505 lines
19 KiB
Python
505 lines
19 KiB
Python
#!/usr/bin/env python3
|
||
"""MAOMOMO article image generation CLI.
|
||
|
||
This is a lightweight OpenAI-compatible fallback for article assets. Prefer the
|
||
agent's built-in image tool when it is available; use this script when the user
|
||
explicitly wants API/CLI generation or the built-in backend is unavailable.
|
||
"""
|
||
|
||
from __future__ import annotations
|
||
|
||
import argparse
|
||
import base64
|
||
import json
|
||
import os
|
||
from pathlib import Path
|
||
import sys
|
||
import time
|
||
from typing import Any, Dict, Iterable, List, Optional
|
||
import urllib.error
|
||
import urllib.request
|
||
|
||
|
||
DEFAULT_BASE_URL = "https://api.openai.com/v1"
|
||
DEFAULT_MODEL = "gpt-image-2"
|
||
DEFAULT_SIZE = "1536x1024"
|
||
DEFAULT_QUALITY = "medium"
|
||
DEFAULT_STYLE = "warm-fintech-guide"
|
||
|
||
|
||
STYLE_PRESETS: Dict[str, str] = {
|
||
"warm-fintech-guide": (
|
||
"明亮白底，暖橙强调色，干净卡片布局，友好的白橘猫作为向导，"
|
||
"适合港卡、银行活动、返现攻略和教程总览。"
|
||
),
|
||
"clean-editorial": (
|
||
"大标题、留白充足、少量暖橙和黑灰文字，像一张信息密度适中的中文攻略头图。"
|
||
),
|
||
"data-card-dashboard": (
|
||
"指标卡、时间线、计算公式和对比表清晰分区，适合返现、费用、门槛和路径对比。"
|
||
),
|
||
"handdrawn-note": (
|
||
"白底纸感、手绘箭头、便利贴、重点圈注和轻量猫咪贴纸，适合避坑经验和保姆式步骤。"
|
||
),
|
||
"xiaohongshu-vertical": (
|
||
"3:4 或 9:16 原生竖版，手机端可读，大字少字，强标题钩子，多图系列一致视觉，"
|
||
"默认 2D 扁平信息卡片，用户确认后可用 3D 轻拟物风，带 MAOMOMO 标识。"
|
||
),
|
||
"clean-professional": (
|
||
"浅色背景、蓝绿或暖橙强调、结构化信息卡片和清晰层级，适合正式银行规则、开户教程和综合攻略。"
|
||
),
|
||
"creative-magazine": (
|
||
"大标题、强留白、编辑部排版和轻视觉冲击，适合传播型封面、观点总结和活动盘点。"
|
||
),
|
||
"retro-flat-illustration": (
|
||
"低饱和暖色、扁平金融小物件、轻复古海报感，适合轻松羊毛攻略和经验分享。"
|
||
),
|
||
"e-ink-editorial": (
|
||
"纸感背景、黑白灰为主、少量强调色、强标题和元信息条，适合深度解释和观点型长文。"
|
||
),
|
||
"scientific-defense": (
|
||
"严谨浅色版式、证据图、流程框和来源标注，适合规则拆解、政策解读和多来源核对。"
|
||
),
|
||
"mckinsey-brief": (
|
||
"结论先行、矩阵、2x2、瀑布图和高对比商业配色，适合方案对比、路径选择和决策建议。"
|
||
),
|
||
}
|
||
|
||
|
||
def die(message: str, code: int = 1) -> None:
|
||
print(f"错误：{message}", file=sys.stderr)
|
||
raise SystemExit(code)
|
||
|
||
|
||
def warn(message: str) -> None:
|
||
print(f"警告：{message}", file=sys.stderr)
|
||
|
||
|
||
def read_text(path: Path) -> str:
|
||
return path.read_text(encoding="utf-8")
|
||
|
||
|
||
def write_text(path: Path, text: str) -> None:
|
||
path.parent.mkdir(parents=True, exist_ok=True)
|
||
with path.open("w", encoding="utf-8", newline="\n") as handle:
|
||
handle.write(text)
|
||
|
||
|
||
def read_json(path: Path) -> Any:
|
||
return json.loads(read_text(path))
|
||
|
||
|
||
def strip_toml_comment(line: str) -> str:
|
||
quote = ""
|
||
escaped = False
|
||
for index, char in enumerate(line):
|
||
if escaped:
|
||
escaped = False
|
||
continue
|
||
if quote and char == "\\":
|
||
escaped = True
|
||
continue
|
||
if char in ("'", '"'):
|
||
if not quote:
|
||
quote = char
|
||
elif quote == char:
|
||
quote = ""
|
||
continue
|
||
if char == "#" and not quote:
|
||
return line[:index]
|
||
return line
|
||
|
||
|
||
def parse_toml_scalar(value: str) -> Any:
|
||
raw = strip_toml_comment(value).strip()
|
||
if len(raw) >= 2 and raw[0] == raw[-1] and raw[0] in ("'", '"'):
|
||
inner = raw[1:-1]
|
||
if raw[0] == '"':
|
||
return inner.replace(r"\"", '"').replace(r"\\", "\\")
|
||
return inner
|
||
if raw.lower() == "true":
|
||
return True
|
||
if raw.lower() == "false":
|
||
return False
|
||
return raw
|
||
|
||
|
||
def parse_basic_toml(text: str) -> Dict[str, Any]:
|
||
"""解析 Codex 配置需要的 TOML 子集，避免 Python 3.10 及以下缺少 tomllib。"""
|
||
data: Dict[str, Any] = {}
|
||
current: Dict[str, Any] = data
|
||
for raw_line in text.splitlines():
|
||
line = strip_toml_comment(raw_line).strip()
|
||
if not line:
|
||
continue
|
||
if line.startswith("[") and line.endswith("]"):
|
||
section = line.strip("[]").strip()
|
||
current = data
|
||
for part in section.split("."):
|
||
key = part.strip().strip("'\"")
|
||
if not key:
|
||
continue
|
||
child = current.get(key)
|
||
if not isinstance(child, dict):
|
||
child = {}
|
||
current[key] = child
|
||
current = child
|
||
continue
|
||
if "=" not in line:
|
||
continue
|
||
key, value = line.split("=", 1)
|
||
current[key.strip().strip("'\"")] = parse_toml_scalar(value)
|
||
return data
|
||
|
||
|
||
def load_prompt_file(path: Path) -> str:
|
||
text = read_text(path).strip()
|
||
stripped = text.lstrip()
|
||
if path.suffix.lower() == ".json" or stripped.startswith("{"):
|
||
try:
|
||
data = json.loads(text)
|
||
except json.JSONDecodeError as exc:
|
||
die(f"--prompt-file 看起来是 JSON，但解析失败：{path}: {exc}")
|
||
if not isinstance(data, dict) or not isinstance(data.get("prompt"), str) or not data["prompt"].strip():
|
||
die("--prompt-file 不能传入缺少 prompt 字段的 JSON。请先抽取 prompt 字段生成 .prompt.txt。")
|
||
prompt = data["prompt"].strip()
|
||
prompt_out = path.with_suffix(".prompt.txt")
|
||
write_text(prompt_out, prompt + "\n")
|
||
warn(f"--prompt-file 收到 JSON，已抽取 prompt 字段到：{prompt_out}；后续请直接传入该 .prompt.txt。")
|
||
return prompt
|
||
return text
|
||
|
||
|
||
def load_codex_config_base_url() -> Optional[str]:
|
||
config_path = Path.home() / ".codex" / "config.toml"
|
||
if not config_path.exists():
|
||
return None
|
||
text = read_text(config_path)
|
||
try:
|
||
import tomllib
|
||
data = tomllib.loads(text)
|
||
except ModuleNotFoundError:
|
||
data = parse_basic_toml(text)
|
||
except Exception:
|
||
data = parse_basic_toml(text)
|
||
try:
|
||
if not isinstance(data, dict):
|
||
data = parse_basic_toml(text)
|
||
except Exception:
|
||
data = parse_basic_toml(text)
|
||
|
||
provider_name = data.get("model_provider")
|
||
providers = data.get("model_providers")
|
||
if isinstance(provider_name, str) and isinstance(providers, dict):
|
||
provider = providers.get(provider_name)
|
||
if isinstance(provider, dict) and isinstance(provider.get("base_url"), str):
|
||
return provider["base_url"].strip()
|
||
|
||
if isinstance(data.get("base_url"), str):
|
||
return data["base_url"].strip()
|
||
if isinstance(providers, dict):
|
||
for provider in providers.values():
|
||
if isinstance(provider, dict) and isinstance(provider.get("base_url"), str):
|
||
return provider["base_url"].strip()
|
||
return None
|
||
|
||
|
||
def find_secret(value: Any) -> Optional[str]:
|
||
if isinstance(value, dict):
|
||
for key in ("MAOMOMO_IMAGE_API_KEY", "OPENAI_API_KEY", "api_key", "openai_api_key", "token"):
|
||
raw = value.get(key)
|
||
if isinstance(raw, str) and raw.strip():
|
||
return raw.strip()
|
||
for child in value.values():
|
||
found = find_secret(child)
|
||
if found:
|
||
return found
|
||
if isinstance(value, list):
|
||
for child in value:
|
||
found = find_secret(child)
|
||
if found:
|
||
return found
|
||
return None
|
||
|
||
|
||
def load_codex_auth_api_key() -> Optional[str]:
|
||
auth_path = Path.home() / ".codex" / "auth.json"
|
||
if not auth_path.exists():
|
||
return None
|
||
try:
|
||
return find_secret(read_json(auth_path))
|
||
except Exception:
|
||
return None
|
||
|
||
|
||
def api_key() -> str:
|
||
value = os.getenv("MAOMOMO_IMAGE_API_KEY") or os.getenv("OPENAI_API_KEY") or load_codex_auth_api_key()
|
||
if not value:
|
||
die(
|
||
"未找到 API Key。请设置 MAOMOMO_IMAGE_API_KEY 或 OPENAI_API_KEY，"
|
||
"或确保 ~/.codex/auth.json 中存在可用 key。"
|
||
)
|
||
return value
|
||
|
||
|
||
def base_url() -> str:
|
||
return (
|
||
os.getenv("MAOMOMO_IMAGE_BASE_URL")
|
||
or os.getenv("OPENAI_BASE_URL")
|
||
or load_codex_config_base_url()
|
||
or DEFAULT_BASE_URL
|
||
).rstrip("/")
|
||
|
||
|
||
def image_model() -> str:
|
||
return os.getenv("MAOMOMO_IMAGE_MODEL") or os.getenv("CODEX_PPT_IMAGE_MODEL") or DEFAULT_MODEL
|
||
|
||
|
||
def style_prompt(style: str) -> str:
|
||
if style in STYLE_PRESETS:
|
||
return STYLE_PRESETS[style]
|
||
return style
|
||
|
||
|
||
def build_prompt(
|
||
*,
|
||
title: str,
|
||
image_type: str,
|
||
core_text: str,
|
||
style: str,
|
||
aspect_ratio: str,
|
||
visual_direction: str = "",
|
||
portrait_rules: str = "",
|
||
extra: str = "",
|
||
) -> str:
|
||
allowed_visible = [value for value in (title, core_text, "MAOMOMO") if value.strip()]
|
||
allowed_visible_text = "、".join(f"「{value}」" for value in allowed_visible)
|
||
parts = [
|
||
"生成一张 MAOMOMO 中文金融实操文章配图。",
|
||
f"图片类型：{image_type}",
|
||
f"标题 / 主题：{title}",
|
||
f"核心文案：{core_text}",
|
||
f"画幅：{aspect_ratio}",
|
||
f"画面呈现：{style_prompt(style)}",
|
||
f"可见文字限制：画面只允许出现这些指定业务文案：{allowed_visible_text}；如补充要求明确列出已核对的品牌或支付标识，可作为小型辅助标识出现；不要添加副标题、风格标签、prompt 描述、说明文字、额外 slogan 或无关文字。",
|
||
"硬性要求：包含清晰可见的 MAOMOMO 主标识；中文文字清楚可读；信息层级明确；港币金额统一使用 HKD 写法；不要伪造真实 App 截图；Mastercard、Visa、UnionPay、银行、支付宝等 Logo 或支付网络标识只有在官方材料、用户素材或补充要求明确核对正确时才可出现，并且只能作为辅助识别，不能抢占或替代 MAOMOMO 主标识位置；未确认时使用通用卡片或通用支付符号，不要凭空加入或替换卡组织；不要出现真实个人信息；猫咪不穿戴第三方 Logo。",
|
||
]
|
||
if visual_direction.strip():
|
||
parts.append(f"确认视觉方向：{visual_direction.strip()}。整组图片必须保持同一视觉方向，不得混用 2D 和 3D。")
|
||
if style == "xiaohongshu-vertical" or aspect_ratio in {"3:4", "9:16"}:
|
||
rules = portrait_rules.strip() or (
|
||
"native portrait composition；no squeezed elements；"
|
||
"no stretched card / logo / text；vertical layout redesigned for Xiaohongshu；"
|
||
"禁止把横版设计硬塞进竖版画幅。"
|
||
)
|
||
parts.append(f"小红书竖图要求：{rules}")
|
||
if extra.strip():
|
||
parts.append(f"补充要求：{extra.strip()}")
|
||
return "\n".join(parts)
|
||
|
||
|
||
def request_image(prompt: str, *, size: str, quality: str, output_format: str) -> bytes:
|
||
payload = {
|
||
"model": image_model(),
|
||
"prompt": prompt,
|
||
"size": size,
|
||
"quality": quality,
|
||
"n": 1,
|
||
"response_format": "b64_json",
|
||
}
|
||
if output_format:
|
||
payload["output_format"] = output_format
|
||
|
||
body = json.dumps(payload, ensure_ascii=False).encode("utf-8")
|
||
req = urllib.request.Request(
|
||
base_url() + "/images/generations",
|
||
data=body,
|
||
headers={
|
||
"Authorization": f"Bearer {api_key()}",
|
||
"Content-Type": "application/json",
|
||
},
|
||
method="POST",
|
||
)
|
||
|
||
try:
|
||
with urllib.request.urlopen(req, timeout=300) as resp:
|
||
data = json.loads(resp.read().decode("utf-8"))
|
||
except urllib.error.HTTPError as exc:
|
||
detail = exc.read().decode("utf-8", "replace")
|
||
die(f"图片接口返回 HTTP {exc.code}: {detail}")
|
||
except Exception as exc:
|
||
die(f"图片接口请求失败：{exc}")
|
||
|
||
items = data.get("data")
|
||
if not isinstance(items, list) or not items:
|
||
die("图片接口响应缺少 data。")
|
||
first = items[0]
|
||
if not isinstance(first, dict):
|
||
die("图片接口响应格式不正确。")
|
||
if isinstance(first.get("b64_json"), str):
|
||
return base64.b64decode(first["b64_json"])
|
||
if isinstance(first.get("url"), str):
|
||
with urllib.request.urlopen(first["url"], timeout=300) as resp:
|
||
return resp.read()
|
||
die("图片接口响应中没有 b64_json 或 url。")
|
||
return b""
|
||
|
||
|
||
def normalize_output(path: str, *, base_dir: Optional[Path]) -> Path:
|
||
out = Path(path)
|
||
if not out.is_absolute() and base_dir is not None:
|
||
out = base_dir / out
|
||
return out.resolve()
|
||
|
||
|
||
def generate_one(args: argparse.Namespace) -> Dict[str, str]:
|
||
if args.prompt_file:
|
||
prompt = load_prompt_file(Path(args.prompt_file))
|
||
elif args.prompt:
|
||
prompt = args.prompt.strip()
|
||
else:
|
||
prompt = build_prompt(
|
||
title=args.title,
|
||
image_type=args.image_type,
|
||
core_text=args.core_text,
|
||
style=args.style,
|
||
aspect_ratio=args.aspect_ratio,
|
||
visual_direction=args.visual_direction,
|
||
portrait_rules=args.portrait_rules,
|
||
extra=args.extra,
|
||
)
|
||
|
||
out = normalize_output(args.out, base_dir=None)
|
||
if args.dry_run:
|
||
write_text(out.with_suffix(".prompt.txt"), prompt + "\n")
|
||
return {"out": str(out), "prompt": str(out.with_suffix(".prompt.txt")), "status": "dry-run"}
|
||
|
||
image_bytes = request_image(prompt, size=args.size, quality=args.quality, output_format=args.output_format)
|
||
out.parent.mkdir(parents=True, exist_ok=True)
|
||
out.write_bytes(image_bytes)
|
||
return {"out": str(out), "status": "generated"}
|
||
|
||
|
||
def manifest_items(path: Path) -> List[Dict[str, Any]]:
|
||
data = read_json(path)
|
||
if isinstance(data, dict):
|
||
items = data.get("images")
|
||
else:
|
||
items = data
|
||
if not isinstance(items, list):
|
||
die("manifest 必须是图片数组，或包含 images 数组的对象。")
|
||
normalized: List[Dict[str, Any]] = []
|
||
for index, item in enumerate(items, start=1):
|
||
if not isinstance(item, dict):
|
||
die(f"manifest 第 {index} 项不是对象。")
|
||
normalized.append(item)
|
||
return normalized
|
||
|
||
|
||
def batch(args: argparse.Namespace) -> List[Dict[str, str]]:
|
||
manifest_path = Path(args.manifest).resolve()
|
||
base_dir = Path(args.base_dir).resolve() if args.base_dir else manifest_path.parent
|
||
results: List[Dict[str, str]] = []
|
||
|
||
for index, item in enumerate(manifest_items(manifest_path), start=1):
|
||
filename = item.get("file_name") or item.get("filename") or item.get("out")
|
||
if not isinstance(filename, str) or not filename.strip():
|
||
die(f"manifest 第 {index} 项缺少 file_name。")
|
||
style = str(item.get("style") or args.style or DEFAULT_STYLE)
|
||
aspect_ratio = str(item.get("aspect_ratio") or args.aspect_ratio)
|
||
visual_direction = str(item.get("visual_direction") or args.visual_direction or "")
|
||
raw_portrait_rules = item.get("portrait_composition_rules") or args.portrait_rules
|
||
if isinstance(raw_portrait_rules, list):
|
||
portrait_rules = "；".join(str(value) for value in raw_portrait_rules)
|
||
else:
|
||
portrait_rules = str(raw_portrait_rules or "")
|
||
prompt = item.get("prompt")
|
||
if not isinstance(prompt, str) or not prompt.strip():
|
||
prompt = build_prompt(
|
||
title=str(item.get("title") or item.get("alt_text") or filename),
|
||
image_type=str(item.get("type") or "文章配图"),
|
||
core_text=str(item.get("core_text") or item.get("caption") or item.get("alt_text") or ""),
|
||
style=style,
|
||
aspect_ratio=aspect_ratio,
|
||
visual_direction=visual_direction,
|
||
portrait_rules=portrait_rules,
|
||
extra=str(item.get("extra") or item.get("requirements") or ""),
|
||
)
|
||
|
||
out = normalize_output(filename, base_dir=base_dir)
|
||
if args.dry_run:
|
||
write_text(out.with_suffix(".prompt.txt"), prompt + "\n")
|
||
results.append({"out": str(out), "prompt": str(out.with_suffix(".prompt.txt")), "status": "dry-run"})
|
||
continue
|
||
|
||
image_bytes = request_image(
|
||
prompt,
|
||
size=str(item.get("size") or args.size),
|
||
quality=str(item.get("quality") or args.quality),
|
||
output_format=str(item.get("output_format") or args.output_format),
|
||
)
|
||
out.parent.mkdir(parents=True, exist_ok=True)
|
||
out.write_bytes(image_bytes)
|
||
results.append({"out": str(out), "status": "generated"})
|
||
if args.sleep > 0:
|
||
time.sleep(args.sleep)
|
||
return results
|
||
|
||
|
||
def print_styles(_: argparse.Namespace) -> int:
|
||
print(json.dumps(STYLE_PRESETS, ensure_ascii=False, indent=2))
|
||
return 0
|
||
|
||
|
||
def build_parser() -> argparse.ArgumentParser:
|
||
parser = argparse.ArgumentParser(description="Generate MAOMOMO article images with an OpenAI-compatible API.")
|
||
subparsers = parser.add_subparsers(dest="command", required=True)
|
||
|
||
common = argparse.ArgumentParser(add_help=False)
|
||
common.add_argument("--style", default=DEFAULT_STYLE, help="Style preset name or custom style text.")
|
||
common.add_argument("--size", default=DEFAULT_SIZE)
|
||
common.add_argument("--quality", default=DEFAULT_QUALITY)
|
||
common.add_argument("--output-format", default="png")
|
||
common.add_argument("--aspect-ratio", default="16:9")
|
||
common.add_argument("--visual-direction", default="")
|
||
common.add_argument("--portrait-rules", default="")
|
||
common.add_argument("--dry-run", action="store_true", help="Only write prompt files; do not call the image API.")
|
||
|
||
generate = subparsers.add_parser("generate", parents=[common], help="Generate one image.")
|
||
generate.add_argument("--out", required=True)
|
||
generate.add_argument("--prompt")
|
||
generate.add_argument("--prompt-file")
|
||
generate.add_argument("--title", default="MAOMOMO 文章配图")
|
||
generate.add_argument("--image-type", default="文章配图")
|
||
generate.add_argument("--core-text", default="")
|
||
generate.add_argument("--extra", default="")
|
||
|
||
batch_parser = subparsers.add_parser("batch", parents=[common], help="Generate images from a JSON manifest.")
|
||
batch_parser.add_argument("--manifest", required=True)
|
||
batch_parser.add_argument("--base-dir", help="Base directory for relative file_name paths. Defaults to manifest dir.")
|
||
batch_parser.add_argument("--sleep", type=float, default=0.0, help="Seconds to sleep between API calls.")
|
||
|
||
subparsers.add_parser("styles", help="Print bundled style presets.")
|
||
return parser
|
||
|
||
|
||
def main(argv: Optional[Iterable[str]] = None) -> int:
|
||
parser = build_parser()
|
||
args = parser.parse_args(list(argv) if argv is not None else None)
|
||
if args.command == "styles":
|
||
return print_styles(args)
|
||
if args.command == "generate":
|
||
result = generate_one(args)
|
||
print(json.dumps(result, ensure_ascii=False, indent=2))
|
||
return 0
|
||
if args.command == "batch":
|
||
results = batch(args)
|
||
print(json.dumps(results, ensure_ascii=False, indent=2))
|
||
return 0
|
||
parser.print_help()
|
||
return 1
|
||
|
||
|
||
if __name__ == "__main__":
|
||
raise SystemExit(main())
|