for_prompt += "注意:可用字体列表似乎未正确加载,请基于通用字体(如SimHei)给出建议。\n"
|
||||
|
||||
|
||||
# 提取背景颜色描述到变量中
|
||||
background_color_desc = '浅色' if background_is_light else '深色'
|
||||
|
||||
prompt_parts = [
|
||||
f"你是一个专业的海报文案及设计元素建议助手,请为关于“{poster_theme}”的主题海报创作宣传文案和设计建议。",
|
||||
f"海报整体风格要求:{style_desc}。",
|
||||
f"设计背景是{background_color_desc}。",
|
||||
f"Logo文字已确定为:“{logo_text_to_display}”。",
|
||||
"--------------------------------------------------",
|
||||
"可用字体参考(请为下面的 'font_name' 字段从以下列表的 '名称(name)' 中选择一个最合适的):",
|
||||
font_list_for_prompt, # 注入格式化后的字体列表
|
||||
"--------------------------------------------------",
|
||||
"请严格按照以下JSON格式返回结果,仅包含针对指定图层的文本内容、从上述列表中选择的具体字体名称和颜色。所有建议需与海报主题、风格及背景色协调:",
|
||||
"{",
|
||||
f' "layer5_logo_content": {{ "text": "{logo_text_to_display}", "color": "这里是为Logo文字“{logo_text_to_display}”在背景色 ({background_color_desc}) 上推荐的颜色。例如:浅色背景可使用主题色如南开紫(#7E0C6E)或黑色(#000000),深色背景可使用白色(#FFFFFF)。请提供符合要求的具体十六进制颜色值。" }},',
|
||||
f' "layer6_title_content": {{ "content": "这里是为“{poster_theme}”生成的主标题。如果用户输入中明确指定了主题(例如:世界读书日、端午、秋分),则标题必须直接使用该主题,且长度严格控制在2到8个汉字之间。", "font_name": "从上面提供的可用字体列表中为标题选择的具体字体\'名称(name)\' (例如: SimHei, FZLanTingHei-ExtraBold-GB)。", "color": "这里是建议的标题文字颜色,需确保在背景上高对比度且与整体风格和谐,可以根据背景选择白色(#FFFFFF)、米色(#f4efd9)、黑色(#000000)等。请提供具体十六进制颜色值。" }},',
|
||||
f''' "layer7_subtitle_content": {{ "content": "这里是为“{poster_theme}”生成的副标题或说明文案。请根据主题选择以下一种形式并创作,确保内容富有文化内涵或情感共鸣,避免空洞口号:\n 1. 点题短语或对偶句(总计10-25字)。\n 2. 精美描述性或介绍性文字(一段话,20-35字)。\n 3. 若主题适合诗歌,可为原创短诗(例如绝句的前两句或类似形式,总计10-20字)。", "font_name": "从上面提供的可用字体列表中为副标题选择的具体字体\'名称(name)\',应与主标题协调。", "color": "这里是建议的副标题文字颜色,确保清晰可读,并与整体色调和谐,避免过于鲜艳的颜色。请提供具体十六进制颜色值。" }}''',
|
||||
"}",
|
||||
"重要提示:",
|
||||
"1. 为 'content' 字段生成的文本必须是直接的文字内容,不应包含诸如“这里是...”之类的描述性前缀。",
|
||||
"2. 'font_name' 字段的值必须是从上面提供的可用字体列表中的一个确切的 '名称(name)'。",
|
||||
"3. 'color' 字段应提供明确的十六进制颜色代码(例如“#FFFFFF”)。",
|
||||
"4. 确保最终输出是完整且严格符合上述格式的JSON对象,不要在JSON内容之外添加任何解释、注释、Markdown标记(如 ```json ```)或其他任何非JSON字符。"
|
||||
]
|
||||
user_prompt = "\n".join(prompt_parts)
|
||||
system_prompt_content = "你是一个专业的海报文案及设计元素建议助手。请严格按照用户要求的JSON格式输出,确保内容符合设计原则、主题文化内涵,并直接输出JSON内容,不要包含任何修饰性或解释性的文字,例如 '好的,这是您要求的JSON:' 或 '```json' 等标记。"
|
||||
|
||||
try:
|
||||
completion = client.chat.completions.create(
|
||||
model="moonshot-v1-8k",
|
||||
messages=[
|
||||
{"role": "system", "content": system_prompt_content},
|
||||
{"role": "user", "content": user_prompt}
|
||||
],
|
||||
temperature=0.4
|
||||
)
|
||||
response_content = completion.choices[0].message.content
|
||||
|
||||
json_str = response_content.strip()
|
||||
if "```json" in json_str:
|
||||
json_str = json_str.split("```json")[1].split("```")[0].strip()
|
||||
elif json_str.startswith("```") and json_str.endswith("```"):
|
||||
json_str = json_str[3:-3].strip()
|
||||
|
||||
parsed_json = json.loads(json_str)
|
||||
return parsed_json
|
||||
except Exception as e:
|
||||
print(f"调用Kimi或解析JSON时发生错误: {e}\n原始回复内容: {response_content if 'response_content' in locals() else 'N/A'}")
|
||||
return {"error": f"LLM调用或解析失败: {e}"}
|
||||
|
||||
def get_poster_content_suggestions(user_input_string: str):
|
||||
"""
|
||||
主接口函数:接收用户输入,提取参数,确定Logo文字,并生成海报图层内容建议。
|
||||
返回包含建议内容的字典,或包含错误信息的字典。
|
||||
"""
|
||||
print(f"--- 开始处理用户输入: {user_input_string} ---")
|
||||
# 1. 提取参数
|
||||
extracted_params = extract_parameters_from_input(user_input_string)
|
||||
|
||||
if "error" in extracted_params:
|
||||
print(f"参数提取过程中发生错误: {extracted_params['error']}")
|
||||
return extracted_params # 返回包含错误信息的字典
|
||||
|
||||
poster_theme = extracted_params.get("poster_theme", "未知主题")
|
||||
style_desc = extracted_params.get("style_desc", "默认风格")
|
||||
elements_to_include = extracted_params.get("elements_to_include", [])
|
||||
background_is_light = extracted_params.get("background_is_light", True)
|
||||
custom_logo_text = extracted_params.get("custom_logo_text")
|
||||
|
||||
print(f"\n--- 提取到的参数用于生成文案 ---")
|
||||
print(f"主题: {poster_theme}")
|
||||
print(f"风格: {style_desc}")
|
||||
print(f"背景是否浅色: {background_is_light}")
|
||||
print(f"用户指定Logo文字: {custom_logo_text if custom_logo_text else '未指定'}")
|
||||
|
||||
# 2. Logo文字选择逻辑
|
||||
final_logo_text = DEFAULT_LOGO_TEXT # 默认为 ""
|
||||
if custom_logo_text and custom_logo_text.strip(): # 如果用户指定了且不为空
|
||||
final_logo_text = custom_logo_text.strip()
|
||||
print(f"--- 使用用户自定义Logo文字: “{final_logo_text}” ---")
|
||||
else:
|
||||
print(f"--- 使用默认Logo文字: “{final_logo_text}” ---")
|
||||
print("--------------------------------\n")
|
||||
|
||||
# 3. 调用生成函数
|
||||
generated_content = generate_text_content_for_layers(
|
||||
poster_theme,
|
||||
style_desc,
|
||||
background_is_light,
|
||||
final_logo_text, # 传递最终确定的Logo文字
|
||||
elements_to_include
|
||||
)
|
||||
|
||||
if "error" in generated_content:
|
||||
print(f"内容生成过程中发生错误: {generated_content['error']}")
|
||||
else:
|
||||
print("\n--- 成功生成的文案及设计建议 ---")
|
||||
# print(json.dumps(generated_content, indent=2, ensure_ascii=False))
|
||||
|
||||
return generated_content
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
# 用户自由输入示例
|
||||
# user_input_example1 = "帮我设计一个关于'南开百年校庆'的海报,风格要喜庆、现代,Logo文字用'南开大学百年华诞',背景用浅红色。"
|
||||
# user_input_example2 = "我要一个端午节的海报,中国风,深色背景。"
|
||||
# user_input_example3 = "世界读书日海报,简约风格,元素包含打开的书和思考的人。"
|
||||
|
||||
load_config_from_file("./configs/font.yaml")
|
||||
|
||||
user_input = input("请输入您的海报需求:")
|
||||
|
||||
# 调用主接口函数
|
||||
suggestions = get_poster_content_suggestions(user_input)
|
||||
|
||||
# 打印结果
|
||||
if suggestions:
|
||||
# print("\n--- 最终输出给调用者的内容 ---")
|
||||
# print(json.dumps(suggestions, indent=2, ensure_ascii=False))
|
||||
output_folder = "outputs"
|
||||
output_file = os.path.join(output_folder, "poster_content.json")
|
||||
try:
|
||||
with open(output_file, "w", encoding="utf-8") as file:
|
||||
json.dump(suggestions, file, indent=2, ensure_ascii=False)
|
||||
print(f"\n--- 内容已成功保存到文件: {output_file} ---")
|
||||
except Exception as e:
|
||||
print(f"\n--- 保存到文件时发生错误: {e} ---")
|
||||
else:
|
||||
print("未生成任何内容,无法保存到文件。")
|
73
serve.py
73
serve.py
@ -1,73 +0,0 @@
|
||||
# server.py
|
||||
|
||||
from flask import Flask, request, jsonify
|
||||
from flask_cors import CORS
|
||||
import requests # 用于发起HTTP请求(比如调用外部接口)
|
||||
import json
|
||||
|
||||
app = Flask(__name__)
|
||||
CORS(app) # 如果前后端分离需要跨域,启用CORS
|
||||
|
||||
@app.route('/generate', methods=['POST'])
|
||||
def generate():
|
||||
try:
|
||||
# 1. 从请求体中获取参数
|
||||
data = request.json
|
||||
prompt = data.get('prompt', '')
|
||||
width = data.get('width', 512)
|
||||
height = data.get('height', 512)
|
||||
|
||||
print(f'收到参数: prompt={prompt}, width={width}, height={height}')
|
||||
|
||||
# 2. 调用大模型API或其他逻辑,生成图片
|
||||
|
||||
images = generate_images_coze(prompt, width, height)
|
||||
|
||||
# 返回 JSON 格式数据给前端
|
||||
return jsonify({'images': images})
|
||||
|
||||
except Exception as e:
|
||||
print('图片生成出错:', e)
|
||||
return jsonify({'error': '图片生成出错'}), 500
|
||||
|
||||
|
||||
def generate_images_coze(prompt, width, height):
|
||||
print(f'正在根据 "{prompt}" 生成内容 (宽{width},高{height})...')
|
||||
|
||||
url = 'https://api.coze.cn/v3/chat'
|
||||
headers = {
|
||||
'Authorization': 'Bearer pat_7G46ritHNFFvDnbeIIzXWaoPabf8ocbHsGfnAl20te1dsdHX2aTiRKR0XTLbauHW',
|
||||
'Content-Type': 'application/json'
|
||||
}
|
||||
|
||||
data = {
|
||||
"bot_id": "7486021323127980071",
|
||||
"user_id": "123",
|
||||
"stream": True,
|
||||
"auto_save_history": True,
|
||||
"additional_messages": [
|
||||
{
|
||||
"role": "user",
|
||||
"content": prompt,
|
||||
"content_type": "text"
|
||||
}
|
||||
]
|
||||
}
|
||||
|
||||
try:
|
||||
resp = requests.post(url, headers=headers, data=json.dumps(data))
|
||||
resp.raise_for_status() # 如果响应非200,会抛异常
|
||||
|
||||
result_text = resp.text
|
||||
print('Coze API 返回结果:', result_text)
|
||||
|
||||
# 假设并没有真正返回图片链接,就先返回一个测试
|
||||
return [f"大模型返回的内容:{result_text}"]
|
||||
except Exception as e:
|
||||
print('调用 Coze API 失败:', e)
|
||||
# 出错的话就返回空列表
|
||||
return []
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
app.run(host='0.0.0.0', port=9000, debug=False)
|
Loading…
Reference in New Issue
Block a user